A Laplace-DLTS Defect Parameter Extraction Method and System

Abstract

This invention provides a Laplace-DLTS defect parameter extraction method and system. The method includes: acquiring capacitance transient data obtained from deep-level transient spectrum testing; preprocessing the capacitance transient data; constructing a Laplace transform kernel function based on the capacitance transient data, and obtaining a discrete Laplace transform model through discretization; introducing regularization constraint parameters to construct an objective function, and obtaining the corresponding emissivity spectrum through numerical solution, wherein the regularization constraint parameters are adaptively determined according to the signal characteristics of the capacitance transient data; automatically identifying peaks in the emissivity spectrum based on preset criteria; and automatically extracting and outputting the physical parameters of the corresponding defects based on the identified effective defect peaks. By introducing adaptive regularization constraint parameters, the degree of regularization can be dynamically adjusted according to the actual signal characteristics of the capacitance transient data, effectively suppressing noise amplification and solution instability problems, and improving the resolution and reliability of the emissivity spectrum analysis results.

Description

Technical Field

[0001] This invention relates to the field of semiconductor material defect characterization technology, and in particular to a Laplace-DLTS defect parameter extraction method and system. Background Technology

[0002] Deep level transient spectroscopy (DLTS) is a commonly used analytical method for characterizing deep level defects in semiconductors. It can obtain defect-related information by analyzing the transient capacitance response of devices and has wide applications in semiconductor materials and device research.

[0003] However, standard DLTS typically performs spectral analysis based on rate windows or correlation functions, which has limited resolution for adjacent deep-level defects. When defect energy levels are close, spectral peaks overlap, or defect characteristics are similar, standard DLTS often struggles to achieve fine differentiation and accurate extraction. To improve the resolution of deep-level defects, Laplace-DLTS performs an inverse Laplace transform on the transient capacitance signal, obtaining a high-resolution spectral distribution related to defect emissivity, thereby enabling fine analysis of similar defect characteristics.

[0004] However, in existing Laplace-DLTS data processing, inverse Laplace inversion typically relies on regularization parameter settings. These parameters often require manual adjustment based on experience or repeated trial calculations. The lack of a unified standard for parameter selection under different data conditions can easily affect the consistency of inversion results and processing efficiency. Furthermore, existing analysis workflows lack sufficient automation in peak identification, spurious peak suppression, and overlapping peak separation. The extraction of defective parameters still largely depends on manual judgment, hindering the formation of a stable and standardized data analysis workflow.

[0005] Therefore, a Laplace-DLTS defect parameter extraction method and system are needed to achieve standardized processing of transient data, automatic analysis of emissivity spectra, and effective extraction of defect physical parameters, thereby improving the reliability, consistency, and automation level of Laplace-DLTS data processing. Summary of the Invention

[0006] The purpose of this invention is to solve the problems in existing Laplace-DLTS data processing, such as the reliance on manual setting of regularization parameters, insufficient inversion adaptability under different transient data conditions, and insufficient automatic identification capability of defect peaks, thereby improving the reliability and processing efficiency of defect parameter extraction.

[0007] A first aspect of the present invention provides a method for extracting Laplace-DLTS defect parameters, comprising:

[0008] Acquire transient capacitance data obtained from deep-level transient spectrum testing;

[0009] The transient capacitance data is preprocessed.

[0010] A Laplace transform kernel function is constructed based on the transient capacitance data, and a discrete Laplace transform model is obtained after discretization.

[0011] A regularization constraint parameter is introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained by numerical solution. The regularization constraint parameter is adaptively determined according to the signal characteristics of the capacitance transient data.

[0012] Automatic peak identification is performed on the emissivity spectrum based on preset criteria;

[0013] Based on the identified valid defect peaks, the physical parameters of the corresponding defects are automatically extracted and output.

[0014] Furthermore, the preprocessing includes at least one or more of the following: baseline correction, noise suppression, normalization, and outlier removal.

[0015] Furthermore, the Laplace transform kernel function is an exponential kernel function, which is constructed based on the multi-exponential decay characteristics of the capacitance transient data.

[0016] Furthermore, the regularization constraint parameters are adaptively determined based on characteristics such as the noise level, signal-to-noise ratio, or fitting residual of the capacitance transient data.

[0017] Furthermore, after constructing the objective function, an inverse Laplace transform operation is performed using a numerical solution method to obtain the corresponding emissivity spectrum.

[0018] Furthermore, the signal characteristics of the capacitor transient data include one or more of the following: multi-exponential decay characteristics, signal amplitude distribution, time constant distribution, or noise distribution pattern.

[0019] Furthermore, the preset criteria include at least one of spectral intensity, spectral width, and spacing between adjacent spectral peaks.

[0020] Furthermore, during the automatic peak identification process, spurious peaks caused by noise are suppressed, and overlapping spectral peaks are separated and identified.

[0021] A second aspect of the present invention provides a Laplace-DLTS defect parameter extraction system, comprising:

[0022] The test module is used to perform deep-level transient spectrum testing on semiconductor devices;

[0023] The data acquisition module is used to receive the transient capacitance data output by the test module;

[0024] The data analysis and processing module is used to preprocess the transient capacitance data; construct an inversion model based on the Laplace transform; introduce regularization constraint parameters into the inversion model to construct an objective function, and obtain the corresponding emissivity spectrum through numerical solution; identify peaks in the emissivity spectrum; and extract and output defect physical parameters.

[0025] Furthermore, the data analysis and processing module includes:

[0026] A data preprocessing unit is used to preprocess the transient capacitance data to reduce noise and interference from abnormal signals.

[0027] The inverse Laplace transform unit is used to construct a Laplace transform kernel function based on the transient capacitance data, and then discretize it to obtain a discrete Laplace transform model. Regularization constraint parameters are introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameters are adaptively determined according to the signal characteristics of the transient capacitance data.

[0028] An emissivity spectrum analysis unit is used to automatically identify peaks in the emissivity spectrum based on preset criteria.

[0029] The defect parameter extraction unit is used to automatically extract the physical parameters of the corresponding defects based on the identified valid defect peaks.

[0030] The result display unit is used to output the physical parameters of the defect.

[0031] Compared to existing technologies, this invention offers at least the following advantages: By introducing adaptive regularization constraint parameters and combining them with the numerical inverse Laplace transform method, the degree of regularization can be dynamically adjusted according to the actual signal characteristics of the capacitance transient data. This helps to mitigate the impact of solution instability and improve the resolution and reliability of the emissivity spectrum analysis results. Combined with automatic peak identification technology, this method achieves an automated analysis process from capacitance transient data processing to defect parameter output, avoiding the subjectivity and errors of manual interpretation and improving the efficiency and reliability of deep-level transient spectrum data analysis. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained as provided without creative effort.

[0033] Figure 1 This is a schematic diagram of the steps of the Laplace-DLTS defect parameter extraction method in one embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the Laplace-DLTS defect parameter extraction system in one embodiment of the present invention. Detailed Implementation

[0035] The present invention will now be described in more detail with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. It should be understood that those skilled in the art can modify the invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as being broadly known to those skilled in the art and is not intended to limit the invention.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0037] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer as explained below. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0038] Example 1

[0039] In a first aspect, this invention provides a method for extracting Laplace-DLTS defect parameters. Please refer to [reference needed]. Figure 1 ,include:

[0040] Obtain transient capacitance data from deep-level transient spectrum testing.

[0041] The transient capacitance data is preprocessed.

[0042] Based on the transient capacitance data, a Laplace transform kernel function is constructed, and a discrete Laplace transform model is obtained after discretization.

[0043] A regularization constraint parameter is introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameter is adaptively determined according to the signal characteristics of the capacitance transient data.

[0044] Automatic peak identification is performed on the emissivity spectrum based on preset criteria.

[0045] Based on the identified valid defect peaks, the physical parameters of the corresponding defects are automatically extracted and output.

[0046] First, it is necessary to obtain the transient capacitance data from deep-level transient spectroscopy (DLS) testing. This data can be obtained by applying excitation pulses to the semiconductor device under multiple temperature conditions using a DLS testing device, and then collecting the transient signal of the corresponding junction capacitance changing over time after the pulse ends.

[0047] Subsequently, the transient data of the capacitor is preprocessed, and a Laplace transform kernel function is constructed based on this data. The resulting discrete Laplace transform model is then obtained through discretization. Discretization can employ uniform sampling or fixed-step discretization methods to transform the continuous mathematical expression into a discrete form suitable for computer processing.

[0048] Based on this, a regularization constraint parameter is introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. This regularization constraint parameter is adaptively determined based on the noise level, signal-to-noise ratio, fitting residual, or other signal characteristics of the capacitor transient data to ensure that the regularization strength matches the characteristics of the input data. After the objective function is constructed, the emissivity spectrum can be obtained using iterative optimization, matrix solving, or other numerical solution methods.

[0049] Next, the emissivity spectrum is automatically peak-identified based on preset criteria. Finally, the physical parameters of the corresponding defects are automatically extracted and output based on the identified valid defect peaks. Physical parameters such as activation energy, trapping cross-section, or defect concentration of the corresponding defects can be extracted based on the identified peak positions, peak characteristics, and test conditions.

[0050] By introducing adaptive regularization constraint parameters and combining them with numerical inverse Laplace transform processing, the resolution and parameter extraction effects of deep-level transient spectrum data analysis are improved, overcoming the limitations of traditional methods in distinguishing similar defects and being sensitive to noise. Simultaneously, the automatic peak identification mechanism reduces reliance on human experience, increases the automation level of data processing, and enhances the reliability of defect parameter extraction.

[0051] Furthermore, the preprocessing includes at least one or more of the following: baseline correction, noise suppression, normalization, and outlier removal.

[0052] Specifically, baseline correction aims to eliminate slow drift or DC offset present in capacitance transient data. This drift may be caused by changes in the test environment, instrument stability, or the characteristics of the semiconductor device itself. Commonly used baseline correction methods include, but are not limited to, polynomial fitting (such as least squares), moving average, wavelet transform baseline correction, or iterative polynomial fitting. By identifying and separating the low-frequency baseline components in the signal, the effective portion of the transient signal can be prevented from being disturbed by the baseline, thus providing a stable zero-point reference for subsequent analysis.

[0053] The noise suppression process is used to reduce random noise and high-frequency interference in transient capacitance data. Noise may originate from test circuits, environmental electromagnetic interference, or quantization errors during data acquisition. In this embodiment, the impact of random noise can be reduced using smoothing or statistical methods. During implementation, appropriate filtering algorithms and parameters must be selected based on the characteristics of the noise and the frequency range of the signal to effectively suppress noise while preserving the characteristic information of the transient signal to the maximum extent possible, thus avoiding signal distortion.

[0054] The normalization process is applied to capacitance data at different temperatures. This process helps eliminate the influence of dimensions, allowing the data to be analyzed on a uniform scale. Common normalization methods include min-max normalization (scaling the data to the [0, 1] or [-1, 1] interval), Z-score normalization (giving the data zero mean and unit variance), or DecimalScaling normalization. Normalization ensures that subsequent Laplace transform model construction and emissivity spectrum calculation are unaffected by fluctuations in the original signal amplitude.

[0055] Outlier removal is used to identify and remove outliers in transient capacitance data caused by transient interference, sensor malfunction, or measurement errors. These outliers can have a significant negative impact on data analysis results. Common removal methods include statistical methods (such as the 3σ criterion and box plots), distance-based methods (such as the Local Outlier Factor (LOF)), or density-based methods. In practice, it is necessary to set reasonable thresholds or criteria to distinguish between normal fluctuations and true outliers to avoid mistakenly deleting valid data or retaining harmful data.

[0056] Furthermore, the Laplace transform kernel function is an exponential kernel function, which is constructed based on the multi-exponential decay characteristics of the capacitance transient data.

[0057] Specifically, in deep-level transient spectroscopy (DLTS) measurements, the capacitance transient data of semiconductor devices typically exhibits an exponential decay over time, determined by the electron or hole emission processes of deep-level defects. Therefore, setting the Laplace transform kernel function to an exponential kernel function ensures its mathematical form closely matches the physical decay mechanism of capacitance transient data. This matching helps to more accurately describe and analyze capacitance transient signals, providing a physically clear and mathematically applicable foundation for subsequent inverse transforms. An exponential kernel function can typically be expressed in the form exp(-st), where s represents the emission rate and t represents time, directly corresponding to the physical model of DLTS signals. Furthermore, actual capacitance transient data often consists of the superposition of multiple deep-level defect signals, exhibiting a multi-exponential decay characteristic. Constructing an exponential kernel function based on this characteristic means that multiple exponentially decaying components that may exist in the signal are fully considered when designing or selecting the kernel function. This construction method enables the kernel function to accurately capture the rich defect information contained in the signal, thereby improving the inverse transform's ability to distinguish different defect peaks.

[0058] Furthermore, the regularization constraint parameters are adaptively determined based on characteristics such as the noise level, signal-to-noise ratio, or fitting residual of the capacitance transient data.

[0059] Specifically, because this problem is often ill-conditioned—meaning it is highly sensitive to small perturbations (such as noise) in the input data—it can lead to unstable or non-unique solutions. The role of regularization constraint parameters is to introduce additional constraints to stabilize the solution process, making the solution more physically meaningful and robust. It avoids overfitting noise by balancing data fitting terms and constraints such as the smoothness or sparsity of the solution in the objective function, thereby obtaining a more reliable emissivity spectrum. The adaptive determination means that the regularization constraint parameters are not fixed values ​​but dynamically adjusted according to the specific characteristics of the input data. This dynamic adjustment allows the regularization strength to match the data characteristics, avoiding the loss of signal details due to overly strong regularization or the inability to effectively suppress noise due to overly weak regularization.

[0060] Furthermore, after constructing the objective function, an inverse Laplace transform operation is performed using a numerical solution method to obtain the corresponding emissivity spectrum.

[0061] Specifically, the numerical solution method refers to using computer algorithms to find approximate solutions to equations or optimization problems in discrete space through mathematical methods such as iteration and approximation. In the inverse Laplace transform, due to its ill-conditioned nature, direct analytical solutions are often infeasible or unstable; therefore, numerical methods are necessary to obtain reliable solutions. The numerical solution method enables the stable and accurate retrieval of physically meaningful emissivity spectra from transient capacitance data, even in the presence of noise and data incompleteness.

[0062] Furthermore, the signal characteristics of the capacitor transient data include one or more of the following: multi-exponential decay characteristics, signal amplitude distribution, time constant distribution, or noise distribution pattern.

[0063] Specifically, the signal characteristics of capacitance transient data refer to the inherent properties and behavioral patterns exhibited by capacitance transient data in the time domain. These characteristics directly reflect the physical mechanisms of deep-level defects in semiconductor materials and the environmental influences during the measurement process. Multi-exponential decay characteristics refer to the fact that capacitance transient data typically exhibits the superposition of multiple exponential decay processes, with each exponential decay term corresponding to the emission process of a deep-level defect. Analyzing its decay constant, amplitude, and number can reveal the type and concentration information of defects. This characteristic can be obtained through fitting analysis of transient data or by using a multi-exponential fitting algorithm. Signal amplitude distribution refers to the amplitude variation range and statistical distribution of the capacitance transient signal at different time points, such as the initial amplitude, decay amplitude, and overall dynamic range of the signal. This helps to evaluate the signal strength and effectiveness and provides a basis for subsequent normalization or quantization processing. Time constant distribution refers to the decay time constants and their relative distribution characteristics corresponding to each decay component in the capacitance transient data obtained from deep-level transient spectrum testing. This distribution can reflect the differences in the speed of response processes of different defects in the transient signal and can serve as one of the bases for adaptively determining regularization constraint parameters. Noise distribution patterns refer to the statistical characteristics of noise in transient capacitance data, such as the type of noise (e.g., Gaussian noise, Poisson noise, 1 / f noise), the amplitude of the noise, the frequency range, and its distribution pattern or statistical regularity over time.

[0064] Furthermore, the preset criteria include at least one of spectral intensity, spectral width, and spacing between adjacent spectral peaks.

[0065] In this context, spectral intensity refers to the amplitude or height of a peak in the emissivity spectrum, reflecting the concentration or density of defects. In practical applications, a minimum spectral intensity threshold can be set to filter out low-intensity spurious peaks caused by random noise; only peaks with sufficient signal strength are considered valid defect peaks. For example, the maximum height of the peak relative to the baseline can be calculated, or the integral area of ​​the peak can be calculated and compared with a preset threshold.

[0066] Spectral width refers to the width of a peak in the emission spectrum, usually measured by the full width at half maximum (FWHM) or the width at another specific height. Spectral width can provide information about the discreteness of defect energy levels or the uniformity of defect distribution. For typical deep-level defects, their emission spectrum peaks typically have a specific range of spectral widths. Therefore, a reasonable spectral width range can be set as a criterion to exclude peaks that are too narrow (potentially caused by transient noise) or too wide (potentially caused by multiple unresolved defects or measurement artifacts), thereby improving the accuracy of identification.

[0067] The distance between adjacent peaks in an emission spectrum refers to the distance between two adjacent peaks, such as the difference in their peak positions. In some cases, different defects may have very close energy levels, causing their emission spectrum peaks to overlap. By introducing the distance between adjacent peaks as a criterion, we can assess whether adjacent peaks are sufficiently separated to be identified as independent defects. For example, a minimum distance threshold can be set; if the distance between two adjacent peaks is less than this threshold, further analysis or the use of specific deconvolution algorithms for separation and identification may be required, or the peak may be considered a composite peak.

[0068] Furthermore, during the automatic peak identification process, spurious peaks caused by noise are suppressed, and overlapping spectral peaks are separated and identified.

[0069] Specifically, suppressing spurious peaks caused by noise aims to eliminate false signals in the emissivity spectrum resulting from measurement noise or data processing errors, thereby ensuring the accuracy and reliability of subsequent defect parameter extraction. This suppression can be achieved through various methods. For example, a spectral intensity threshold can be set, and peaks below this threshold can be considered spurious peaks caused by noise and eliminated; alternatively, peaks that do not conform to the characteristics of true defect spectral peaks can be suppressed by considering the shape characteristics of the peaks, such as their symmetry and width.

[0070] Meanwhile, the purpose of separating and identifying overlapping spectral peaks is to decompose the composite peak formed by the superposition of multiple defect emissivity spectral peaks on the frequency axis into independent single peaks. This allows for accurate identification of the presence of each defect and lays the foundation for subsequent precise extraction of their respective physical parameters. Various numerical deconvolution techniques can be employed for separating and identifying overlapping spectral peaks. For example, Gaussian functions, Lorentz functions, or other suitable mathematical models can be used to fit the composite peak, and iterative optimization algorithms can be used to find the optimal fitting parameters, thereby decomposing the composite peak into multiple independent single peaks.

[0071] Furthermore, the physical parameters include at least one of activation energy, capture cross section, or defect concentration.

[0072] Specifically, activation energy is a key physical parameter of deep-level defects, characterizing the energy depth of the defect level relative to the conduction or valence band edge. Activation energy extraction is typically based on Arrhenius plot analysis, which involves measuring the defect emissivity at different temperatures and fitting the result to the logarithm of the emissivity versus temperature. After obtaining the emissivity spectrum, the peak emissivity at different temperatures can be obtained by changing the test temperature and repeating the above method, and then the activation energy can be calculated using the Arrhenius equation. The trapping cross-section is a physical quantity that measures the ability of a defect to trap charge carriers, reflecting the effective area of ​​interaction between the defect and the charge carriers. Simultaneously with activation energy extraction, the trapping cross-section can be further calculated using the intercept of the Arrhenius plot. This usually requires derivation using known parameters such as the effective density of states of the semiconductor material and the thermal velocity of charge carriers. Defect concentration represents the number of defects per unit volume and is an important indicator for evaluating the quality of semiconductor materials and the reliability of devices. Defect concentration extraction is typically based on the relationship between the amplitude of the capacitance transient signal and the defect concentration. After obtaining the emissivity spectrum and identifying the effective defect peaks, the defect concentration can be calculated using relevant formulas by analyzing the intensity of the corresponding defect peaks or the initial amplitude of the transient capacitance signal, combined with parameters such as the device's doping concentration and junction area.

[0073] Example 2

[0074] A second aspect of the present invention provides a Laplace-DLTS defect parameter extraction system, please refer to... Figure 2 ,include:

[0075] The test module is used to perform deep-level transient spectrum testing on semiconductor devices;

[0076] The data acquisition module is used to receive the transient capacitance data output by the test module;

[0077] The data analysis and processing module is used to preprocess the transient capacitance data; construct an inversion model based on the Laplace transform; introduce regularization constraint parameters into the inversion model to construct an objective function, and obtain the corresponding emissivity spectrum through numerical solution; identify peaks in the emissivity spectrum; and extract and output defect physical parameters.

[0078] The data analysis and processing module includes:

[0079] The data preprocessing unit is used to preprocess the transient capacitance data to reduce noise and interference from abnormal signals.

[0080] The inverse Laplace transform unit is used to construct a Laplace transform kernel function based on the transient capacitance data, and then discretize it to obtain a discrete Laplace transform model. Regularization constraint parameters are introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameters are adaptively determined according to the signal characteristics of the transient capacitance data.

[0081] The emissivity spectrum analysis unit is used to automatically identify peaks in the emissivity spectrum based on preset criteria.

[0082] The defect parameter extraction unit is used to automatically extract the physical parameters of the corresponding defects based on the identified valid defect peaks.

[0083] The result display unit is used to output the physical parameters of the defect.

[0084] In this system, the data preprocessing unit first performs baseline correction, noise suppression, and outlier removal on the capacitance transient data obtained from deep-level transient spectrum testing to reduce noise interference. The Laplace inverse transform unit constructs an exponential kernel function model based on the processed data and uses adaptively determined regularization constraint parameters based on signal characteristics to construct the objective function and perform numerical solution to obtain the corresponding emissivity spectrum. The emissivity spectrum analysis unit automatically identifies valid defect peaks based on preset criteria including at least spectral intensity, spectral width, and the spacing between adjacent peaks, while suppressing noise spurious peaks and separating overlapping peaks. The defect parameter extraction unit automatically calculates physical parameters such as activation energy, trapping cross-section, or defect concentration based on the identification results. Finally, the results display unit outputs the defect physical parameters and corresponding analysis results. This technical solution achieves an automated analysis process from capacitance transient data processing to defect parameter output, effectively overcoming the reliance on manual experience in traditional methods and improving the reliability and engineering applicability of defect parameter extraction.

[0085] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A method for extracting defect parameters in Laplace-DLTS, characterized in that, include: Acquire transient capacitance data obtained from deep-level transient spectrum testing; The transient capacitance data is preprocessed. Based on the preprocessed transient capacitance data, a Laplace transform kernel function is constructed, and a discrete Laplace transform model is obtained after discretization. A regularization constraint parameter is introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameter is adaptively determined based on one or more of the following: noise level, signal-to-noise ratio, fitting residual, signal amplitude distribution, time constant distribution, or noise distribution law of the capacitor transient data, so that the regularization intensity matches the signal characteristics of the capacitor transient data. The emissivity spectrum is automatically peaked based on at least two criteria among spectral intensity, spectral width, and the spacing between adjacent spectral peaks. Noise spurious peaks below the spectral intensity threshold are suppressed, and overlapping spectral peaks with the spacing between adjacent spectral peaks less than a preset spacing threshold are separated and identified. Based on the spectral peak position, spectral characteristics, and test conditions of the identified valid defect peaks, the physical parameters of the corresponding defects are automatically extracted and output.

2. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, The preprocessing includes at least one or more of the following: baseline correction, noise suppression, normalization, and outlier removal.

3. The Laplace-DLTS defect parameter extraction method as described in claim 2, characterized in that, The baseline correction includes identifying and separating the low-frequency baseline components in the capacitance transient data through one of the following methods: polynomial fitting, moving average method, wavelet transform baseline correction, or iterative polynomial fitting; the outlier removal includes using methods based on the 3σ criterion, box plot method, local outlier factor, or density-based method.

4. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, The Laplace transform kernel function is an exponential kernel function, which is constructed based on the multi-exponential decay characteristics of capacitance transient data.

5. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, After constructing the objective function, the inverse Laplace transform operation is performed using a numerical solution method to obtain the corresponding emissivity spectrum.

6. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, The signal characteristics of the capacitor transient data include one or more of the following: multi-exponential decay characteristics, signal amplitude distribution, time constant distribution, or noise distribution pattern.

7. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, The spectral intensity criterion includes comparing the peak height or peak integral area in the emissivity spectrum with a preset spectral intensity threshold, and filtering out spectral peaks that are lower than the preset spectral intensity threshold. The spectral width criterion includes determining whether a spectral peak is within a preset spectral width range based on the full width at half maximum (FWHM), and excluding spectral peaks that are below or above the preset spectral width range. The criterion for determining the spacing between adjacent spectral peaks includes comparing the difference between the peak positions of adjacent spectral peaks with the preset spacing threshold. When the difference is less than the preset spacing threshold, the corresponding spectral peak is separated and identified or regarded as a composite peak.

8. The Laplace-DLTS defect parameter extraction method as described in claim 1, characterized in that, The separation and identification of overlapping spectral peaks includes fitting the composite peak with a Gaussian function or a Lorentz function, and obtaining multiple independent single peaks through iterative optimization.

9. A Laplace-DLTS defect parameter extraction system, characterized in that, include: The test module is used to perform deep-level transient spectrum testing on semiconductor devices; The data acquisition module is used to receive the transient capacitance data output by the test module; The data analysis and processing module is used to preprocess the transient capacitance data; Constructing an inversion model based on the Laplace transform; In the inversion model, a regularization constraint parameter is introduced to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameter is adaptively determined based on one or more of the following: noise level, signal-to-noise ratio, fitting residual, signal amplitude distribution, time constant distribution, or noise distribution pattern of the capacitor transient data, so that the regularization intensity matches the signal characteristics of the capacitor transient data. Automatic peak identification is performed on the emissivity spectrum based on at least two criteria among spectral intensity, spectral width, and the spacing between adjacent spectral peaks. Noise spurious peaks below the spectral intensity threshold are suppressed, and overlapping spectral peaks with a spacing between adjacent spectral peaks less than a preset spacing threshold are separated and identified. Defect physical parameters are extracted and output.

10. The Laplace-DLTS defect parameter extraction system as described in claim 9, characterized in that, The data analysis and processing module includes: A data preprocessing unit is used to preprocess the transient capacitance data to reduce noise and interference from abnormal signals. The inverse Laplace transform unit is used to construct a Laplace transform kernel function based on the transient capacitance data, and then discretize it to obtain a discrete Laplace transform model. Regularization constraint parameters are introduced into the discrete Laplace transform model to construct an objective function, and the corresponding emissivity spectrum is obtained through numerical solution. The regularization constraint parameters are adaptively determined according to the signal characteristics of the transient capacitance data. An emissivity spectrum analysis unit is used to automatically identify peaks in the emissivity spectrum based on preset criteria. The defect parameter extraction unit is used to automatically extract the physical parameters of the corresponding defects based on the identified valid defect peaks. The result display unit is used to output the physical parameters of the defect.

Images