A method for modeling and estimating failure of MOS field effect transistor considering quality consistency

Abstract

The application discloses a MOS field effect transistor failure physical modeling and estimation method considering quality consistency, and belongs to the technical field of MOS field effect transistor performance and reliability analysis. In order to solve the problem that existing MOS field effect transistor failure physical modeling schemes do not consider the performance parameter degradation process and the model coefficients in the degradation model cannot reflect batch product consistency, the application firstly determines a general expression form of a failure physical model considering quality consistency according to typical failure mechanisms of the MOS field effect transistor, carries out MOS field effect transistor multi-stress accelerated test design and obtains accelerated test results, calculates failure physical model coefficients by using a multi-parameter correlation method based on Copula, obtains an electronic component failure physical model, and further determines the failure physical model considering quality consistency. Finally, estimation is carried out according to the failure physical model considering quality consistency.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of MOS field effect transistor performance and reliability analysis, and relates to a MOS field effect transistor failure physical modeling method. BACKGROUND

[0002] MOS field effect transistors are widely used in various power electronic converters. With the continuous improvement of application requirements and the rapid development of process technology, the performance of MOS field effect transistors is getting better and better. However, in actual application, MOS field effect transistors are affected by electricity, heat and various complex environmental stresses, and have a high failure rate. Relevant statistics show that among all the failures of power electronic converters, more than 1 / 3 are caused by semiconductor power devices. It is of great significance and application value to study the mechanisms of various failures of MOS field effect transistors, establish failure physical models for main failure mechanisms, and apply them to the reliability analysis of MOS field effect transistors or power electronic converters.

[0003] At present, stage achievements have been made in MOS field effect transistor failure analysis, accelerated degradation test and failure physical modeling: the main failure mechanisms of MOS field effect transistors such as hot carrier injection, gate oxide layer time-dependent breakdown, single particle gate penetration and bias temperature instability have been widely studied, the occurrence process of various failure mechanisms has been explained through simulation, test and other ways, and they have been used for actual evaluation; based on the performance parameter degradation data obtained from accelerated degradation test, combined with failure mechanisms, failure physical models are established, and on this basis, the reliability of devices is further evaluated.

[0004] However, there are still some problems in the current research: (1) Most of the failure physical models established for MOS field effect transistors are life models, which only focus on the relationship between the life (failure time) of the device and the stress, and do not focus on the degradation process of the performance parameters of the device, ignoring the reliability information implied in the degradation process of the device; (2) There is still a blank in the failure physical modeling of MOS field effect transistors considering quality consistency. For domestic MOS field effect transistors, how to consider individual differences in the failure physical model and integrate quality consistency information needs to be solved. Therefore, combined with the structural characteristics and failure mechanisms of MOS field effect transistors, the form of the degradation model function is determined, and based on the manufacturing process data of MOS field effect transistors, the estimation of the distribution of random influence coefficients in the MOS field effect transistor degradation model is realized through the maximum likelihood method, so as to construct a degradation model integrating quality consistency information and lay a foundation for further accurate evaluation of the reliability level of MOS field effect transistors. SUMMARY

[0005] The application provides a MOS field effect tube failure physical modeling and estimation method considering quality consistency.

[0006] A MOS field effect tube failure physical modeling method considering quality consistency comprises the following steps:

[0007] Step one: according to typical failure mechanisms of the MOS field effect tube, a general expression form of the failure physical model considering quality consistency is determined;

[0008] The failure physical model considering quality consistency comprises a performance degradation type failure physical model and a functional failure type failure physical model; the performance degradation type failure physical model coefficient represents all model coefficients in the corresponding model, and model quality consistency is represented by an initial value Y0 of a characteristic output and a degradation rate A of the performance degradation type failure physical model D obeys a two-dimensional normal distribution, wherein μ A represents a mean value of the degradation rate A D , and σ A represents a standard deviation of the degradation rate A D ; the functional failure type failure physical model coefficient represents all model coefficients in the corresponding model, and model quality consistency information is represented by a rate coefficient A F of a function F(·) obeying a normal distribution;

[0009] Step two: a MOS field effect tube multi-stress accelerated test design is carried out in combination with the general expression form of the failure physical model considering quality consistency constructed in step one;

[0010] Step three: an accelerated test result is obtained according to the accelerated test design in step two, a failure physical model coefficient is calculated by using a multi-parameter correlation method based on Copula, an electronic component failure physical model is obtained, and then the failure physical model considering quality consistency is determined.

[0011] Further, the electronic component characteristic output of the performance degradation type failure physical model is as follows:

[0012] Y(t) = Y0 + D(S, t) (1)

[0013] wherein Y(t) represents an electronic component characteristic output at t, Y0 represents an initial value of a characteristic output, D(·) represents a performance degradation type failure physical model, S represents a stress type set, and t represents time;

[0014] Model quality consistency of the performance degradation type failure physical model is represented by Y0 and a degradation rate A Dobeys two-dimensional normal distribution, i.e. satisfies formula (2):

[0015]

[0016] In the formula, μ0 represents the mean of Y0, σ0 represents the standard deviation of Y0, μ A represents the mean of the degradation rate A D , σ A represents the standard deviation of the degradation rate A D , and ρ represents a correlation coefficient.

[0017] The functional failure type failure physical model is constructed as follows:

[0018] τ = F (s) (3)

[0019] Wherein, τ represents the life of an electronic component, S represents a stress combination, and F(·) represents a failure mechanism for a functional failure type failure physical model.

[0020] The quality consistency information of the functional failure type failure physical model is represented as the rate coefficient A F obeys a normal distribution, i.e. satisfies

[0021]

[0022] Wherein, μ F and σ F respectively represent the mean and the standard deviation of A F of the functional failure type failure physical model.

[0023] Further, the specific process of designing a multi-stress accelerated test of a MOS field effect transistor includes:

[0024] Selecting a MOS field effect transistor accelerated test limit stress and group number: the selected maximum stress should not exceed the proportional threshold of the device limit stress, and the selected minimum stress is the device limit rated working condition.

[0025] The stress level in the middle is determined by the reciprocal stress equal interval formula, as shown in formula (4):

[0026]

[0027] Wherein, Δ is the reciprocal stress interval; l is the number of stress selected; S1 is the rated stress level; S Max is the limit stress level.

[0028] Selecting a MOS field effect transistor accelerated test sample: based on the mean and the standard deviation of the sample life, the MOS field effect transistor accelerated test sample is selected according to the confidence interval.

[0029] Determine MOS field effect tube acceleration test test time interval: based on MOS field effect tube quality consistency determines MOS field effect tube acceleration test test time interval.

[0030] Further, the device limit stress ratio threshold is 90%.

[0031] Further, based on the sample life mean and standard deviation, the MOS field effect tube acceleration test sample selection process according to the confidence interval includes:

[0032] First, the degradation data is determined according to the life data of the failure threshold or the same representation of the electronic component life data as the life data, assuming that the life mean is μ and the standard deviation is σ, then the average life true value of the electronic component can be determined as μ;

[0033] Second, extract n samples from all life data, n samples present normal distribution with mean μ and variance σ 2 / n; using the distribution of the average value of the above n samples, the confidence interval of the confidence level P is calculated by formula (5);

[0034]

[0035] Where z is the standard score corresponding to the corresponding confidence level;

[0036] Finally, according to the confidence interval of the confidence level P obtained, the error error between the upper and lower limits of the confidence interval and the true life is calculated by formula (6);

[0037]

[0038] Based on error, the confidence level P average life sampling estimation limit relative error under different quality consistency is generated, and then the MOS field effect tube acceleration test sample is selected.

[0039] Further, the process of determining the MOS field effect tube acceleration test test time interval based on the MOS field effect tube quality consistency includes the following steps:

[0040] Statistical initial performance parameter or life data mean and standard deviation, characteristic output initial value mean and standard deviation are denoted as μ0 and σ0, and life mean and standard deviation are denoted as μ τ 、σ τ ;

[0041] If σ0 / μ0≤characteristic related coefficient of variation threshold, or σ τ / μ≤life related coefficient of variation threshold, the test interval gradually increases;

[0042] If σ0 / μ0>characteristic related coefficient of variation threshold, or στ If the life-related coefficient of variation threshold is not reached, the test interval should be gradually reduced.

[0043] Further, the characteristic-related coefficient of variation threshold is 0.15.

[0044] Further, the life-related coefficient of variation threshold is 0.15.

[0045] Further, the process of calculating the failure physical model coefficient comprises:

[0046] Suppose a set of failure physical model coefficients θ=(θ1, θ2, …, θ n , the test data obeys a normal distribution; a likelihood function L(θ) is given, which represents the probability of the observed data under the given parameter θ, and the calculation formula is shown in equation (7):

[0047]

[0048] In the formula, x i represents the degradation or life data of the sample size, represents the joint probability distribution, and L(θ) represents the likelihood function.

[0049] Taking the logarithm of the likelihood function, the log-likelihood function is obtained:

[0050]

[0051] Based on the optimization algorithm, the optimal parameter is found, and the parameter that maximizes the log-likelihood function logL(θ) is calculated

[0052]

[0053] Where arg max represents the value of the variable that makes the logL(θ) reach the maximum value.

[0054] According to the estimated parameter , the failure physical model parameter θ=(θ1, θ2, …, θ n ) is obtained.

[0055] A MOS field effect transistor failure estimation method considering quality consistency, for a MOS field effect transistor, first, a MOS field effect transistor failure physical modeling method considering quality consistency is used for modeling, to obtain the MOS field effect transistor corresponding to the MOS field effect transistor failure model considering quality consistency, and then according to the MOS field effect transistor failure model considering quality consistency, the MOS field effect transistor is estimated.

[0056] Advantages:

[0057] In the process of failure physical modeling of MOS field effect tube, the failure physical model expression form considering quality consistency is constructed, the difference between MOS field effect tube individuals can be more accurately represented, the degradation distribution of MOS field effect performance parameters under specific stress can be more accurately represented, and the reliability prediction of MOS field effect tube and power electronic converter has important significance. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 The figure is a flow chart of the MOS field effect tube failure physical modeling method considering quality consistency.

[0059] Figure 2 The upper and lower limits of the 90% confidence interval corresponding to model A.

[0060] Figure 3 The upper and lower limits of the 90% confidence interval corresponding to model B.

[0061] Figure 4 The upper and lower limits of the 90% confidence interval corresponding to model C.

[0062] Figure 5 The upper and lower limits of the 90% confidence interval corresponding to model D.

[0063] Figure 6 V is the V of model A corresponding to the 90% confidence interval. th Degradation model verification result.

[0064] Figure 7 V is the V of model B corresponding to the 90% confidence interval. th Degradation model verification result.

[0065] Figure 8 V is the V of model C corresponding to the 90% confidence interval. th Degradation model verification result.

[0066] Figure 9 V is the V of model D corresponding to the 90% confidence interval. th Degradation model verification result. DETAILED DESCRIPTION

[0067] The technical solutions of the present application will be further described below in combination with the drawings, but are not limited thereto, any modification or equivalent replacement of the technical solutions of the present application without departing from the spirit and scope of the present application shall be covered in the protection scope of the present application.

[0068] The application firstly constructs a basic form of a failure physical model between threshold voltage and stress according to typical failure mechanisms of MOS field effect tubes, then analyzes influences of key material, structure and process vector fluctuations on model coefficients, introduces quality consistency information into the failure physical model, and constructs a failure physical model considering quality consistency. Secondly, a multi-stress accelerated test design of MOS field effect tubes is carried out. Finally, a maximum likelihood method is used for failure physical model parameter identification, model coefficients with distribution characteristics are obtained, and the failure physical modeling of MOS field effect tubes considering quality consistency is completed. The main steps include: step one: constructing a general form of a failure physical model of MOS field effect tubes according to failure mechanisms; step two: accelerated test design; step three: failure physical model parameter identification and verification (comparison without considering quality consistency + comparison only with mean value). The application is further described in combination with specific embodiments.

[0069] The embodiment is a failure physical modeling method of MOS field effect tubes considering quality consistency and a corresponding failure estimation method. Firstly, the application is combined with Figure 1 The application is a failure physical modeling method of MOS field effect tubes considering quality consistency, which includes the following steps:

[0070] Step one: according to typical failure mechanisms of MOS field effect tubes, a general expression form of a failure physical model considering quality consistency is determined;

[0071] According to typical failure mechanisms of MOS field effect tubes, the failure physical model considering quality consistency is specifically divided into a performance degradation type failure physical model and a functional failure type failure physical model, and specific models are determined according to actual electronic component failure mechanisms;

[0072] A, the electronic component characteristic output of the performance degradation type failure physical model is as follows:

[0073] Y(t) = Y0 + D(S, t) (1)

[0074] Wherein, Y(t) represents an electronic component characteristic output at t time, Y0 represents an initial value of the characteristic output, D(·) represents a performance degradation type failure physical model, S represents a stress type set (sensitive stress types are selected according to electronic component specific failure mechanisms), and t represents time. The performance degradation type failure physical model coefficient represents all model coefficients in D(·), and the quality consistency of the model is represented by Y0 and the degradation rate A of the function D(·) D obeys two-dimensional normal distribution, that is, formula (2) is satisfied:

[0075]

[0076] In the formula, μ0 represents a mean value of Y0, σ0 represents a standard deviation of Y0, μ Aμ D represents the mean of the degradation rate A A σ D represents the standard deviation of the degradation rate A F , the correlation coefficient 0<|p|<1, when p>0, it indicates that the quality consistency at the initial moment of the product is positively correlated with the degradation rate, when p<0, it indicates that the quality consistency at the initial moment of the product is negatively correlated with the degradation rate, when |p|>0.5, it indicates that the correlation between the two is strong, and when |p|<0.5, it indicates that the correlation between the two is weak.

[0077] B, the physical model of functional failure type failure is constructed as formula (3):

[0078] τ=F(s) (3)

[0079] Wherein, τ represents the life of electronic components, S represents the stress combination, F(·) represents the physical model of failure type failure, the coefficient of functional failure type failure physical model represents all model coefficients in F(·), and the quality consistency information of the model is represented by the rate coefficient A F of the function F(s), which is subject to normal distribution, that is μ F ,σ F respectively represent the mean and standard deviation of the rate coefficient A F of the function F(s) corresponding to the physical model of functional failure type failure; the rate coefficient of the function F(s) refers to the coefficient corresponding to the change rate of the physical model of failure;

[0080] Step two: combined with the general expression form of the failure physical model considering quality consistency constructed in step one, the multi-stress accelerated test design of MOS field effect transistor is carried out, and the test design range of this step covers performance degradation type failure mechanism and functional type failure mechanism:

[0081] Select the limit stress and group number of MOS field effect transistor accelerated test: the maximum stress selected should not exceed 90% of the limit stress of the device, and the minimum stress is the limit rated working condition of the device.

[0082] The rated stress of MOS field effect transistor is S0, the limit stress keeping the failure mechanism unchanged is S Max . The stress level in the middle is determined by the reciprocal stress interval formula, as shown in formula (4):

[0083]

[0084] Wherein, Δ is the reciprocal stress interval; l is the number of stress selection; S1 is the rated stress level; S Max is the limit stress level.

[0085] Select the MOS field effect transistor accelerated test sample:

[0086] Firstly, when there are enough electronic component life data, the life data represents electronic component life data, or represents degradation data according to failure threshold to determine life data, assuming that the life mean is μ and the standard deviation is σ, then the average life true value of the electronic component can be determined as μ.

[0087] Secondly, n samples are extracted from all life data, according to the central limit theorem, the average values of the n samples will be around the population sample average μ, and present a normal distribution with mean μ and variance σ 2 / n.

[0088] Then, the confidence interval of the confidence level P is calculated by formula (5) using the distribution of the n sample average values.

[0089]

[0090] Where z is the standard score corresponding to the corresponding confidence level, which can be obtained by looking up the table, when the sample size n is less than 30, z is obtained by looking up the t distribution table; when the sample size n is not less than 30, z is obtained by looking up the standard normal distribution table.

[0091] Finally, according to the confidence interval of the confidence level P obtained, the error error between the upper and lower limits of the confidence interval and the true life can be calculated by formula (6), and the relative error of the confidence level P average life sampling estimation limit under different quality consistency is generated.

[0092]

[0093] In the formula, μ is the actual mean of life, σ 2 / n is the actual standard deviation, z is the standard score corresponding to the corresponding confidence level, and n is the number of samples used for verification. If the error value is very small, it means that the boundaries of the interval are very close to the average life true value, which means that the average value of the sampling sample size n is very close to the average life true value with a probability P. Therefore, in the case that the error value is very small, the average value distribution of n samples can be used to represent the true average life distribution of electronic components. As can be seen from formula (6), the error value is related to the standard score z, the sampling sample size n, and the population sample dispersion (i.e. quality consistency) σ / μ.

[0094] Determination of MOS field effect transistor accelerated test test time interval: the mean and standard deviation of the performance parameter or life data at the initial moment are counted, and the mean and standard deviation of the initial value of the characteristic output are denoted as μ0, σ0, and the mean and standard deviation of the life are denoted as μ τ , σ τ .

[0095] If the quality consistency of MOS field effect transistor is good (σ0 / μ0≤0.15 or στ / μ τ ≤0.15), then the test interval gradually increases: Δt={Δt1<Δt2<Δt3…<Δt n-2 <Δt n-1};

[0096] If the MOS field effect transistor quality consistency is poor (σ0 / μ0>0.15 or σ τ / μ τ >0.15), then the test interval should gradually decrease: Δt={Δt1>Δt2>Δt3…>Δt n-2 >Δt n-1};

[0097] Step three, according to the accelerated test design of step two, the accelerated test results are obtained, and the multi-parameter correlation method based on Copula is used to calculate the failure physical model coefficient (the model coefficient represents the performance degradation failure physical model coefficient or the functional failure type failure physical model coefficient), and the electronic component failure physical model is obtained.

[0098] The process of calculating the failure physical model coefficient includes:

[0099] Suppose that the failure physical model coefficient set to be identified θ=(θ1, θ2, …, θ n ), the test data obeys the normal distribution. Given the likelihood function L(θ) represents the probability of the observed data under the given parameter θ, and the calculation formula is shown in equation (7):

[0100]

[0101] In the formula, x i represents the degradation or life data of the sample size, represents the joint probability distribution, and L(θ) represents the likelihood function.

[0102] In order to simplify the calculation, the logarithm of the likelihood function is usually taken to obtain the log-likelihood function:

[0103]

[0104] Based on the optimization algorithm, the optimal parameters are found, and the parameters that maximize the log-likelihood function logL(θ) are calculated

[0105]

[0106] Where, arg max represents the value of the variable when the logL(θ) reaches the maximum value.

[0107] According to the estimated parameters obtained failure physical model parameter θ = (θ1, θ2, …, θ n ) is obtained.

[0108] The failure physical modeling method considering the quality consistency of the MOS field effect tube is completed, and a failure physical model considering the quality consistency is obtained. The model can be used for failure estimation of the MOS field effect. Specifically, in actual estimation, for the MOS field effect tube, first, a failure physical modeling method considering the quality consistency of the MOS field effect tube is used for modeling to obtain a MOS field effect tube failure model considering the quality consistency of the MOS field effect tube, and then the MOS field effect tube is estimated according to the MOS field effect tube failure model considering the quality consistency of the MOS field effect tube.

[0109] Embodiment:

[0110] This embodiment takes the typical failure mechanism of the MOS field effect tube, bias temperature instability (BTI), as an example to verify the effectiveness of the proposed method, and the bias temperature instability is a performance degradation type failure physical model. The modeling ideas of the functional failure type failure physical model and the performance degradation type failure physical model are the same.

[0111] The threshold voltage V th and the failure physical model basic form of the sensitive stress V and temperature T affecting V th :

[0112]

[0113] In the formula, V th (t) is the threshold voltage at time t, corresponding to Y(t) in formula (1); V th (0) is the initial value of the threshold voltage, corresponding to Y0 in formula (1); A BTI is the BTI effect factor, B is the electric field acceleration factor, E a is the failure activation energy, k is the Boltzmann constant, and C is the time power index. A BTI obeys the normal distribution,

[0114] MOS field effect tube accelerated test sample selection: for the number of samples used in the test, the quality consistency information of the initial threshold voltage of the MOS field effect tube is 0.1 (σ0 / μ0). According to formula (5), the confidence interval under the 90% confidence level of the MOS field effect tube is calculated. According to the obtained confidence interval under the 90% confidence level, the error between the upper and lower limits of the confidence interval and the true value can be calculated by formula (6), thereby generating a 90% confidence level average life sampling estimation limit relative error lookup table, and part of the results are shown in Table 1.

[0115] Table 1 90% confidence level average life sampling estimation limit relative error lookup table

[0116]

[0117] According to the results shown in Table 1, in order to ensure that the test data is representative, at least 22 samples under the verification stress.

[0118] Determination of MOS field effect tube acceleration test stress: On this basis, the level and combination of applied stress are considered. Under the NBTI effect, the key impact stress of PMOSFET is the gate-source driving voltage and the environmental temperature. In combination with the data needs of the relevant standards, the existing test sample quantity and the subsequent modeling work, it can be determined that each stress should be divided into three levels. According to the degradation amount of the device under different stresses in the preliminary test, the degradation rate of the device can be estimated, and then the range of the applied stress is determined. Considering the above information comprehensively, the driving voltage is divided into three levels: -36V, -38V, -40V, the environmental temperature is divided into three levels: 348K, 378K, 408K, and cross combination into 5 groups

[0119] Table 2 Sample quantity statistics table under each stress of the acceleration test (specific data “ / ” indicates no sample)

[0120]

[0121] Determination of MOS field effect tube acceleration test test interval: Since the quality consistency information of the initial value of the threshold voltage of the MOS field effect tube is 0.1 (σ0 / μ0), the test interval of this acceleration test is: the early test interval is one day, the duration is 14 days, and then test once every two or three days until the threshold voltage reaches the failure threshold.

[0122] According to formula (10) and the data obtained by the acceleration test, μ A , σ A , B, E a and C are obtained by maximum likelihood estimation, wherein the threshold voltage degradation model of a certain type of MOS field effect tube is:

[0123]

[0124] 1000, 5000 and 10000 degradation trajectories are randomly generated under the verification stress, and the upper and lower limits of the 90% confidence interval are drawn, as shown in Figures 2-5 , Figures 2-5 The upper and lower limits of the 90% confidence interval corresponding to types A-D, respectively, can be seen from Figures 2-5 , and the comparison results are shown in Figures 6-9 , Figures 6-9 The V thThe results of the degradation model verification show that most of the actual degradation data are located in the confidence interval.

[0125] Assuming that the time required for 5% threshold voltage degradation is t 5％ , the average values of the predicted t 5％ and the actual t 5％ under the verification stress are calculated respectively, and the relative error of the average values is calculated, and the results are shown in Table 1.

[0126] Table 1 Comparison of predicted values and actual values

[0127]

[0128] According to the results in the table, when the mean value is the model accuracy evaluation index, the model error is less than 10%.

[0129] The above examples of the present application are only used to illustrate the calculation model and the calculation process of the present application, and are not limited to the embodiments of the present application. For those skilled in the art, other different forms of changes or variations can be made on the basis of the above description, and it is impossible to enumerate all the embodiments here, and any obvious changes or variations derived from the technical solutions of the present application still fall within the protection scope of the present application.

Claims

1. A method for physical modeling of MOS field effect transistor failure considering quality uniformity, characterized in that, The method comprises the following steps: Step 1: according to the typical failure mechanism of MOS field effect tubes, a general expression form of a failure physical model considering quality consistency is determined; The failure physical model considering the quality consistency includes a performance degradation type failure physical model and a functional failure type failure physical model; the performance degradation type failure physical model coefficient represents all model coefficients in the corresponding model, and the model quality consistency is characterized by an initial value Y0 of a characteristic output and a degradation rate A of the performance degradation type failure physical model D obeys a two-dimensional normal distribution, and represents a mean value of the degradation rate A D , represents a standard deviation of the degradation rate A D ; the functional failure type failure physical model coefficient represents all model coefficients in the corresponding model, and the model quality consistency information is characterized by a function rate coefficient A F obeys a normal distribution; The characteristic output of the performance degradation type failure physical model is as follows: (1) wherein Y(t) represents the characteristic output of the electronic component at time t, Y0 represents the initial value of the characteristic output, denotes a performance degradation type failure physical model, S denotes a set of stress types, and t denotes time. Model quality consistency characterization for performance degradation type failure physical model is Y0 and function of degradation rate A D obeys two-dimensional normal distribution, that is, satisfies formula (2):​ (2) In the formula, This represents the mean of Y0. This represents the standard deviation of Y0. Indicates the degradation rate A D The mean, Indicates the degradation rate A D standard deviation Represents the correlation coefficient; The functional failure type failure physical model is constructed as follows: (3) wherein τ represents the electronic component lifetime, S represents the stress combination, represents a failure mechanism for failure type failure physical model; The functional failure type failure physical model quality consistency information is characterized as a function Rate coefficient A F obeys a normal distribution, that is, satisfies wherein, A and A represent the mean and standard deviation of the functional failure type failure physical model, respectively F of the mean and standard deviation; Step 2: in combination with the general expression form of the failure physical model considering quality consistency constructed in step 1, a MOS field effect tube multi-stress accelerated test is designed; Step 3: according to the accelerated test result obtained in step 2, a failure physical model coefficient is calculated by using a multi-parameter correlation method based on Copula, and a failure physical model of the electronic component is obtained, and then the failure physical model considering quality consistency is determined.

2. The method of claim 1, wherein the method further comprises: The specific process of designing the MOS field effect tube multi-stress accelerated test comprises the following steps: The limit stress and the number of groups of the MOS field effect tube accelerated test are selected: the maximum stress selected should not exceed the proportion threshold of the limit stress of the device, and the minimum stress selected is the limit rated working condition of the device; The stress levels in the middle are determined by using a stress reciprocal equal interval formula, as shown in formula (4): (4) wherein, is the stress reciprocal interval; / is the stress selection number; S1is the rated stress level; S Max is the limit stress level; The MOS field effect tube accelerated test sample is selected: based on the mean and standard deviation of the sample life, the MOS field effect tube accelerated test sample is selected according to the confidence interval; The MOS field effect tube accelerated test time interval is determined: the MOS field effect tube accelerated test time interval is determined based on the quality consistency of the MOS field effect tube.

3. The method of claim 2, wherein the method further comprises: The proportion threshold of the limit stress of the device is 90%.

4. The method of claim 2, wherein the method further comprises: The process of selecting the MOS field effect tube accelerated test sample based on the mean and standard deviation of the sample life and according to the confidence interval comprises the following steps: Firstly, the degradation data is determined as the life data according to the life data determined by the failure threshold or the same representation of the electronic component life data, assuming that the mean is μ and the standard deviation is σ, then the average life true value of the electronic component can be determined as μ; Second, n samples are extracted from the entire life data, and the n samples are shown to obey a normal distribution with a mean of μ and a variance of The confidence interval of the confidence level P is calculated by using the distribution of the mean of the n samples and by using Equation (5). (5) Wherein, z is the standard score corresponding to the corresponding confidence level; Finally, the error error between the upper and lower limits of the confidence interval and the true life is calculated by formula (6) according to the confidence interval of the obtained confidence level P; (6) Based on error, the confidence level P average life sampling estimation limit relative error under different quality consistency is generated, and then the MOS field effect tube accelerated test sample is selected.

5. The method of claim 2, wherein the method further comprises: The process of determining the MOS field effect tube accelerated test time interval based on the quality consistency of the MOS field effect tube comprises the following steps: Statistical initial time performance parameter or life data mean and standard deviation, characteristic output initial value mean and standard deviation is recorded as , , life mean and standard deviation is recorded as , ; If a characteristic-related coefficient of variation threshold, or a life-related coefficient of variation threshold, the test interval is gradually increased; If a characteristic-related coefficient of variation threshold, or a life-related coefficient of variation threshold, the test interval should be gradually reduced.

6. The method of claim 5, wherein the method further comprises: The threshold of the coefficient of variation related to the characteristics is 0.

15.

7. The method of claim 5, wherein the method further comprises: The threshold of the coefficient of variation related to the life is 0.

15.

8. The method according to any one of claims 1 to 7, wherein the method is a method for modeling failure physics of a MOS field effect transistor considering quality uniformity, characterized in that, The process of calculating the failure physical model coefficient comprises: Assume a set of failure physical model coefficients to be identified , the test data obey normal distribution; given the likelihood function represents the probability of the observed data under the given parameters , the calculation formula is shown in (7): (7) where x i denotes degradation or lifetime data of the sample size, denotes the joint probability distribution, denotes the likelihood function; The logarithm of the likelihood function is taken to obtain the logarithm likelihood function: (8) Finding optimal parameters based on optimization algorithm, computing log-likelihood function Parameters for maximization : (9) where arg max denotes the value of the variable that maximizes the posterior ​ According to the estimated parameters , the failure physical model parameters are obtained.

9. A method of estimating failure of a MOS field effect transistor taking into account mass uniformity, characterized by, For the MOS field effect tube, first, a MOS field effect tube failure physical modeling method considering quality consistency is used to model, and a MOS field effect tube failure model considering quality consistency corresponding to the MOS field effect tube is obtained, and then the MOS field effect tube is estimated according to the MOS field effect tube failure model considering quality consistency.

Images