Fault diagnosis method and system for low-on-resistance shield gate trench MOSFET (Metal Oxide Semiconductor Field Effect Transistor)

Abstract

The invention provides a low-on-resistance shield gate trench MOSFET fault diagnosis method and system, and relates to the technical field of MOSFET fault diagnosis, and the method comprises the steps: applying preset transient electrical excitation to a contact interface of a probe and a wafer bonding pad and an MOSFET before the steady-state on resistance of the MOSFET is measured; according to the obtained electrical response information of the contact interface and the MOSFET under each transient electrical excitation effect, constructing a transient feature vector; analyzing the transient feature vector according to a preset discrimination rule for contact part abnormity and a discrimination rule for device structure defects to determine whether an abnormity source is the contact part abnormity or the device structure defects; according to the determined abnormal source, the measured on-resistance value is adjusted, and the problem that the test system misjudges that the on-resistance of the device is abnormally increased due to instantaneous fluctuation of the contact resistance caused by the change of the microscopic geometrical shape of the tip of the probe is solved.

Description

Technical Field

[0001] This application relates to the field of MOSFET fault diagnosis technology, and more specifically, to a method and system for diagnosing faults in low on-resistance shielded gate trench MOSFETs. Background Technology

[0002] Low on-resistance shielded gate trench MOSFETs are widely used in power management, motor drives, and new energy inverters due to their excellent switching performance and high efficiency. Their on-resistance (RDS(on)) is a core performance indicator; even a slight change can directly affect system efficiency and reliability. Therefore, accurate measurement and fault diagnosis of this parameter are crucial in manufacturing.

[0003] On high-throughput wafer testing production lines, automated test systems rely on probe cards to contact wafer test pads for electrical measurements. However, slight batch-to-batch stability of the probe tip plating material can cause the actual wear rate of the tips to be slightly faster than expected. This results in subtle changes to their micro-geometry, making the contact with the pads less than ideal and causing random, transient fluctuations in contact resistance at the milliohm or even microohm level. These fluctuations are extremely small and random, often making them difficult for traditional test systems to identify as anomalies during routine self-checks or calibrations. When measuring devices with on-resistance of only a few milliohms or even lower, these minute contact resistance fluctuations can be misinterpreted by the test system as an increase in the device's own resistance, severely impacting measurement accuracy.

[0004] Furthermore, when introducing a new product with a lower target on-resistance, engineers used the same test procedure as the previous generation to expedite the verification process. The signal acquisition stabilization time in this procedure was set based on the previous generation's higher resistance value. For the new product with extremely low resistance, the relative weight of any noise or transient effects is significantly amplified, and the previously short stabilization time is already at a critical point, making the ability to suppress transient noise clearly insufficient.

[0005] Under these test conditions, the system's ability to suppress probe contact fluctuations decreases. If the device under test itself has minor structural defects, such as the gate shield thickness being at the tolerance edge or the micro-roughness of the trench sidewall slightly exceeding the range, its actual on-resistance will be slightly higher than the normal value. This, combined with the random fluctuations in probe contact resistance, causes some devices that are actually qualified but at the tolerance edge to be misjudged as having abnormal resistance and thus incorrectly rejected. This misjudgment caused by the combined effect of multiple factors results in unstable "false fluctuations" in the production line yield.

[0006] Ultimately, production management faced a dilemma: although key process parameters were all within normal monitoring ranges and routine defect scans revealed no obvious problems, the yield of certain batches of products consistently fell short of expectations. Even more perplexing was that when rejected devices were retested in the laboratory, most of the on-resistance returned to normal or was only slightly elevated, making it difficult for engineers to pinpoint specific process defects. This problem severely hampered the accurate diagnosis of the correlation between trench micro-defects and on-resistance, while also increasing unnecessary production costs and market delivery risks. Summary of the Invention

[0007] This application provides a method and system for fault diagnosis of low on-resistance shielded gate trench MOSFETs, aiming to solve the problem of misjudgment of on-resistance measurement caused by probe contact resistance fluctuations and slight structural defects in the device itself on the power MOSFET wafer testing production line, as well as the resulting instability of yield data and difficulty in fault diagnosis.

[0008] On the one hand, this application provides a fault diagnosis method for low on-resistance shielded gate trench MOSFETs, including: Before measuring the steady-state on-resistance of the MOSFET, a pre-set transient electrical excitation is applied to the contact interface between the probe and the wafer pad and the MOSFET. Acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation, the electrical response information including voltage response data, impedance response data and contact resistance fluctuation data; Key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the electrical response information to construct a transient feature vector; Based on preset discrimination rules for contact component anomalies and discrimination rules for device structural defects, the transient feature vector is analyzed to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. Adjust the measured on-resistance value based on the identified source of the anomaly.

[0009] On the other hand, this application provides a low on-resistance shielded gate trench MOSFET fault diagnosis system, the system comprising: An excitation application module is used to apply a pre-set transient electrical excitation to the contact interface between the probe and the wafer pad and the MOSFET before measuring the steady-state on-resistance of the MOSFET. The response acquisition module is used to acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation. The electrical response information includes voltage response data, impedance response data, and contact resistance fluctuation data. The parameter extraction module is used to extract key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information, and to construct a transient feature vector. The anomaly detection module is used to analyze the transient feature vector according to preset detection rules for contact component anomalies and device structural defects, so as to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. The resistance adjustment module is used to adjust the measured on-resistance value based on the identified source of the anomaly.

[0010] The low on-resistance shielded gate trench MOSFET fault diagnosis method and system disclosed in this application effectively solves the problems in the prior art caused by the instantaneous fluctuation of contact resistance due to changes in the micro-geometry of the probe tip, and the insufficient signal stabilization waiting time caused by the continued use of new product test procedures, which leads to the test system misjudging the abnormal increase in the on-resistance of the device. Attached Figure Description

[0011] To illustrate this application more clearly, the accompanying drawings used in the embodiments will be briefly described below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0012] Figure 1 The diagram above illustrates a flowchart of a low on-resistance shielded gate trench MOSFET fault diagnosis method. Figure 2 The diagram above illustrates a structural schematic of a low on-resistance shielded gate trench MOSFET fault diagnosis system.

[0013] Reference numerals: 100, Low on-resistance shielded gate trench MOSFET fault diagnosis system; 10, Excitation application module; 20, Response acquisition module; 30, Parameter extraction module; 40, Anomaly detection module; 50, Resistance adjustment module. Detailed Implementation

[0014] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0015] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0016] Traditional low on-resistance shielded gate trench MOSFETs suffer from significant limitations in their on-resistance measurement capabilities. These limitations arise from the microscopic geometry of the probe tip, causing instantaneous fluctuations in contact resistance. Furthermore, insufficient stabilization time before signal acquisition in the test procedure further reduces the system's ability to suppress these contact resistance fluctuations. When the MOSFET device itself exhibits minor trench structure defects, these defects, combined with the random fluctuations in probe contact resistance, can lead to some devices that are actually within acceptable tolerances being incorrectly identified as having abnormally high on-resistance and thus wrongly rejected.

[0017] like Figure 1 The diagram illustrates a flowchart of a low on-resistance shielded gate trench MOSFET fault diagnosis method. This application proposes a low on-resistance shielded gate trench MOSFET fault diagnosis method, comprising: S10, Before measuring the steady-state on-resistance of the MOSFET, a preset transient electrical excitation is applied to the contact interface between the probe and the wafer pad and the MOSFET. The contact interface between the probe and the wafer pad refers to the area where the probe contacts the metal pads on the MOSFET wafer during testing. The contact state of this interface directly affects the accuracy of electrical measurements. Transient electrical excitation refers to a non-steady-state electrical signal, such as a pulsed current or AC signal, applied for a short period of time, with the aim of stimulating the transient response of the device and the contact interface.

[0018] S20, acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation, the electrical response information including voltage response data, impedance response data and contact resistance fluctuation data; S30, extract key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information, and construct a transient feature vector; The transient feature vector is a set of key transient features extracted from electrical response information, used to characterize contact state and device characteristics.

[0019] S40, based on preset discrimination rules for contact component anomalies and discrimination rules for device structural defects, the transient feature vector is analyzed to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. Contact component abnormalities refer to problems with contact components such as probes or pads, such as probe wear or contamination. Device structural defects refer to defects in the internal structure of the MOSFET, such as gate oxide layer defects or abnormal trench morphology.

[0020] S50, adjusts the measured on-resistance value based on the identified source of the anomaly.

[0021] Before measuring the steady-state on-resistance of a MOSFET, a pre-defined transient electrical stimulus needs to be applied to the contact interface between the probe and the wafer pad, as well as to the MOSFET. This transient electrical stimulus can take various forms. For example, a voltage pulse with a specific amplitude and duration can be applied, or an AC current signal with a specific frequency and amplitude can be applied. Alternatively, a fast-rising-edge current pulse can be injected into the contact interface and the MOSFET using a controllable current source. The purpose of these stimuli is to perturb the contact interface and the MOSFET for a short period of time, causing them to produce a measurable transient electrical response.

[0022] After applying transient electrical excitation, it is necessary to acquire the electrical response information of the contact interface and MOSFET under each transient electrical excitation. The electrical response information includes voltage response data, impedance response data, and contact resistance fluctuation data. Voltage response data can be obtained by recording the voltage changes across the contact interface or MOSFET in real time using a high-bandwidth oscilloscope or data acquisition system. Impedance response data can be obtained by measuring the complex impedance of the contact interface or MOSFET at different frequencies using an impedance analyzer. Contact resistance fluctuation data can be obtained by monitoring minute changes in contact resistance in real time while the excitation is applied using a high-precision micro-ohmmeter or a four-wire measurement system. For example, when a current pulse is applied, the voltage change curve over time can be recorded simultaneously to obtain voltage response data; when AC signals of different frequencies are applied, the impedance value at each frequency can be measured and recorded to obtain impedance response data; while continuously applying excitation, minute fluctuations in contact resistance can be captured using a high-sensitivity resistance measurement device to obtain contact resistance fluctuation data.

[0023] From electrical response information, it is necessary to extract key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET, and construct a transient feature vector. Various methods can be used to extract key transient features. For example, for voltage response data, parameters such as voltage settling time, overshoot amplitude, and oscillation frequency can be extracted. For impedance response data, parameters such as impedance value at a specific frequency, slope of the impedance spectrum, and resonant frequency can be extracted. For contact resistance fluctuation data, parameters such as fluctuation amplitude, frequency, and duration can be extracted. These extracted parameters can be combined into a transient feature vector. For example, the time required for the voltage to rise from zero to 90% of its steady-state value can be calculated from the voltage response data as the voltage settling time; the impedance change rate within a specific frequency range can be calculated from the impedance response data; and the root mean square value of the resistance fluctuation can be calculated from the contact resistance fluctuation data. These calculated values ​​collectively constitute the transient feature vector.

[0024] Based on pre-defined discrimination rules for contact component anomalies and device structural defects, transient feature vectors need to be analyzed to determine whether the anomaly originates from a contact component anomaly or a device structural defect. Discrimination rules can be based on empirical thresholds, statistical models, or machine learning algorithms. For example, a series of thresholds can be set; if a feature value in the transient feature vector exceeds a preset range, it is judged as an anomaly. Furthermore, a decision tree or support vector machine model can be built based on the sensitivity of different features to contact component anomalies and device structural defects. For example, if the voltage settling time is significantly prolonged and the contact resistance fluctuation is large, it may indicate a contact component anomaly; if the impedance spectrum shows anomalies at a specific frequency and the voltage overshoot is small, it may indicate a device structural defect. By inputting the transient feature vectors into these discrimination rules, the determination result of the anomaly source can be output.

[0025] Based on the identified source of the anomaly, the measured on-resistance value needs to be adjusted. If the anomaly is determined to be a contact component malfunction, the measured on-resistance value can be compensated according to the degree of the malfunction. For example, if probe wear is determined to be causing increased contact resistance, a preset compensation value can be subtracted from the total measured value. If the anomaly is determined to be a device structural defect, the on-resistance value may not need to be adjusted; instead, the device should be marked as defective, and the defect type should be further analyzed. For example, if a contact component malfunction is determined to be causing the measured value to be too high, the measured on-resistance value can be corrected based on historical data or calibration curves to obtain an on-resistance value closer to the true value. If a device structural defect is determined to be causing the measured value to be too high, this measured value may be the device's true on-resistance; in this case, no adjustment is needed, and the device should be marked as defective.

[0026] The advantage of this application lies in its ability to effectively solve the misjudgment problems caused by probe contact resistance fluctuations and insufficient testing procedures in traditional methods. By introducing transient electrical excitation and comprehensively analyzing voltage response data, impedance response data, and contact resistance fluctuation data, this application can identify the unique "fingerprints" of contact component anomalies and device structural defects from transient characteristics. For example, contact resistance fluctuations caused by probe wear typically exhibit characteristics within a specific frequency range and amplitude, while internal trench structure defects may produce unique changes in voltage settling time or impedance spectrum at specific frequency bands. By constructing transient feature vectors and applying specialized discrimination rules, this application can accurately locate the source of anomalies, thereby avoiding incorrect adjustments to on-resistance values ​​or misjudgments of qualified devices. This method not only improves the accuracy of fault diagnosis but also provides a more reliable quality control means for production lines, significantly reducing production costs and market delivery risks.

[0027] In some embodiments, the transient electrical excitation includes a current pulse with a fast rise time, an AC signal with a different frequency, and an electrical response acquired when a small mechanical disturbance is applied; the transient eigenvector includes voltage settling time, current decay constant, characteristic parameters of the impedance response curve, characteristic parameters of resistance fluctuation, and nonlinear response coefficients.

[0028] The current pulse with a fast rise edge refers to a current signal that rises from zero to a preset amplitude in an extremely short time (e.g., nanoseconds or microseconds). This pulse is used to excite transient charge dynamics processes within the MOSFET, such as carrier injection, trap filling or detrapping, and transient resistance changes at the probe-wafer pad contact interface. The AC signal with different frequencies refers to a sinusoidal or other periodic electrical signal with an adjustable frequency range. This signal is used to detect frequency-dependent electrical characteristics of the MOSFET device's internal structure and contact interface, such as dielectric constant, conductivity, and interface capacitance, through impedance spectroscopy analysis. During the probe-wafer pad contact process, minute, controllable mechanical vibrations or pressure changes are applied to the probe or wafer through external or internal mechanisms, and their electrical responses are measured simultaneously. This simulates or amplifies mechanical contact instabilities that may exist in actual testing, thereby obtaining information on minute fluctuations in contact resistance.

[0029] Furthermore, the voltage settling time in the transient characteristic vector refers to the time required from the start of the voltage waveform to the voltage reaching its steady-state value (e.g., 90% or 95%), reflecting the device's response speed and parasitic capacitance effect. The current decay constant refers to the time constant required for the current to decay from its peak value to a specific value (e.g., 1 / e) after the current pulse is removed; it is related to factors such as charge relaxation processes and trap density. The characteristic parameters of the impedance response curve refer to key values ​​extracted from the impedance versus frequency curve, such as the slope of the curve, inflection frequency, and resonant peak value. These parameters reveal the dielectric properties of the material, the interface charge transport mechanism, and defect distribution. The resistance fluctuation characteristic parameter refers to the amplitude and main frequency components of the contact resistance fluctuation over time when a small mechanical disturbance is applied, directly reflecting the physical contact quality and stability of the probe-pad contact interface. The nonlinear response coefficient is a quantity calculated by comparing the proportional relationship of electrical responses under different excitation intensities; it quantifies the nonlinear behavior of the device or contact interface, such as micro-tunneling effects and trap saturation effects.

[0030] Through the above technical solution, this application can comprehensively probe the internal structure and electrical characteristics of MOSFET devices and probe contact interfaces from multiple dimensions by introducing various transient electrical excitations. By extracting key transient features such as voltage settling time, current decay constant, characteristic parameters of impedance response curves, characteristic parameters of resistance fluctuations, and nonlinear response coefficients, it can more precisely distinguish contact component anomalies and device structural defects, thereby significantly improving the accuracy and reliability of fault diagnosis. These rich characteristic parameters help identify potential defects that are difficult to detect by traditional steady-state measurements, thereby improving the testing efficiency and product yield of low on-resistance shielded gate trench MOSFETs.

[0031] In some embodiments, the step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: Before the probe contacts the wafer pad of the MOSFET under test, a micro plasma generator is activated to perform transient cleaning of the probe tip. After the transient cleaning of the probe tip is completed, a current pulse with a fast rise edge is applied and the voltage response data is acquired. From the voltage response data, the starting point of the voltage waveform and the time point when the voltage reaches 90% of its steady-state value are identified to calculate the voltage settling time; the decay part of the current pulse is subjected to exponential fitting to extract the current decay constant. Apply AC signals with different frequencies to obtain impedance response data. Calculate the slope of impedance change with frequency from the impedance response data and identify the inflection frequency in the curve. Use the slope of impedance change with frequency and the inflection frequency as characteristic parameters of the impedance response curve. Apply a small mechanical disturbance to obtain contact resistance fluctuation data, calculate the resistance fluctuation amplitude and the main frequency components of the fluctuation signal from the contact resistance fluctuation data, and use the resistance fluctuation amplitude and the main frequency components as resistance fluctuation characteristic parameters. The nonlinear response coefficients are calculated by comparing the proportional relationships of responses under different excitation intensities. Based on the voltage settling time, current decay constant, characteristic parameters of the impedance response curve, characteristic parameters of resistance fluctuation, and nonlinear response coefficient, a transient characteristic vector is constructed.

[0032] Specifically, before the probe contacts the wafer pads of the MOSFET under test, a micro-plasma generator is activated. The purpose is to generate high-energy plasma to transiently clean the probe tip, effectively removing oxides, organic contaminants, or other impurities from the tip surface. This ensures a stable and low-impedance electrical contact between the probe and the wafer pads. Transient cleaning refers to completing the cleaning process in a very short time, avoiding damage to the probe material or introducing new contamination. After the transient cleaning of the probe tip is completed, a current pulse with a fast rise time is applied. The purpose is to apply a steep current excitation to the MOSFET in a very short time to fully stimulate the transient electrical response inside the device and at the contact interface, and to acquire the corresponding voltage response data.

[0033] The process of identifying the starting point of the voltage waveform and the time point when the voltage reaches 90% of its steady-state value from the voltage response data aims to accurately capture the time required for the voltage to rise from zero to near its steady-state value; this time is known as the voltage settling time. The voltage settling time can be understood as a key parameter characterizing the capacitive effect and transient response characteristics of the device's contact interface. An exponential fit is performed on the decay portion of the current pulse to accurately extract the characteristics of current decay through a mathematical model, thereby extracting the current decay constant. The current decay constant can be understood as a key parameter characterizing the inductive effect and resistive characteristics of the device.

[0034] In practical applications, AC signals of different frequencies are applied to comprehensively probe the frequency response characteristics of MOSFET devices and their contact interfaces through frequency scanning, and to obtain impedance response data. From this impedance response data, the slope of the impedance change with frequency is calculated to characterize the device's impedance behavior at different frequencies, such as the characteristics of regions dominated by capacitance or inductance. Identifying the inflection point frequency in the curve aims to reveal potential resonance, relaxation, or other frequency-sensitive physical mechanisms within the device or at the contact interface. Using the slope of the impedance change with frequency and the inflection point frequency as characteristic parameters of the impedance response curve aims to comprehensively characterize the device's frequency response characteristics.

[0035] Furthermore, minute mechanical disturbances are applied to simulate the minute mechanical stresses that may exist in actual testing or working environments, thereby stimulating and acquiring contact resistance fluctuation data. From this data, the amplitude of the resistance fluctuation is calculated to quantify the stability of the contact interface. The main frequency components of the fluctuation signal are calculated to reveal the physical mechanisms leading to the resistance fluctuations, such as fretting friction or contact point deformation. The amplitude and main frequency components are used as characteristic parameters of the resistance fluctuation to comprehensively characterize the mechanical stability of the contact interface.

[0036] Furthermore, by comparing the proportional relationships of responses under different excitation intensities, the aim is to evaluate the nonlinear behavior of the device or contact interface and to calculate the nonlinear response coefficient. The nonlinear response coefficient can be understood as a sensitive indicator characterizing internal defects, interface traps, or contact nonlinear effects within the device.

[0037] Through the above technical solutions, this application can extract key transient features characterizing the contact interface state and MOSFET device characteristics from transient electrical responses more comprehensively and accurately. Transient cleaning of the probe tip effectively avoids uncertainties in contact resistance, improving measurement accuracy. Diverse transient electrical excitations (current pulses, AC signals, mechanical disturbances) ensure sufficient excitation of different physical mechanisms of the device and contact interface. Refined data processing methods (such as the extraction of voltage settling time, current decay constant, impedance curve characteristics, resistance fluctuation parameters, and nonlinear response coefficients) enable the separation of unique "fingerprints" associated with specific fault modes from complex responses, thereby constructing more discriminative transient feature vectors. This significantly improves the accuracy and reliability of fault diagnosis, especially in distinguishing between contact component anomalies and device structural defects, providing strong support for quality control and failure analysis of low on-resistance MOSFETs.

[0038] In some embodiments, the step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: The electrical response information is subjected to multi-stage narrowband filtering to eliminate electromagnetic interference and residual mechanical vibration noise within a specific frequency range, thereby obtaining filtered electrical response information. Adaptive baseline drift correction is performed on the filtered electrical response information to compensate for signal baseline changes caused by device noise, thereby obtaining the corrected electrical response information; The corrected electrical response information is subjected to transient signal enhancement processing. By using time-domain averaging or wavelet transform, the signal-to-noise ratio of the real signal at the micro-ohm level is improved, and the enhanced electrical response information is obtained. From the enhanced electrical response information, a method based on both threshold and slope is used to accurately identify the starting and stabilizing points of the voltage waveform to calculate the voltage settling time. A statistical outlier detection method is employed to distinguish between real fluctuations and random noise to calculate characteristic parameters of resistance fluctuations, as well as characteristic parameters of the current decay constant, impedance response curve, and nonlinear response coefficients. Based on the calculation results, a transient feature vector is constructed.

[0039] Specifically, multi-stage narrowband filtering refers to using multiple filters connected in series or parallel to precisely suppress known interference sources (such as power frequency harmonics, mechanical resonant frequencies, etc.) within different frequency ranges. Its purpose is to remove electromagnetic interference and residual mechanical vibration noise to the maximum extent possible, thereby obtaining purer electrical response information. Adaptive baseline drift correction can be understood as a method for dynamically adjusting the signal baseline. It can automatically identify and compensate for signal baseline changes caused by factors such as equipment thermal noise and environmental drift, based on the real-time characteristics of the signal and the background noise level, aiming to ensure the accuracy of the reference for subsequent feature extraction. In practical applications, transient signal enhancement processing specifically uses techniques such as time-domain averaging or wavelet transform to amplify and denoise weak transient electrical signals. For example, time-domain averaging reduces the impact of random noise by averaging the signal after multiple measurements; wavelet transform separates noise and effective components in the signal through multi-scale analysis, aiming to significantly improve the signal-to-noise ratio of the real signal in micro-ohm measurements, enabling the clear identification of minute transient features. Furthermore, in the enhanced electrical response information, a dual threshold and slope-based judgment method is employed. This method combines two conditions: the signal amplitude exceeding a preset threshold and the signal rate of change reaching a specific slope, to accurately determine the starting and stable points of the voltage waveform. The aim is to avoid misjudgments caused by noise or slowly varying signals, thereby accurately calculating the voltage settling time. A statistical outlier detection method is used. By analyzing the statistical distribution of resistance fluctuation data, outlier data points that significantly deviate from the normal range are identified and eliminated, treated as random noise. This allows for more accurate calculation of the resistance fluctuation amplitude and main frequency components. The purpose is to distinguish between genuine contact resistance fluctuations and occasional random noise, ensuring the reliability of the resistance fluctuation characteristic parameters.

[0040] Through the above technical solution, this application can significantly improve the accuracy and robustness of extracting key transient features from electrical response information. Compared with directly extracting features from raw data, this technical solution effectively suppresses external interference through multi-stage narrowband filtering, eliminates the influence of device noise through adaptive baseline drift correction, and improves the signal-to-noise ratio of weak signals through transient signal enhancement processing. This enables high-quality electrical response data to be obtained even in complex testing environments at the micro-ohm level. As a result, the calculation accuracy of key transient features such as voltage settling time, current decay constant, characteristic parameters of impedance response curves, characteristic parameters of resistance fluctuations, and nonlinear response coefficients is greatly improved, effectively avoiding misjudgments caused by noise and drift. This precise feature extraction capability allows the subsequent anomaly detection module to more accurately identify whether the anomaly originates from contact component abnormalities or device structural defects, thus providing a more reliable data foundation for fault diagnosis of low on-resistance shielded gate trench MOSFETs. Ultimately, it enables more precise adjustment of the measured on-resistance value, improving the overall reliability and effectiveness of the diagnostic method.

[0041] In some embodiments, the step of analyzing the transient feature vector based on preset discrimination rules for contact component anomalies and discrimination rules for device structural defects to determine whether the anomaly originates from a contact component anomaly or a device structural defect includes: An initial weight is assigned to each key transient feature in the transient feature vector. The initial weight reflects the predetermined importance of the key transient feature in distinguishing between contact component abnormalities and device structural defects. The weight of each key transient feature is dynamically adjusted based on the degree of deviation between the real-time measured value and the preset normal range. The adjusted weight reflects the actual indicative strength of the key transient feature under the current abnormal state. By combining the adjusted weights and the deviation directions of each key transient feature, a contact component anomaly score and a device structure defect score are calculated. By comparing the relative magnitudes of the contact component anomaly score and the device structure defect score, it can be determined whether the anomaly originates from a contact component anomaly or a device structure defect.

[0042] Specifically, an initial weight is assigned to each key transient feature in the transient feature vector. Before fault diagnosis, based on domain expert experience, historical fault data analysis, or statistical learning methods, a preset value is assigned to each key transient feature in the transient feature vector (e.g., voltage settling time, current decay constant, characteristic parameters of the impedance response curve, resistance fluctuation characteristic parameters, and nonlinear response coefficient). This initial weight aims to quantify the prior importance or indicative ability of the feature in distinguishing between contact component abnormalities and device structural defects. For example, some features may show significant changes when there is poor contact, but not significant changes when there are internal device defects; therefore, their initial weight in distinguishing contact component abnormalities will be higher.

[0043] The dynamic adjustment of the weight of each key transient feature based on its real-time measured value deviating from a preset normal range can be understood as follows: during actual measurement, the real-time measured value of each key transient feature is acquired and compared with a pre-established feature range under normal operating conditions (e.g., the mean and standard deviation obtained through statistical analysis). The degree of deviation can be quantified in various ways, such as calculating the distance between the measured value and the boundary of the normal range, the standardized difference, or the percentage deviation. Based on this degree of deviation, the initial weight of the feature is adjusted in real time. The adjusted weight aims to reflect the actual contribution of the abnormal performance of the feature to determining the source of the fault under the current specific measurement conditions. For example, if the measured value of a feature deviates significantly from the normal range, its weight will be significantly increased, indicating that the feature has a stronger indicative power under the current abnormal state.

[0044] In practical applications, by combining adjusted weights and the deviation direction of each key transient feature, a contact component anomaly score and a device structural defect score are calculated. After obtaining the adjusted weights for each key transient feature, these weights are combined with the deviation direction of each feature (e.g., whether it is higher or lower than the normal value, and the magnitude of the deviation). Through weighted summation, fuzzy logic reasoning, or machine learning models, scores representing the probability of contact component anomalies and the probability of device structural defects are calculated respectively. Considering the deviation direction helps distinguish between different types of anomalies; for example, an excessively long voltage settling time may indicate poor contact, while an abnormal current decay constant may indicate an internal device defect.

[0045] Furthermore, by comparing the relative magnitudes of the contact component anomaly score and the device structure defect score, it is determined whether the anomaly originates from a contact component anomaly or a device structure defect. After calculating the two scores, the final anomaly source is determined by setting a threshold, comparing their magnitudes, or calculating their ratio. For example, if the contact component anomaly score is significantly higher than the device structure defect score, the anomaly is determined to primarily originate from the contact component; conversely, if the contact component anomaly score is significantly higher than the device structure defect score, the anomaly is determined to primarily originate from the device structure defect.

[0046] Through the above technical solution, this application can significantly improve the accuracy and robustness of fault diagnosis for low on-resistance shielded gate trench MOSFETs. Compared with methods relying solely on preset fixed rules, the dynamic weight adjustment mechanism can better adapt to the complexity of different fault modes and measurement conditions, especially in cases of multiple faults or unclear fault characteristics, enabling more effective identification of the dominant anomaly source. This technical solution, by finely evaluating the contribution of each key transient feature and combining its deviation direction, makes the distinction between contact component anomalies and device structural defects more accurate, thereby avoiding production efficiency reduction or product reliability risks caused by misjudgment. Consequently, the measured on-resistance value can be adjusted more accurately, ensuring the reliability of MOSFET performance evaluation.

[0047] In some embodiments, the step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: After the transient electrical excitation is applied, a miniature ultrasonic transmitter is activated above the test contact area of ​​the MOSFET to apply transient acoustic perturbation to the MOSFET at a preset ultrasonic frequency and power. Simultaneously acquire electrical response information under the transient acoustic disturbance; The electrical response information is analyzed in the time and frequency domains to extract specific acoustic-electric coupling features that characterize the transient charge dynamics effect caused by internal defects. These specific acoustic-electric coupling features include the rate of change of voltage response caused by sound waves, specific frequency shifts or broadening of the impedance spectrum, and acoustic modulation depth of resistance fluctuations. Based on a pre-defined discrimination rule for transient charge dynamics effects caused by internal defects, the specific acoustic-electric coupling characteristics are analyzed to determine whether there is a significant contribution from transient charge dynamics effects caused by internal defects. If it is determined that there is a significant contribution from the transient charge dynamics effect caused by internal defects, then the contribution of this effect is separated from the total measurement results. From the separated electrical response information, key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted to obtain the transient feature vector.

[0048] Specifically, after applying transient electrical excitation, a miniature ultrasonic transmitter is activated above the test contact area of ​​the MOSFET, allowing transient acoustic perturbation to be applied to the MOSFET at a preset ultrasonic frequency and power. The miniature ultrasonic transmitter can be understood as a device capable of generating high-frequency mechanical waves. Its purpose is to interact with microstructural defects (such as voids, lattice defects, and interface traps) within the MOSFET through mechanical vibration, thereby exciting or modulating the charge dynamics of these defects. The preset ultrasonic frequency and power are selected to optimize the coupling efficiency between the acoustic waves and specific types of internal defects to produce detectable changes in electrical response.

[0049] The process involves simultaneously acquiring electrical response information under transient acoustic disturbances. While the ultrasonic disturbance is applied, changes in the MOSFET's voltage, current, impedance, and other electrical parameters are precisely recorded. The purpose of this simultaneous acquisition is to ensure that the observed changes in electrical response can be directly attributed to the interaction between the acoustic disturbance and internal defects, thereby avoiding interference from other irrelevant factors.

[0050] In practical applications, time-domain and frequency-domain analyses are performed on the electrical response information to extract specific acoustic-electric coupling characteristics related to internal defects from complex signals. These characteristics include the rate of change of voltage response induced by acoustic waves, specific frequency shifts or broadenings of the impedance spectrum, and the acoustic modulation depth of resistance fluctuations. The rate of change of voltage response induced by acoustic waves refers to the rate or amplitude of transient changes in the MOSFET terminal voltage under the influence of acoustic waves, which may reflect the instantaneous impact of internal defects on the charge transport path. Specific frequency shifts or broadenings of the impedance spectrum may reveal the modulation of the device's equivalent circuit parameters (such as capacitance and resistance) by internal defects, leading to changes in its response characteristics at specific frequencies. The acoustic modulation depth of resistance fluctuations refers to the degree of change in the amplitude of contact resistance or device resistance fluctuations under the influence of acoustic waves, which may be related to the acceleration or suppression of charge trapping-detrapping processes at defects by acoustic waves.

[0051] Furthermore, based on pre-defined discrimination rules for transient charge dynamics effects caused by internal defects, the aforementioned specific acoustic-electric coupling characteristics are analyzed. The purpose is to quantify and identify the contribution of internal defects to the electrical response. These discrimination rules can be established based on historical data, physical models, or machine learning algorithms, and are used to distinguish the differences in acoustic-electric coupling characteristics between normal devices and devices with internal defects.

[0052] If a significant contribution from transient charge dynamics caused by internal defects is determined, this contribution is removed from the overall measurement results. This removal process aims to eliminate or reduce the interference of internal defects on the overall electrical response, thereby obtaining purer electrical response information that primarily reflects the contact interface state and the macroscopic characteristics of the device. From the removed electrical response information, key transient features characterizing the contact state of the aforementioned contact interface and the device characteristics of the aforementioned MOSFET can be extracted more accurately, thus constructing a more precise transient feature vector.

[0053] Through the above technical solution, this application can significantly improve the accuracy and reliability of fault diagnosis for low on-resistance shielded gate trench MOSFETs. Specifically, by introducing transient acoustic disturbances and analyzing their coupling characteristics with the electrical response, this application can effectively identify and quantify the transient charge dynamics effects caused by microscopic defects (such as lattice defects, voids, interface traps, etc.) inside the MOSFET. This capability is not available in traditional methods that rely solely on electrical excitation, because the electrical characteristics of internal defects are often masked or confused by other macroscopic effects. By separating the contribution of internal defects from the overall measurement results, this application ensures that subsequent judgments of contact component anomalies and device structural defects are based on purer and more accurate electrical response information, thereby avoiding misjudgments. As a result, when finally adjusting the measured on-resistance value, it can more accurately reflect the true state of the device, improving the precision of fault diagnosis, which is of great significance for improving the production yield and reliability of MOSFETs.

[0054] In some embodiments, the step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, an ultraviolet light pulse with a specific waveform and amplitude is applied at the contact interface between the probe and the wafer pad. The energy and duration of the ultraviolet light pulse can excite and accelerate the charge trapping-detrapping process in the micro-oxide layer or passivation layer. The electrical response information under the action of the ultraviolet light pulse is acquired synchronously. The electrical response information under the action of the ultraviolet light pulse is analyzed in the time domain and frequency domain. Specific photo-electric coupling features characterizing the interface charge trapping-de-trap effect are extracted. The specific photo-electric coupling features include the transient change of voltage response caused by ultraviolet light, the specific photoinduced frequency shift or broadening of the impedance spectrum, and the optical modulation depth of the resistance fluctuation. Based on the preset discrimination rules for interface charge trapping-detrap effects, the specific opto-electric coupling characteristics are analyzed to determine whether there is a significant contribution from the interface charge trapping-detrap effect. If a significant contribution from the interface charge trapping-detrap effect is determined, this contribution is removed from the total measurement results. Multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the removed electrical response information to construct a transient feature vector.

[0055] Specifically, after the probe contacts the pads of the MOSFET under test, a micro-ultraviolet light pulse is emitted towards the contact interface between the probe and the wafer pads simultaneously with the application of transient electrical excitation. The waveform, amplitude, and duration of this ultraviolet light pulse are precisely controlled to ensure that its energy is sufficient to excite and accelerate the charge trapping-detrapping process in the micro-oxide or passivation layer at the contact interface. For example, the ultraviolet light pulse can have a pulse width in the nanosecond or microsecond range, and its wavelength can be selected between 200 nm and 400 nm to effectively excite electron-hole pairs or interface defect states in the semiconductor material.

[0056] Specifically, during the application of the ultraviolet light pulse, in sync with the transient electrical excitation, the voltage response data, impedance response data, and contact resistance fluctuation data of the contact interface and MOSFET are continuously monitored and recorded to synchronously acquire electrical response information under the ultraviolet light pulse. This data includes additional electrical changes caused by the charge trapping-detrap effect induced by ultraviolet light.

[0057] In practical applications, time-domain and frequency-domain analyses of the electrical response information under ultraviolet (UV) light pulses aim to identify and quantify specific photoelectric coupling characteristics of the interface charge trapping-detrapping effect. For example, in time-domain analysis, transient changes in the voltage response before and after the UV pulse application can be observed, such as rapid voltage rises or falls and their recovery times. In frequency-domain analysis, methods such as Fourier transforms can be used to analyze whether the impedance spectrum exhibits specific frequency shifts or broadening under UV light, which is usually related to changes in the interface charge state. Furthermore, the depth and frequency components of the resistance fluctuation signal modulated by UV light can be analyzed to characterize the impact of the charge trapping-detrapping process on contact resistance stability. These characteristics, including transient changes in the voltage response induced by UV light, specific photoinduced frequency shifts or broadening of the impedance spectrum, and the optical modulation depth of the resistance fluctuation, collectively constitute the photoelectric coupling characteristics characterizing the interface charge trapping-detrapping effect.

[0058] Furthermore, based on pre-defined discrimination rules for interface charge trapping-detrapping effects, the extracted specific photoelectric coupling features are analyzed. These discrimination rules can be established based on historical data, theoretical models, or machine learning algorithms to determine whether there is a significant contribution from interface charge trapping-detrapping effects. For example, a threshold can be set; when the transient change in voltage response caused by ultraviolet light exceeds this threshold, or when a specific frequency shift in the impedance spectrum reaches a certain level, a significant contribution is determined.

[0059] If a significant contribution from the interface charge trapping-detrap effect is determined, this effect's contribution is removed from the overall measurement results. This removal process can be achieved using signal processing techniques, such as establishing a mathematical model of the interface effect and subtracting the model's predicted contribution from the original electrical response information; or using differential measurements, i.e., comparing the differences in electrical responses under ultraviolet light and without ultraviolet light, to separate the contribution of the interface effect. From the removed electrical response information, several key transient features characterizing the contact state of the interface and the device characteristics of the MOSFET are extracted to construct a transient feature vector.

[0060] Through the above technical solution, this application can effectively identify and quantify the charge trapping-detrap effect induced by ultraviolet light at the probe-wafer pad contact interface. Compared with traditional methods that rely solely on transient electrical excitation, this application significantly improves the accuracy of extracting key transient features from electrical response information by actively exciting and removing the contribution of these interface effects. It can more accurately distinguish between measurement deviations caused by interface effects and actual contact component anomalies or device structural defects, thereby avoiding misjudgments and improving the reliability and accuracy of fault diagnosis for low on-resistance shielded gate trench MOSFETs.

[0061] In some embodiments, the step of extracting multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the stripped electrical response information and constructing a transient feature vector includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, multiple miniature piezoelectric sensors and miniature electromagnetic field sensors are arranged at the edge of the contact interface between the probe and the wafer pad to monitor local micro-vibrations and electromagnetic field fluctuations in real time. During the testing process, the probe is periodically scanned to obtain the microscopic fatigue state parameters of the probe material. The electrical response information after stripping is analyzed in the time and frequency domains. At the same time, combined with the data monitored by the micro piezoelectric sensor and the micro electromagnetic field sensor and the micro fatigue state parameters of the probe material, specific environment-probe coupling features characterizing nonlinear drift are extracted. The specific environment-probe coupling features include transient changes in voltage response caused by vibration, impedance spectrum shift caused by electromagnetic field fluctuations, and resistance fluctuation trend caused by probe fatigue. Based on the preset discrimination rules for nonlinear drift caused by the environment or probe material, the specific environment-probe coupling characteristics are analyzed to determine whether there is a significant contribution of nonlinear drift caused by the environment or probe material. If a significant contribution of nonlinear drift caused by the environment or probe material is determined, the contribution of this effect is separated from the total measurement results. Multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the separated electrical response information to construct a transient feature vector.

[0062] Specifically, after the probe contacts the pads of the MOSFET under test, multiple miniature piezoelectric sensors and miniature electromagnetic field sensors are arranged at the edge of the contact interface between the probe and the wafer pads while applying transient electrical excitation. The miniature piezoelectric sensors are used to monitor local micro-vibrations at the contact interface in real time. Their principle is to convert mechanical vibrations into electrical signals using the piezoelectric effect, thereby quantifying the amplitude and frequency of the vibrations. The miniature electromagnetic field sensors are used to monitor electromagnetic field fluctuations around the contact interface in real time, aiming to capture external electromagnetic interference that may affect electrical measurements. These sensors provide real-time data on the stability of the test environment, laying the foundation for subsequent nonlinear drift analysis.

[0063] Furthermore, during testing, the probe undergoes periodic microscopic morphology scanning. This scanning can be achieved using atomic force microscopy (AFM) or other high-resolution imaging techniques to obtain microscopic fatigue state parameters of the probe material, such as the degree of tip wear, deformation, or surface roughness changes. The microscopic fatigue state of the probe directly affects its contact resistance and stability with the pads, thus introducing measurement drift. By periodically monitoring these parameters, the impact of the probe's own state on the measurement results can be quantified.

[0064] Among these, the environment-probe coupling characteristics refer to the changes in electrical response caused by environmental factors (such as vibration and electromagnetic field fluctuations) and probe material fatigue. Specifically, vibration-induced transient changes in voltage response refer to the instantaneous changes in contact resistance or internal charge distribution of the device under the action of micro-vibration, resulting in additional transient fluctuations in the voltage response. Electromagnetic field fluctuation-induced impedance spectrum shifts refer to the possibility that external electromagnetic interference may cause the impedance spectrum of the MOSFET to shift or broaden at specific frequencies. Probe fatigue-induced resistance fluctuation trends refer to the possibility that, with the increase of probe usage time and the accumulation of wear, the contact resistance between the probe and the pad may exhibit a nonlinear drift or fluctuation pattern. These coupling characteristics are key indicators for identifying and quantifying nonlinear drift.

[0065] In practical applications, the specific environment-probe coupling characteristics are analyzed based on preset discrimination rules for nonlinear drift caused by the environment or probe materials. These discrimination rules can be established based on historical data, statistical models, or machine learning algorithms to evaluate whether the extracted environment-probe coupling characteristics significantly deviate from the normal range, thereby determining whether there is a nonlinear drift contribution caused by the environment or probe materials. For example, a threshold can be set; when the rate of change of voltage response caused by vibration, impedance spectrum shift, or resistance fluctuation trend caused by probe fatigue exceeds a specific threshold, it is determined that there is a significant nonlinear drift contribution.

[0066] If a significant contribution of nonlinear drift caused by the environment or probe material is determined, this effect is removed from the total measurement results. This removal process can be achieved through signal processing techniques, such as using vibration and electromagnetic field data monitored by sensors, and by establishing corresponding mathematical models or employing adaptive filtering algorithms to subtract or compensate for these nonlinear drift components from the original electrical response information. This yields purer electrical response information, allowing for more accurate extraction of multiple key transient features characterizing the contact state of the interface and the device characteristics of the MOSFET, ultimately constructing a more precise transient feature vector.

[0067] Through the above technical solution, this application effectively solves the problem of interference to measurement results caused by nonlinear drift due to environmental factors (such as micro-vibration and electromagnetic field fluctuations) and micro-fatigue of probe materials in the fault diagnosis of low on-resistance shielded gate trench MOSFETs. Compared with the technical solution that only strips the interface charge trap-detrap effect, this application further considers and quantifies the influence of the environment and the probe's own state on the measurement. By extracting and stripping the environment-probe coupling features, the purity of the electrical response information and the accuracy of key transient feature extraction are greatly improved. As a result, the constructed transient feature vector can more accurately characterize the true state of the contact components and the inherent characteristics of the device, thereby enabling the anomaly discrimination module to more reliably distinguish between contact component anomalies and device structural defects, ultimately improving the overall accuracy and reliability of low on-resistance MOSFET fault diagnosis.

[0068] For example, suppose that during fault diagnosis of a batch of low on-resistance shielded gate trench MOSFETs, the interface charge trap-detrap effect was first removed using ultraviolet light pulses. However, during subsequent feature extraction, irregular drift in the measurement results of some batches was found, suggesting an environmental or probe problem. At this point, the technical solution of this application is employed.

[0069] Specifically, after the probe contacts the pads of the MOSFET under test, while applying transient electrical excitation, miniature piezoelectric sensors are placed at the four edges of the probe-pad contact interface to monitor micro-vibrations in the vertical and horizontal directions. Simultaneously, three miniature electromagnetic field sensors are arranged around the test stage to monitor AC electromagnetic field fluctuations at 50Hz and 100Hz. Throughout the test, every 100 measurements, a miniature atomic force microscope integrated into the test system is used to scan the probe tip's morphology, recording its wear depth and surface roughness.

[0070] After acquiring the electrical response information stripped of the opto-electric coupling effect, time-domain and frequency-domain analyses are performed by combining data monitored by piezoelectric and electromagnetic field sensors, as well as fatigue parameters obtained from probe morphology scanning. For example, voltage response data is analyzed using Fourier transform to identify voltage fluctuation components related to environmental vibration frequencies (such as 20Hz vibration caused by equipment fans); the spectral shift caused by electromagnetic field fluctuations is quantified by comparing the impedance spectrum with and without electromagnetic field fluctuations; and the impact of probe fatigue on resistance fluctuation trends is assessed by analyzing the correlation between probe wear depth and resistance fluctuation amplitude.

[0071] Based on preset discrimination rules, for example, if the transient change in voltage response caused by vibration exceeds 5 microvolts, or the offset of the impedance spectrum in a specific frequency band exceeds 0.1 ohms, or the slope of the resistance fluctuation trend caused by probe wear is greater than 0.01 ohms / hour, then a significant nonlinear drift contribution caused by the environment or probe material is determined. Once determined, an adaptive Kalman filter algorithm is used, combined with sensor data, to extract these nonlinear drift components from the current electrical response information in real time. For example, the voltage fluctuation caused by vibration is modeled using a Kalman filter and subtracted from the total voltage response, and the impedance shift caused by the electromagnetic field is compensated.

[0072] In this way, the final electrical response information will be purer, which will enable more accurate calculation of voltage settling time, current decay constant, characteristic parameters of impedance response curve, resistance fluctuation characteristic parameters, and nonlinear response coefficients, thus constructing a more reliable transient characteristic vector and providing a solid data foundation for subsequent fault diagnosis.

[0073] In some embodiments, the step of extracting multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the stripped electrical response information and constructing a transient feature vector includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, a miniature atomic force microscope probe is activated to scan the microstructure of the contact interface between the probe and the wafer pad in real time with a preset scanning frequency and mechanical mode, and to obtain the microstructure parameters of the contact interface. During the testing process, the probe is periodically scanned for its microstructure to obtain the micro-fatigue state parameters of the probe material. The electrical response information after stripping is analyzed in the time and frequency domains. At the same time, combined with the microscopic morphology parameters monitored by the micro atomic force microscope probe and the microscopic fatigue state parameters of the probe material, specific morphology-electrical coupling features characterizing the nonlinear memory effect are extracted. The specific morphology-electrical coupling features include the transient change of voltage response caused by morphology change, the impedance spectrum shift caused by morphology change, and the resistance fluctuation trend caused by probe fatigue. Based on the preset discrimination rules for nonlinear memory effects caused by changes in probe micromorphology, the specific morphology-electrical coupling characteristics are analyzed to determine whether there is a significant contribution from nonlinear memory effects caused by changes in probe micromorphology. If it is determined that there is a significant contribution from the nonlinear memory effect caused by the micromorphological changes of the probe, then the contribution of this effect is separated from the total measurement results. From the separated electrical response information, multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted to construct a transient feature vector.

[0074] Specifically, the micro atomic force microscope probe is a high-resolution surface morphology measurement tool that uses a tiny probe tip to interact with the sample surface at the atomic level, thereby acquiring three-dimensional morphology information of the sample surface. In this application, it is used to monitor the microscopic morphology of the probe-wafer pad contact interface in real time, such as the roughness, wear, and changes in contact area at the contact point. The preset scanning frequency and mechanical mode refer to selecting an appropriate scanning speed and probe-surface interaction mode (such as contact mode, tapping mode, or non-contact mode) without affecting electrical testing, to ensure the accuracy of morphology measurement and minimal disturbance to the contact interface.

[0075] The microscopic morphology parameters of the contact interface may include, but are not limited to, surface roughness (Ra, Rz), peak and valley height, contact area, and the depth and width of wear marks. These parameters directly reflect the physical state of the contact interface. The microscopic fatigue state parameters of the probe material, such as the amount of wear at the probe tip, the degree of deformation, and the density of material lattice defects, can be obtained through periodic microscopic morphology scanning or in combination with other material analysis techniques (such as scanning electron microscopy). The purpose is to assess the potential degradation of the probe's physical properties during long-term use.

[0076] In practical applications, the nonlinear memory effect refers to the phenomenon where, due to subtle changes in the microstructure of the probe-pad contact interface during testing (e.g., probe tip wear, plastic deformation of the pad surface), and the fatigue accumulation of the probe material itself, the electrical response is no longer a simple function of instantaneous excitation, but rather related to the historical contact state and morphological evolution path. This effect introduces additional nonlinear errors, affecting the judgment of the true electrical characteristics of the MOSFET.

[0077] The specific morphology-electrical coupling characteristics are key indicators for characterizing this nonlinear memory effect. Specifically, the transient change in voltage response caused by morphology changes refers to the observable transient shift or distortion in the voltage response waveform (e.g., rising edge, falling edge) under transient electrical excitation when the microstructure of the contact interface changes specifically. The impedance spectrum shift caused by morphology changes refers to the identifiable frequency shift or broadening of the impedance spectrum (e.g., resonant frequency, quality factor) of the contact interface under different microstructure states. The resistance fluctuation trend caused by probe fatigue refers to the specific changing patterns in the amplitude and frequency characteristics of the contact resistance as fatigue accumulates in the probe material.

[0078] The discrimination rules for nonlinear memory effects caused by changes in probe microstructure can be constructed based on machine learning algorithms (such as support vector machines and neural networks) or statistical methods (such as principal component analysis and cluster analysis). These rules determine whether a significant nonlinear memory effect contributes to the current measurement by analyzing the correlation between morphology-electrical coupling characteristics and known nonlinear memory effect patterns. For example, a threshold can be set; when the transient change in voltage response caused by morphology changes exceeds a certain preset value, or the impedance spectrum shift reaches a specific range, a nonlinear memory effect is determined to exist.

[0079] Through the above technical solution, this application effectively solves the measurement error problem caused by nonlinear memory effects resulting from probe micromorphology changes and material fatigue in low on-resistance MOSFET fault diagnosis. Compared with technical solutions that only remove interface charge traps and detrap effects, this application further considers the physical dynamic changes of the probe-pad contact interface. By introducing a micro atomic force microscope probe and probe fatigue state monitoring, it achieves accurate extraction and analysis of morphology-electrical coupling characteristics. This allows for a more comprehensive and accurate identification and removal of interference from nonlinear memory effects, significantly improving the purity and reliability of extracting key transient features from electrical response information. Ultimately, this makes the identification of contact component anomalies and device structural defects more precise, thereby improving the overall accuracy and reliability of low on-resistance MOSFET fault diagnosis and providing more refined and reliable technical support for semiconductor device quality control and failure analysis.

[0080] This application also proposes a low on-resistance shielded gate trench MOSFET fault diagnosis system, such as... Figure 2 As shown, a low on-resistance shielded gate trench MOSFET fault diagnosis system 100 is provided, the system comprising: The excitation application module 10 is used to apply a preset transient electrical excitation to the contact interface between the probe and the wafer pad and the MOSFET before measuring the steady-state on-resistance of the MOSFET. The response acquisition module 20 is used to acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation. The electrical response information includes voltage response data, impedance response data, and contact resistance fluctuation data. The parameter extraction module 30 is used to extract key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information, and to construct a transient feature vector. The anomaly detection module 40 is used to analyze the transient feature vector according to preset detection rules for contact component anomalies and device structural defects, so as to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. The resistance adjustment module 50 is used to adjust the measured on-resistance value based on the determined source of the anomaly.

[0081] The system in this application captures the unique "fingerprint" information of contact interfaces and devices under transient excitation by working collaboratively between an excitation application module and a response acquisition module. For example, the excitation application module can generate current pulses with fast rise edges, while the response acquisition module can capture transient characteristics such as voltage settling time and current decay constant, which are highly sensitive to distinguishing contact conditions from internal device structural defects. The introduction of a parameter extraction module and anomaly discrimination module enables the system to intelligently identify the unique patterns of contact component anomalies and device structural defects from these complex transient data. Finally, the resistance adjustment module can make targeted corrections based on the discrimination results, avoiding incorrect adjustments to the on-resistance value or misjudging qualified devices. This systematic approach not only significantly improves the accuracy of fault diagnosis but also provides a more reliable and efficient quality control means for power MOSFET production lines, thereby reducing production costs and market delivery risks.

[0082] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A fault diagnosis method for low on-resistance shielded gate trench MOSFETs, characterized in that, include: Before measuring the steady-state on-resistance of the MOSFET, a pre-set transient electrical excitation is applied to the contact interface between the probe and the wafer pad and the MOSFET. Acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation, the electrical response information including voltage response data, impedance response data and contact resistance fluctuation data; Key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the electrical response information to construct a transient feature vector; Based on preset discrimination rules for contact component anomalies and discrimination rules for device structural defects, the transient feature vector is analyzed to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. Adjust the measured on-resistance value based on the identified source of the anomaly.

2. The fault diagnosis method for low on-resistance shielded gate trench MOSFET according to claim 1, characterized in that, The transient electrical excitation includes a current pulse with a fast rise time, an AC signal with different frequencies, and an electrical response acquired when a small mechanical disturbance is applied; the transient eigenvector includes voltage settling time, current decay constant, characteristic parameters of the impedance response curve, characteristic parameters of resistance fluctuation, and nonlinear response coefficient.

3. The low on-resistance shielded gate trench MOSFET fault diagnosis method according to claim 2, characterized in that, The step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: Before the probe contacts the wafer pad of the MOSFET under test, a micro plasma generator is activated to perform transient cleaning of the probe tip. After the transient cleaning of the probe tip is completed, a current pulse with a fast rise edge is applied and the voltage response data is acquired. From the voltage response data, the starting point of the voltage waveform and the time point when the voltage reaches 90% of its steady-state value are identified to calculate the voltage settling time; the decay part of the current pulse is subjected to exponential fitting to extract the current decay constant. Apply AC signals with different frequencies to obtain impedance response data. Calculate the slope of impedance change with frequency from the impedance response data and identify the inflection frequency in the curve. Use the slope of impedance change with frequency and the inflection frequency as characteristic parameters of the impedance response curve. Apply a small mechanical disturbance to obtain contact resistance fluctuation data, calculate the resistance fluctuation amplitude and the main frequency components of the fluctuation signal from the contact resistance fluctuation data, and use the resistance fluctuation amplitude and the main frequency components as resistance fluctuation characteristic parameters. The nonlinear response coefficients are calculated by comparing the proportional relationships of responses under different excitation intensities. Based on the voltage settling time, current decay constant, characteristic parameters of the impedance response curve, characteristic parameters of resistance fluctuation, and nonlinear response coefficient, a transient characteristic vector is constructed.

4. The fault diagnosis method for low on-resistance shielded gate trench MOSFET according to claim 2, characterized in that, The step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: The electrical response information is subjected to multi-stage narrowband filtering to eliminate electromagnetic interference and residual mechanical vibration noise within a specific frequency range, thereby obtaining filtered electrical response information. Adaptive baseline drift correction is performed on the filtered electrical response information to compensate for signal baseline changes caused by device noise, thereby obtaining the corrected electrical response information; The corrected electrical response information is subjected to transient signal enhancement processing. By using time-domain averaging or wavelet transform, the signal-to-noise ratio of the real signal at the micro-ohm level is improved, and the enhanced electrical response information is obtained. From the enhanced electrical response information, a method based on both threshold and slope is used to accurately identify the starting and stabilizing points of the voltage waveform to calculate the voltage settling time. A statistical outlier detection method is employed to distinguish between real fluctuations and random noise to calculate characteristic parameters of resistance fluctuations, as well as characteristic parameters of the current decay constant, impedance response curve, and nonlinear response coefficients. Based on the calculation results, a transient feature vector is constructed.

5. The low on-resistance shielded gate trench MOSFET fault diagnosis method according to claim 2, characterized in that, The step of analyzing the transient feature vector based on preset discrimination rules for contact component anomalies and discrimination rules for device structural defects to determine whether the anomaly originates from a contact component anomaly or a device structural defect includes: An initial weight is assigned to each key transient feature in the transient feature vector. The initial weight reflects the predetermined importance of the key transient feature in distinguishing between contact component abnormalities and device structural defects. The weight of each key transient feature is dynamically adjusted based on the degree of deviation between the real-time measured value and the preset normal range. The adjusted weight reflects the actual indicative strength of the key transient feature under the current abnormal state. By combining the adjusted weights and the deviation directions of each key transient feature, a contact component anomaly score and a device structure defect score are calculated. By comparing the relative magnitudes of the contact component anomaly score and the device structure defect score, it can be determined whether the anomaly originates from a contact component anomaly or a device structure defect.

6. The fault diagnosis method for low on-resistance shielded gate trench MOSFET according to claim 2, characterized in that, The step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: After the transient electrical excitation is applied, a miniature ultrasonic transmitter is activated above the test contact area of ​​the MOSFET to apply transient acoustic perturbation to the MOSFET at a preset ultrasonic frequency and power. Simultaneously acquire electrical response information under the transient acoustic disturbance; The electrical response information is analyzed in the time and frequency domains to extract specific acoustic-electric coupling features that characterize the transient charge dynamics effect caused by internal defects. These specific acoustic-electric coupling features include the rate of change of voltage response caused by sound waves, specific frequency shifts or broadening of the impedance spectrum, and acoustic modulation depth of resistance fluctuations. Based on a pre-defined discrimination rule for transient charge dynamics effects caused by internal defects, the specific acoustic-electric coupling characteristics are analyzed to determine whether there is a significant contribution from transient charge dynamics effects caused by internal defects. If it is determined that there is a significant contribution from the transient charge dynamics effect caused by internal defects, then the contribution of this effect is separated from the total measurement results. From the separated electrical response information, key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted to obtain the transient feature vector.

7. The low on-resistance shielded gate trench MOSFET fault diagnosis method according to claim 2, characterized in that, The step of extracting key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information and constructing a transient feature vector includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, an ultraviolet light pulse with a specific waveform and amplitude is applied at the contact interface between the probe and the wafer pad. The energy and duration of the ultraviolet light pulse can excite and accelerate the charge trapping-detrapping process in the micro-oxide layer or passivation layer. The electrical response information under the action of the ultraviolet light pulse is acquired synchronously. The electrical response information under the action of the ultraviolet light pulse is analyzed in the time domain and frequency domain. Specific photo-electric coupling features characterizing the interface charge trapping-de-trap effect are extracted. The specific photo-electric coupling features include the transient change of voltage response caused by ultraviolet light, the specific photoinduced frequency shift or broadening of the impedance spectrum, and the optical modulation depth of the resistance fluctuation. Based on the preset discrimination rules for interface charge trapping-detrap effects, the specific opto-electric coupling characteristics are analyzed to determine whether there is a significant contribution from the interface charge trapping-detrap effect. If a significant contribution from the interface charge trapping-detrap effect is determined, this contribution is removed from the total measurement results. Multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the removed electrical response information to construct a transient feature vector.

8. The fault diagnosis method for low on-resistance shielded gate trench MOSFET according to claim 7, characterized in that, The step of extracting multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the stripped electrical response information, and constructing a transient feature vector, includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, multiple miniature piezoelectric sensors and miniature electromagnetic field sensors are arranged at the edge of the contact interface between the probe and the wafer pad to monitor local micro-vibrations and electromagnetic field fluctuations in real time. During the testing process, the probe is periodically scanned to obtain the microscopic fatigue state parameters of the probe material. The electrical response information after stripping is analyzed in the time and frequency domains. At the same time, combined with the data monitored by the micro piezoelectric sensor and the micro electromagnetic field sensor and the micro fatigue state parameters of the probe material, specific environment-probe coupling features characterizing nonlinear drift are extracted. The specific environment-probe coupling features include transient changes in voltage response caused by vibration, impedance spectrum shift caused by electromagnetic field fluctuations, and resistance fluctuation trend caused by probe fatigue. Based on the preset discrimination rules for nonlinear drift caused by the environment or probe material, the specific environment-probe coupling characteristics are analyzed to determine whether there is a significant contribution of nonlinear drift caused by the environment or probe material. If a significant contribution of nonlinear drift caused by the environment or probe material is determined, the contribution of this effect is separated from the total measurement results. Multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted from the separated electrical response information to construct a transient feature vector.

9. The low on-resistance shielded gate trench MOSFET fault diagnosis method according to claim 7, characterized in that, The step of extracting multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the stripped electrical response information, and constructing a transient feature vector, includes: After the probe contacts the pad of the MOSFET under test, while applying the transient electrical excitation, a miniature atomic force microscope probe is activated to scan the microstructure of the contact interface between the probe and the wafer pad in real time with a preset scanning frequency and mechanical mode, and to obtain the microstructure parameters of the contact interface. During the testing process, the probe is periodically scanned for its microstructure to obtain the micro-fatigue state parameters of the probe material. The electrical response information after stripping is analyzed in the time and frequency domains. At the same time, combined with the microscopic morphology parameters monitored by the micro atomic force microscope probe and the microscopic fatigue state parameters of the probe material, specific morphology-electrical coupling features characterizing the nonlinear memory effect are extracted. The specific morphology-electrical coupling features include the transient change of voltage response caused by morphology change, the impedance spectrum shift caused by morphology change, and the resistance fluctuation trend caused by probe fatigue. Based on the preset discrimination rules for nonlinear memory effects caused by changes in probe micromorphology, the specific morphology-electrical coupling characteristics are analyzed to determine whether there is a significant contribution from nonlinear memory effects caused by changes in probe micromorphology. If it is determined that there is a significant contribution from the nonlinear memory effect caused by the micromorphological changes of the probe, then the contribution of this effect is separated from the total measurement results. From the separated electrical response information, multiple key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET are extracted to construct a transient feature vector.

10. A fault diagnosis system for low on-resistance shielded gate trench MOSFETs, characterized in that, The system includes: An excitation application module is used to apply a pre-set transient electrical excitation to the contact interface between the probe and the wafer pad and the MOSFET before measuring the steady-state on-resistance of the MOSFET. The response acquisition module is used to acquire electrical response information of the contact interface and the MOSFET under each transient electrical excitation. The electrical response information includes voltage response data, impedance response data, and contact resistance fluctuation data. The parameter extraction module is used to extract key transient features characterizing the contact state of the contact interface and the device characteristics of the MOSFET from the electrical response information, and to construct a transient feature vector. The anomaly detection module is used to analyze the transient feature vector according to preset detection rules for contact component anomalies and device structural defects, so as to determine whether the source of the anomaly is a contact component anomaly or a device structural defect. The resistance adjustment module is used to adjust the measured on-resistance value based on the identified source of the anomaly.

Images