Metal-oxide-semiconductor field effect transistor (MOSFET) service life quantification method based on fusion stress model

Abstract

The invention discloses an MOSFET service life quantitative evaluation method based on a fusion stress model, and the method comprises the steps: S1, building a fusion stress model which is used for describing the mathematical relationship between the degradation amount of key electrical parameters of an MOSFET and time, temperature, voltage and current stress, and the basic form of the fusion stress model is as follows: S1: S1: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S1: S2: S2: S2, multiple groups of accelerated degradation tests under different stress levels are carried out on the MOSFET sample, and performance degradation data are collected regularly; s3, performing parameter identification on the fusion stress model by using the performance degradation data; and S4, extrapolating the identified complete fusion stress model to a target working stress condition of the device, and quantitatively calculating the predicted life of the MOSFET under the target working condition through a preset failure threshold. Accurate service life evaluation of the long-service-life MOSFET device can be completed in a short time, and a scientific basis is provided for design, operation and maintenance of a high-reliability power electronic system.

Description

Technical Field

[0001] This invention relates to the field of power electronic device reliability assessment technology, and in particular to a MOSFET lifetime quantification method based on a fused stress model. Background Technology

[0002] Power metal-oxide-semiconductor field-effect transistors (MOSFETs) are core switching devices in modern power electronic systems, widely used in power controllers, motor drives, and new energy conversion. Their reliability directly determines the stability and lifespan of the entire system. As power electronic devices develop towards higher power density, higher efficiency, and longer lifespan, higher requirements are placed on the reliability assessment of key semiconductor devices.

[0003] In practical applications, power MOSFETs are typically subjected to complex, multi-stress environments, including temperature cycling, high voltage stress, and high current surges. These factors work together to cause gradual performance degradation. However, under normal operating conditions, MOSFETs, as long-life devices, exhibit extremely slow performance degradation, with lifespans reaching several years or even more than a decade. This long-life characteristic presents a significant challenge to reliability verification during the product design phase, as it is impossible to obtain sufficient failure data within a reasonable timeframe using conventional lifetime testing methods.

[0004] Currently, most lifetime assessment methods for power semiconductor devices rely on a single stress acceleration model. For example, the Arrhenius model only considers the impact of temperature stress on device lifetime, which makes it difficult to accurately reflect the complex degradation process caused by the coupling of multiple stresses such as temperature, voltage, and current in actual operation, resulting in significant deviations between lifetime prediction results and actual conditions.

[0005] Furthermore, existing lifetime assessment processes often rely on data from complete device failure, requiring waiting for samples until complete functional loss. This is not only time-consuming and labor-intensive, but also fails to predict lifetime during the performance degradation process.

[0006] Therefore, there is an urgent need for an evaluation method that can integrate the effects of multiple stresses, be based on performance degradation data, and quickly and accurately quantify MOSFET lifetime.

[0007] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a quantitative assessment method for MOSFET lifetime based on a fused stress model, which has clear physical meaning and high assessment accuracy. This method constructs a physical model that simultaneously characterizes the combined effects of time, temperature, and electrical stress. Combined with accelerated degradation test design and parameter identification strategies, it can complete accurate lifetime assessments of long-life MOSFET devices in a relatively short time, providing a scientific basis for the design and maintenance of high-reliability power electronic systems.

[0009] To achieve the above objectives, a MOSFET lifetime quantification assessment method based on a fusion stress model is proposed, comprising the following steps: Step S1: Construct a fusion stress model, which describes the mathematical relationship between the degradation of key electrical parameters of the MOSFET and time, temperature, voltage, and current stress. The basic form of the fusion stress model is as follows: ; Where ΔP is the performance degradation, t is time, T is the absolute operating temperature of the device, V is the operating voltage of the device, I is the operating current, A, n, θ1, θ2, and θ3 are all undetermined model parameters, and f(T; θ1), g(V; θ2), and h(I; θ3) are functions related to the absolute operating temperature of the device, the operating voltage of the device, and the operating current, respectively. Step S2: Perform accelerated degradation tests on the MOSFET samples under different stress levels and collect performance degradation data at regular intervals. Step S3: Using the performance degradation data, perform parameter identification on the fusion stress model; Step S4: Extrapolate the identified complete fusion stress model to the target operating stress conditions of the device, and quantify the predicted lifetime of the MOSFET under the target operating conditions by using a preset failure threshold.

[0010] For example, in step S3, the parameter identification includes: The fusion stress model is linearized, and linear least squares fitting is performed on the linearized model to obtain the initial estimated values ​​of the parameters of the undetermined model. It is then determined whether the fitting accuracy meets the preset requirements. If it does, the linear fitting result is adopted and step S4 is executed. Otherwise, the initial estimated value is used as the initial value of the nonlinear least squares fitting algorithm, and parameter fine-tuning is performed on the original fusion stress model structure.

[0011] For example, the fusion stress model is specifically as follows: ; Where, ΔV th Where B is the threshold voltage change, C and E are acceleration factors related to electrical stress. αThe activation energy is T, where k is the Boltzmann constant. e The absolute temperature corresponding to 25℃ is 298.15K.

[0012] For example, in step S2, the accelerated degradation test is a power cycling test, and the applied stress includes gate voltage, drain-source voltage and junction temperature higher than the device rating.

[0013] For example, the key electrical parameter is the threshold voltage.

[0014] For example, the failure threshold is set as the percentage change of the critical electrical parameter relative to its initial value, and the device is determined to fail when the threshold voltage change reaches 20% of its initial value.

[0015] For example, in the parameter identification process, by fixing the stress conditions, the model is simplified to the form ln(ΔP) = m + nln(t), and the experimental data is linearly fitted to obtain the composite constant term m and the undetermined model parameter n corresponding to the stress level. Then, m and n are expressed as functions of stress, and the parameter identification of the fusion stress model is completed by regression fitting.

[0016] Compared with the prior art, the technical solution of the present invention has at least the following beneficial effects: (1) The integrated stress model established in this invention organically integrates three classical physical models (power law, Arrhenius, and Irene), which can accurately capture the synergistic effect mechanism between different stresses, thereby more realistically characterizing the degradation characteristics of MOSFET devices under actual complex operating conditions. The integrated stress model organically integrates time, temperature, voltage, and current stresses within the same mathematical framework, overcoming the limitation of traditional single stress models that cannot reflect the coupling effect of multiple stresses, making the lifetime prediction results closer to the actual application environment.

[0017] (2) This invention monitors the gradual degradation of key parameters such as threshold voltage instead of waiting for complete failure. Combined with the design of accelerated degradation tests on MOSFET samples under multiple stress levels, it can obtain sufficient degradation data in a short time. This not only significantly shortens the evaluation cycle, but also captures the interaction between different stress factors, providing a high-quality experimental basis for subsequent model parameter identification.

[0018] (3) The present invention also designed a two-stage parameter identification process. When the linear fitting result obtained by using performance degradation data to perform linear least square fitting on the fusion stress model does not meet the accuracy requirements, a nonlinear least square fitting algorithm is used to refine the model parameters, which fundamentally solves the error structure distortion problem caused by model linearization and improves the accuracy of life prediction. Attached Figure Description

[0019] Figure 1 The overall flowchart of a MOSFET lifetime quantification method based on a fusion stress model provided by the present invention is shown below.

[0020] Figure 2 This is a schematic diagram illustrating the lifetime quantification of the fusion stress model according to the present invention. Detailed Implementation

[0021] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a further detailed explanation of the MOSFET lifetime quantification method based on a fusion stress model proposed in this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the accompanying drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the embodiments of this invention. Please refer to the accompanying drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes to aid those skilled in the art and are not intended to limit the implementation conditions of this invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0022] As described in the background section, existing MOSFET lifetime assessment techniques largely rely on a single stress-accelerated model, which makes it difficult to accurately characterize the complex degradation process in actual operation.

[0023] To address this, this invention constructs a physical model capable of simultaneously characterizing the synergistic effects of multiple stresses by hierarchically deconstructing and systematically integrating the MOSFET degradation mechanism. Starting from the three fundamental dimensions driving degradation (time accumulation effect, temperature activation effect, and electrical stress promotion effect), it organically integrates the power-law model, the Arrhenius model, and the Irene model, forming a fused stress model through multiplicative relationships to accurately describe the complex degradation behavior of MOSFETs under actual operating conditions. Simultaneously, a two-stage parameter identification strategy is employed to resolve the error structure distortion problem caused by model linearization.

[0024] This invention provides a method for quantitatively evaluating MOSFET lifetime based on a fusion stress model, comprising the following steps: Step S1: Construct a fusion stress model, which describes the mathematical relationship between the degradation of key electrical parameters of the MOSFET and time, temperature, voltage, and current stress. The basic form of the fusion stress model is as follows: (1-1); Where ΔP is the performance degradation, t is time, T is the absolute operating temperature of the device, V is the operating voltage of the device, I is the operating current, A, n, θ1, θ2, and θ3 are all undetermined model parameters, and f(T; θ1), g(V; θ2), and h(I; θ3) are functions related to the absolute operating temperature of the device (hereinafter referred to as temperature), the operating voltage of the device (hereinafter referred to as voltage), and the operating current (hereinafter referred to as current), respectively.

[0025] In some embodiments, the key electrical parameter is a threshold voltage.

[0026] In some embodiments, the fusion stress model is specifically as follows: (1-2); Where, ΔV th Where B is the threshold voltage change, C and E are acceleration factors related to electrical stress. α The activation energy is T, where k is the Boltzmann constant. e The absolute temperature corresponding to 25℃ is 298.15K.

[0027] Step S2: Perform accelerated degradation tests on the MOSFET samples under different stress levels and collect performance degradation data at regular intervals.

[0028] In some embodiments, the accelerated degradation test is a power cycling test, and the applied stress includes gate voltage, drain-source voltage, and junction temperature higher than the device rating.

[0029] Step S3: Using the performance degradation data, perform parameter identification on the fusion stress model.

[0030] In some embodiments, step S3 includes: Step S3.1: Using the performance degradation data, perform linear least squares fitting on the fusion stress model to obtain initial estimates of the parameters of the undetermined model; determine whether the fitting accuracy meets the preset requirements. If it does, adopt the linear fitting result and proceed to step S4; otherwise, proceed to step 3.2. The basic form of the linear least squares fitting is as follows: (1-3).

[0031] Step S3.2: Use the initial estimated value as the initial value for the nonlinear least squares fitting algorithm, and perform parameter fine-tuning on the original fused stress model structure; the basic form of the nonlinear least squares fitting is as follows: (1-4); Wherein, ΔP i t represents the performance degradation (e.g., threshold voltage change) measured at the i-th data point.i Let T be the time corresponding to the i-th data point. i V represents the absolute operating temperature of the device corresponding to the i-th data point. i I represents the device operating voltage corresponding to the i-th data point. i Let Θ be the operating current corresponding to the i-th data point, and Θ be the parameter vector of the model (e.g., A, n, B, C, E). α etc.), including all model parameters to be identified (to be determined); Φ(t) i , T i V i , I i ; Θ) is the functional expression of the fused stress model, representing the stress condition (t) under given stress conditions. i , T i V i , I i The amount of performance degradation predicted by the model under parameter Θ.

[0032] In some embodiments, during parameter identification, for example under fixed stress conditions, the model is simplified to the form ln(ΔP) = m + nln(t), and the experimental data is linearly fitted to initially obtain the composite constant term m corresponding to the stress level and the undetermined model parameter (time exponent) n; m and n are expressed as functions of stress, and their specific relationship is determined by regression fitting, thereby completing the parameter identification of the entire fused stress model.

[0033] Step S4: Extrapolate the identified complete fusion stress model to the target operating stress conditions of the device, and quantify the predicted lifetime of the MOSFET under the target operating conditions by using a preset failure threshold.

[0034] In some embodiments, the failure threshold is set as the percentage change of the critical electrical parameter relative to its initial value, and the device is determined to have failed when the threshold voltage change reaches 20% of its initial value.

[0035] The technical solution of the present invention will be described in further detail below.

[0036] MOSFET performance degradation is essentially a physicochemical change process that occurs in materials under various environmental stresses. To accurately describe this complex process, this invention organically integrates three classic physical empirical models, starting from the three fundamental dimensions of driving degradation.

[0037] First, degradation is a cumulative process over time. The basis for describing this macro-trend is the power-law model, which takes the form: (2-1); Where ΔP is the performance degradation, t is time, and A and n are undetermined model parameters. This model (2-1) establishes the basic power-law relationship between performance degradation and time. The model parameter n determines the acceleration or deceleration characteristics of the degradation process and constitutes the time baseline for lifetime prediction.

[0038] However, the rate of device degradation is not constant; it is also significantly driven by temperature. This microscopic mechanism can be characterized by the Arrhenius model, the core of which is: (2-2); Where R represents the degradation rate. This model (2-2) reveals that temperature, by providing energy, helps material atoms or defects cross specific energy barriers, thereby exponentially accelerating internal physicochemical reactions such as interface trap formation and gate oxide defect generation. The model (2-2) explains why a small increase in junction temperature leads to a sharp decrease in device lifetime.

[0039] In practical applications, electrical stress also plays a crucial role, and the Irene model, derived from quantum mechanics principles, provides a framework for understanding this effect. Compared to purely empirical models, the Irene model starts from first principles and can simultaneously consider the influence of temperature and non-temperature stresses such as electric fields and humidity on lifetime. Its general form includes the term exp(γ⋅S), where S represents non-temperature stress and γ is the acceleration factor. In applications involving electrical stress, it is often simplified to L∝exp(-γV) or L∝exp(-γE). ox ), where L is the device lifetime, E ox The gate oxide electric field is shown. The Erin model reveals that electric field or current, or other electrical stresses, directly promote degradation processes such as carrier injection and trap generation through quantum mechanical mechanisms.

[0040] In actual operating conditions, the effects of time, temperature, and electrical stress are not isolated but concurrent and synergistically coupled. Therefore, fusing these three models, which separately characterize the stress effects of different dimensions, through a multiplicative relationship is physically complete and reasonable. The complete form of the resulting fused stress model is as follows: (2-3); Where B and C are acceleration factors related to electrical stress. To facilitate linearization and parameter identification, the natural logarithm of both sides of the above model is taken, yielding: (2-4); This form not only highlights the linear relationship between the logarithm of degradation and the logarithm of time, but also clearly demonstrates how current, voltage, and temperature (in their reciprocal form) act as linear terms to modulate the intercept of the entire degradation trajectory, laying a theoretical foundation for the experimental identification of model parameters.

[0041] To obtain the model parameters, it is necessary to design an accelerated degradation experiment of the system. Based on the experimental data, this invention adopts a parameter identification strategy that balances efficiency and accuracy, and specifically considers the error structure changes caused by model linearization. The fused stress model is linearizable in form and can be transformed into equation (2-4) through logarithmic transformation. This linearization technique simplifies the calculation while fundamentally changing the error structure of the optimization problem. To illustrate this point, consider a simplified nonlinear model. This corresponds to a special case of this model under fixed stresses.

[0042] In the original space (the space of performance degradation ΔP), the most natural assumption is that the measurement error ε is additive, independent, and homoscedastic, i.e., the data generation process is as follows: (2-5); Among them, y i is the observed value at the i-th data point, representing the performance degradation of the MOSFET under specific time and stress conditions; c1 is the first parameter of the model, the magnitude coefficient of the exponential function, which is related to material properties and the basic degradation rate in the fused stress model of this invention; c2 is the second parameter of the model, the coefficient of the exponential term, controlling the sensitivity of the degradation rate to changes in the independent variable; x i The independent variable for the i-th data point can be stress t, T, V, or I, or a transformation thereof; ε i Let be the random error term for the i-th data point, representing measurement noise or random fluctuations not captured by the model. In this case, the optimization objective of nonlinear least squares is to minimize the sum of squared absolute prediction residuals of performance degradation: (2-6).

[0043] After taking the logarithm of the model, what is actually processed is Where z is the transformed dependent variable, and y is the natural logarithm of the original observation y. Assuming the original data conforms to the above additive error model, then: (2-7); Among them, z i The transformed dependent variable is y, which is the original observation value of the i-th data point. i The natural logarithm of .

[0044] Through ε i Taylor expansion analysis at =0 shows that: (2-8); Equation (2-8) reveals that the error is no longer a simple additive problem; the error term in the logarithmic space is also related to the predicted value itself. If higher-order terms are ignored, the error variance in the logarithmic space is approximately: (2-9).

[0045] This means that early degenerate data points with smaller predicted values ​​are given higher weights in the logarithmic space, while later degenerate data points with larger predicted values ​​are given lower weights. Therefore, the linearized regression problem essentially minimizes the sum of squared relative errors, rather than the sum of squared absolute errors that the original problem was concerned with. The resulting parameter estimates are not the optimal unbiased estimates of the original nonlinear least squares problem; their bias depends on the data range and the level of data noise.

[0046] To address the aforementioned error structure distortion problem and ensure model prediction accuracy, this invention employs a two-stage refined parameter identification process: (1) Linearization initial estimation: First, use the log-linear model to perform ordinary least squares regression to quickly obtain the model parameters (A, n, B, C, E). a This involves obtaining a set of initial estimates for the stress. Specifically, under fixed stress conditions, the fusion model is simplified to ln(ΔP) = m + nln(t), and regression analysis is used to obtain the time exponent n and the composite constant term m for each stress combination. Subsequently, by analyzing the correlation between the composite constant term m and each stress parameter, all parameters of the model are decoupled and obtained. In this process, robust estimation algorithms such as RANSAC can be introduced to effectively suppress the influence of measurement outliers on parameter identification and improve the reliability of the model. This step is computationally efficient and provides a high-quality iterative starting point close to the true solution.

[0047] (2) Nonlinear optimization and refinement: Starting with the initial estimate obtained from the linearization initial estimate, a nonlinear least squares fitting algorithm (such as the Levenberg-Marquardt algorithm) is used for iterative optimization in the original nonlinear model space. The objective function of this stage is directly defined as the sum of squared absolute residuals in the original space. In this way, the algorithm will directly minimize the absolute error between the predicted and measured values ​​of performance degradation, thereby obtaining the final parameter estimate that conforms to the original problem definition and has better statistical properties.

[0048] The parameter identification strategy employed in this invention not only combines the simplicity of linearization methods with the accuracy of nonlinear optimization, but also fundamentally corrects the error structure bias introduced by linearization. This ensures the mathematical rigor of the established lifetime prediction model, making its prediction results more reliable and accurate for practical engineering applications.

[0049] Finally, by extrapolating the fully identified model to the normal stress conditions under which the device is expected to operate, and by applying a preset failure threshold, the predicted lifetime of the MOSFET can be obtained through inverse solving. This method not only achieves a quantitative assessment of device lifetime but also enables the analysis of B, C, and E... a These parameters quantitatively reveal the contribution of different stresses to the degradation process, providing a deeper scientific basis for the optimized design of devices and the formulation of application conditions.

[0050] Example The implementation of this method begins with defining the device failure modes and key degradation parameters. As described in the background section, the failure of power MOSFETs is mainly manifested as parameter drift, in which the threshold voltage Vf th Extremely sensitive to microscopic degradation mechanisms such as gate oxide charge trapping and interface state density, it was therefore selected as a key degradation characterization parameter in this embodiment. Based on industry consensus and device specifications, when the threshold voltage drift ΔV th When the value reaches 20% of its initial value, the device is considered to have failed. This threshold is defined as the failure criterion D in this evaluation. fail .

[0051] To obtain the degradation data required for modeling within a reasonable timeframe, accelerated degradation tests must be designed and executed. These tests are conducted on an accelerated aging test platform equipped with a precise environmental temperature chamber, programmable power supply and load, gate drive circuitry, and a high-precision data acquisition system. Given that temperature (T), voltage (V), and current (I) are the main external stresses affecting MOSFET degradation, this experiment employs an orthogonal experimental design method to efficiently cover the combined effects of multiple stress levels with a minimal number of test groups. The determination of each stress level relies on pre-tests: the temperature stress range must refer to the system's extreme operating environment and the upper limit of the device junction temperature; the electrical stress level is explored through pre-tests to determine its lower limit (ensuring measurable degradation) and upper limit (avoiding instantaneous breakdown), thereby ensuring that the accelerated test effectively shortens the time without introducing atypical failure mechanisms.

[0052] The experiment employed a power cycling profile to simulate actual operating conditions. Specifically, a square wave signal with a specific frequency (1kHz) and duty cycle (40%) was applied to the gate of the device, while a bias voltage (4Vdc) and a load (0.2Ω) were applied to the drain circuit, causing the device to periodically turn on and off. Closed-loop temperature control was used to cycle the device's casing temperature between preset upper and lower limits, thereby generating junction temperature fluctuations ΔT within the chip. j This is the key mechanism for inducing thermal mechanical fatigue. Throughout the experiment, the test needs to be interrupted periodically, and the sample placed at standard room temperature to measure its threshold voltage V. th The parameters are recorded precisely, including the measurement time t, the corresponding stress conditions (T, V, I), and performance parameter values, to provide a data foundation for subsequent modeling.

[0053] After acquiring the experimental data, the core model parameter identification stage begins, and its overall process follows the following... Figure 1 The logic is as follows. First, the acquired threshold voltage data is preprocessed to calculate its drift ΔV relative to the initial value. th (t). Since all parameter measurements are performed at a uniform room temperature, the instantaneous influence of temperature on semiconductor parameters in online monitoring has been naturally eliminated, and the data has been normalized in the early stages.

[0054] The core of this embodiment lies in directly identifying model parameters through a linear regression method. This method ensures sufficient engineering accuracy while also offering advantages such as computational simplicity and ease of implementation, making it highly suitable for engineering applications. Specifically, it identifies ΔV under different stress combinations. th (t) Data is based on the formula ln(ΔV) th Linear regression is performed using the formula ) = m + nln(t). This step yields estimates of the time exponent n and the composite constant term m for each stress group, as shown in Table 1.

[0055] Table 1. Coefficient Constant Term Table Subsequently, the composite constant term *m* obtained from each group was correlated with the corresponding stress conditions. Through multiple linear regression, the relationship was determined as follows: (3-1) By performing a fitting, the parameters (A, B, C, E) of the fusion stress model can be decoupled in one step. a The time exponent n, based on extensive literature, is related to the device's operating temperature, electrical stress, room temperature, and maximum rated voltage-to-current ratio. For the components used in the experiment, it satisfies the following relationship: (3-2).

[0056] Similar to the method used for the m operation, the parameters (D, Q) of the fused stress model are decoupled. Here, D is the basic parameter, representing the base value of the time exponent n under reference operating conditions (25°C, rated voltage, and current); Q is the acceleration factor parameter, characterizing the combined sensitivity of temperature, voltage, and current to the stress acceleration effect, reflecting the strength of the influence on the time exponent n when the stress conditions deviate from the reference value. To improve the robustness of this method, the RANSAC (Random Sample Consensus) algorithm can be introduced to clean the data before regression. This algorithm effectively enhances the model's robustness to measurement noise.

[0057] It should be noted that while the above linearization method is very convenient in engineering, it minimizes the residual in the logarithmic space, which differs theoretically from the objective of minimizing the absolute error in the original space. For applications requiring extremely high prediction accuracy, a better improvement is to use the parameters identified by the linearization method in this embodiment as the initial values ​​for nonlinear least squares optimization, and then further refine them in the original data space. The objective function is: (3-3).

[0058] This improved approach directly minimizes the absolute prediction error, thereby obtaining parameter estimates with better statistical properties. However, for most engineering applications, the direct linearization method described in this embodiment already provides sufficient accuracy and reliability.

[0059] Finally, lifetime extrapolation and quantification were performed using the identified complete fusion stress model (3-4).

[0060] (3-4).

[0061] It is evident that under this failure mechanism, temperature and current magnitude have the greatest impact on the degradation of the threshold voltage parameter, while voltage has a relatively small influence. Subsequently, the normal operating stress conditions of the MOSFET under evaluation in the target application are substituted into model equation (3-4) to solve for time t. life This is the predicted lifetime of the device under specified operating conditions and threshold voltage drift limits. In this embodiment, the calculated lifetime is approximately 10.65 years, and its average lifetime versus operating temperature curve is shown below. Figure 2 As shown.

[0062] In summary, this invention constructs a complete quantitative assessment technology system for MOSFET lifetime: by integrating power-law, Arrhenius, and Irene models, a physical model is established that simultaneously characterizes the coupling effects of time, temperature, voltage, and current stress; multi-stress accelerated degradation testing is conducted using experimental design; an innovative two-stage parameter identification strategy of "linearized initial estimation + nonlinear fine-tuning" is proposed to effectively solve the problem of error structure distortion; the identification model is extrapolated to normal operating conditions, and lifetime is accurately predicted by combining a preset failure threshold (threshold voltage drift of 20%). Verification shows that this method can accurately quantify the contribution of different stresses and predict lifetime, significantly shortening the assessment cycle and providing a scientific basis for the design and maintenance of high-reliability power electronic systems.

[0063] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0064] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for quantitatively evaluating the lifetime of MOSFETs based on a fusion stress model, characterized in that, Includes the following steps: Step S1: Construct a fusion stress model, which describes the mathematical relationship between the degradation of key electrical parameters of the MOSFET and time, temperature, voltage, and current stress. The basic form of the fusion stress model is as follows: ； Where ΔP is the performance degradation, t is time, T is the absolute operating temperature of the device, V is the operating voltage of the device, I is the operating current, A, n, θ1, θ2, and θ3 are all undetermined model parameters, and f(T; θ1), g(V; θ2), and h(I; θ3) are functions related to the absolute operating temperature of the device, the operating voltage of the device, and the operating current, respectively. Step S2: Perform accelerated degradation tests on the MOSFET samples under different stress levels and collect performance degradation data at regular intervals. Step S3: Using the performance degradation data, perform parameter identification on the fusion stress model; Step S4: Extrapolate the identified complete fusion stress model to the target operating stress conditions of the device, and quantify the predicted lifetime of the MOSFET under the target operating conditions by using a preset failure threshold.

2. The MOSFET lifetime quantitative evaluation method based on the fusion stress model as described in claim 1, characterized in that, In step S3, the parameter identification includes: The fusion stress model is linearized, and linear least squares fitting is performed on the linearized model to obtain the initial estimated values ​​of the parameters of the undetermined model. It is then determined whether the fitting accuracy meets the preset requirements. If it does, the linear fitting result is adopted and step S4 is executed. Otherwise, the initial estimated value is used as the initial value of the nonlinear least squares fitting algorithm, and parameter fine-tuning is performed on the original fusion stress model structure.

3. The MOSFET lifetime quantitative evaluation method based on the fusion stress model as described in claim 1, characterized in that, The fusion stress model is specifically as follows: ； Where, ΔV th Where B is the threshold voltage change, C and E are acceleration factors related to electrical stress. α The activation energy is T, where k is the Boltzmann constant. e The absolute temperature corresponding to 25℃ is 298.15K.

4. The MOSFET lifetime quantitative evaluation method based on the fusion stress model as described in claim 1, characterized in that, In step S2, the accelerated degradation test is a power cycling test, and the applied stress includes gate voltage, drain-source voltage and junction temperature higher than the device's rated values.

5. The MOSFET lifetime quantitative evaluation method based on the fusion stress model as described in claim 1, characterized in that, The key electrical parameter is the threshold voltage.

6. The MOSFET lifetime quantitative evaluation method based on the fusion stress model as described in claim 1, characterized in that, The failure threshold is set as the percentage change of the key electrical parameter relative to its initial value. When the threshold voltage change reaches 20% of its initial value, the device is determined to have failed.

7. The MOSFET lifetime quantitative evaluation method based on the fusion stress model according to claim 1, characterized in that, During the parameter identification process, by fixing the stress conditions, the model is simplified to the form ln(ΔP) = m + nln(t). The experimental data are linearly fitted to obtain the composite constant term m corresponding to the stress level and the undetermined model parameter n. m and n are expressed as functions of stress, and the parameter identification of the fusion stress model is completed by regression fitting.

Images