Integrated circuit defect detection method and system using optical means

Abstract

The invention relates to the technical field of integrated circuit defect detection, and discloses an integrated circuit defect detection method and system using an optical means, and the method comprises the following steps: controlling the emission of incident light, and controlling the polarization state of the incident light, enabling the incident light to irradiate a detection area of the wafer with the integrated circuit in a preset polarization direction; returning light rays reflected or scattered from the detection area are collected, polarization state parameters of the returning light rays are calculated, and the polarization state parameters comprise the polarization degree and the polarization angle; and comparing the polarization state parameter with a reference polarization state parameter of a preset defect-free area so as to judge whether a defect exists on the wafer with the integrated circuit or not. According to the method, high-sensitivity and high-precision defect detection and classification are realized through polarization state analysis, and the depth of defect analysis is enhanced while the detection sensitivity and reliability are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated circuit defect detection, in particular to an integrated circuit defect detection method and system using optical means. BACKGROUND

[0002] In advanced integrated circuit manufacturing environment, detection of microscopic defects is a critical step to ensure product reliability and production yield. A typical, high-throughput defect detection system usually employs sophisticated optical means, with a high-brightness LED light source array providing stable and uniform illumination, light rays passing through a carefully designed objective lens group for capturing high-resolution images of wafer surface, followed by image processing units to identify various abnormalities. However, with the continuous pursuit of higher performance and lower power consumption in modern integrated circuits, manufacturers have generally adopted new materials such as low-k materials. These new materials are highly sensitive to the wavelength of incident light, and even a slight change in the spectral purity of the detection light source can significantly change the way light interacts with and penetrates these layers, affecting the clarity and contrast of internal structural features.

[0003] At the same time, during long-term operation of the detection equipment, the semiconductor material in the LED light source array will experience performance degradation, manifested as a slight shift in the peak emission wavelength. When such spectrally shifted light sources interact with new low-k materials that are sensitive to the spectrum, the image contrast of internal defects is severely lost, making it difficult for existing optical detection methods to reliably identify subsurface microcavities and cracks. For example, a long-serving detection equipment, its LED light source array may experience a peak wavelength shift of about 5 nanometers. When such shifted light is used to detect new low-k material wafers, due to the high sensitivity of low-k materials to specific wavelengths, the attenuation and scattering behavior of light when penetrating the passivation layer and the underlying low-k dielectric layer will change unexpectedly, resulting in a reduction in the amount of light returned from internal features, thus greatly weakening the image intensity difference of internal defects such as tiny cavities or hairline cracks, almost blending into the background noise.

[0004] This loss of contrast makes the defect recognition logic established and trained based on standard images essentially ineffective, and traditional electrical probe testing methods and mechanical detection techniques also cannot effectively detect these subsurface defects. In the face of rising internal defect miss rate, field engineers try to adjust the exposure time, LED light source intensity or replace the objective lens group, but these conventional adjustments cannot solve the fundamental problem, i.e. there is a mismatch between the "spectral quality" of the light source and the "spectral sensitivity" of the new low-k material. Image processing personnel also confirm that advanced image enhancement programs cannot reconstruct the key defect signals that have been lost or severely attenuated in the front-end light-material interaction and image acquisition stage. SUMMARY

[0005] To address the shortcomings of existing technologies, this application discloses an integrated circuit defect detection method and system using optical means. It aims to solve the technical problem that existing integrated circuit defect detection methods are unable to reliably identify subsurface micro-voids and cracks when faced with new materials and declining light source performance, leading to an increased defect false detection rate.

[0006] The technical solution of this application is as follows: In a first aspect, this application discloses a method for detecting defects in integrated circuits using optical means, comprising: The incident light is controlled to be emitted and its polarization state is controlled so that the incident light shines on the detection area of ​​the wafer with integrated circuits in a preset polarization direction. Collect the reflected or scattered rays from the detection area and calculate the polarization state parameters of the returned rays, including the degree of polarization and the polarization angle; The polarization state parameters are compared with the reference polarization state parameters of a preset defect-free region to determine whether there are defects on the wafer containing integrated circuits. The step of comparing the polarization state parameter with the reference polarization state parameter of a preset defect-free region to determine whether there are defects on the wafer with the integrated circuit includes: For each pixel in the detection area, calculate the difference index between its polarization state parameter and the reference polarization state parameter; If the difference index is greater than a preset tolerance threshold, it is determined that the pixel has a potential defect in the corresponding area on the wafer with integrated circuits, and it is marked as a potential defect pixel. If the difference index is less than or equal to the preset tolerance threshold, then the corresponding area is determined to be defect-free. The preset tolerance threshold is determined through the following steps: By performing polarized light detection on wafers with integrated circuits containing subsurface defects of different types and sizes, a large amount of actual defect sample data on the degree of polarization and polarization angle of defect regions was collected. Analyze the actual defect sample data to determine the typical range of polarization degree and polarization angle changes caused by different defect types, and determine the preset tolerance threshold based on the typical range.

[0007] Secondly, this application also discloses an integrated circuit defect detection system using optical means, for performing the aforementioned integrated circuit defect detection method using optical means, the system comprising: The polarized light generation module is used to control the emission of incident light and control the polarization state of the incident light so that the incident light illuminates the detection area of ​​the wafer with integrated circuits in a preset polarization direction. The polarization parameter calculation module is used to collect the returned light reflected or scattered from the detection area and calculate the polarization state parameters of the returned light, including the degree of polarization and the polarization angle. The defect detection module is used to compare the polarization state parameters with the reference polarization state parameters of a preset defect-free region to determine whether there are defects on the wafer containing integrated circuits.

[0008] In summary, this application discloses an integrated circuit defect detection method and system utilizing optical means. The method involves controlling the polarization state of incident light to illuminate the wafer detection area and collecting the polarization state parameters (including degree of polarization and polarization angle) of the returned light. These parameters are then compared with reference polarization state parameters of a defect-free area to determine the presence of defects on the wafer. This method effectively addresses the problem in existing technologies where the high spectral sensitivity of novel low-k materials and the long-term operation of LED light source arrays cause peak emission wavelength drift, resulting in severe loss of contrast in internal defect images. This makes it difficult for traditional optical detection methods to reliably identify subsurface microvoids and cracks. By introducing polarization state parameters as the basis for defect detection, this application can capture weak signal changes that traditional intensity detection methods cannot identify. Even under conditions of light source spectral drift or complex material properties, it can effectively distinguish defective and defect-free areas, significantly improving the detection sensitivity and accuracy of subsurface defects. This reduces the false negative rate of integrated circuits and improves product reliability and production yield. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating an integrated circuit defect detection method using optical means, provided as an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of an integrated circuit defect detection system using optical means, provided as an embodiment of this application.

[0011] Labeling Explanation: 210, Polarized Ray Generation Module; 220, Polarization Parameter Calculation Module; 230, Defect Judgment Module. Detailed Implementation

[0012] 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.

[0013] 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.

[0014] In advanced integrated circuit manufacturing environments, the detection of microscopic defects is a critical step in ensuring product reliability and production yield. Traditional optical inspection systems experience performance degradation in the semiconductor materials within their LED light source arrays over long-term operation, manifesting as a slight shift in peak emission wavelength. When this spectrally shifted light source interacts with novel low-k materials sensitive to spectral density, it leads to a severe loss of image contrast for internal defects, making it difficult for existing optical inspection methods to reliably identify subsurface microvoids and cracks. This loss of contrast renders defect recognition logic built and trained based on standard images essentially ineffective, and traditional electrical probe testing methods and mechanical inspection techniques are also ineffective in detecting these subsurface defects.

[0015] Firstly, please see Figure 1 This application proposes a method for detecting defects in integrated circuits using optical means, comprising: S1. Control the emission of incident light and control the polarization state of the incident light so that the incident light irradiates the detection area of ​​the wafer with integrated circuits in a preset polarization direction. S2. Collect the returned light reflected or scattered from the detection area and calculate the polarization state parameters of the returned light, including the degree of polarization and the polarization angle; S3. Compare the polarization state parameters with the reference polarization state parameters of the preset defect-free region to determine whether there are defects on the wafer with integrated circuits.

[0016] This application, by precisely controlling the polarization state of the incident light and analyzing the polarization state parameters of the light returning from the wafer detection area, can effectively identify subsurface defects that are difficult to detect by traditional intensity or spectral analysis methods, thereby overcoming the problem of loss of defect contrast caused by light source spectral drift and the spectral sensitivity of new materials in the prior art.

[0017] The integrated circuit defect detection method of this application first requires controlling the emitted incident light and its polarization state, so that the incident light illuminates the detection area of ​​the wafer containing the integrated circuit with a preset polarization direction. The polarization state control of the incident light can be achieved in various ways. For example, a broadband light source can be used, whose emitted light first passes through a linear polarizer to generate linearly polarized light. Subsequently, this linearly polarized light can further pass through a rotatable half-wave plate. By precisely adjusting the rotation angle of the half-wave plate, the vibration direction of the linearly polarized light can be rotated to any preset angle, thereby achieving precise control of the polarization direction of the incident light. As another implementation method, an electrically controlled solid-state polarization rotation device, such as a liquid crystal variable retarder or an electro-optic crystal, can also be used. These devices can quickly and accurately control the rotation of the polarization direction by applying different voltage sequences, thereby achieving dynamic adjustment of the polarization state of the incident light. In this way, it can be ensured that the incident light illuminates the detection area of ​​the wafer with a specific, preset polarization direction, providing a stable input for subsequent polarization state analysis.

[0018] Next, the returned light reflected or scattered from the detection area needs to be collected, and the polarization state parameters of the returned light need to be calculated. These parameters include the degree of polarization and the polarization angle. The collection of returned light and the calculation of the polarization state parameters can be achieved using a polarization analyzer. The polarization analyzer can acquire returned light reflected or scattered from the detection area and obtain multiple images with different polarization directions. These images contain intensity information of the returned light in different polarization directions. For example, the polarization analyzer can sequentially acquire images with horizontal polarization, vertical polarization, 45-degree polarization, and 135-degree polarization directions. Based on these multiple images with different polarization directions, the polarization state parameters of each pixel can be calculated. Specifically, the Stokes vector can be calculated based on the light intensity value of each pixel in the multiple images with different polarization directions. The Stokes vector is a quadruple, typically including the total light intensity, the difference between the horizontal and vertical polarization components, and the difference between the 45-degree and 135-degree polarization components. Based on the calculated Stokes vector, the degree of polarization and the polarization angle of each pixel can be further calculated. The degree of polarization can be calculated from the components of the Stokes vector, representing the extent of light polarization; the polarization angle can be calculated from the components of the Stokes vector, representing the polarization direction.

[0019] Finally, the polarization state parameters are compared with the reference polarization state parameters of a preset defect-free region to determine whether defects exist on the wafer containing integrated circuits. Obtaining the reference polarization state parameters is crucial. In the actual inspection process, for each pixel in the inspection area, the difference index between its polarization state parameters and the reference polarization state parameters is calculated. The difference index can be the difference in polarization degree, the difference in polarization angle, or a combination of both. If the difference index is greater than a preset tolerance threshold, it is determined that the corresponding region on the wafer containing integrated circuits has a potential defect, and it is marked as a potential defect pixel. Through this comparison, regions with significantly different polarization state parameters from the defect-free region can be effectively identified, thus indicating potential defects.

[0020] This application's integrated circuit defect detection method effectively solves the problem of defect contrast loss caused by spectral drift of the light source and the spectral sensitivity of novel materials in existing technologies by controlling the polarization state of incident light and analyzing the polarization state parameters of the returned light. Specifically, in advanced integrated circuit manufacturing, novel materials such as low-k dielectrics are highly sensitive to the spectral purity of traditional optical detection, causing even small spectral drifts of the light source to severely affect the image contrast of defects. This application precisely controls the polarization direction of the incident light and uses a polarization analyzer to collect the returned light, calculating its polarization degree and polarization angle, among other polarization state parameters. These polarization state parameters are more sensitive to the microstructure and subsurface defects of materials, and can reveal the presence of defects even when traditional intensity image contrast is insufficient. For example, when incident light illuminates the wafer surface with a specific polarization direction, if subsurface voids or cracks exist, the light will undergo unexpected polarization state changes during penetration and reflection, resulting in the polarization degree and polarization angle of the returned light being significantly different from those of the light from the defect-free region. By comparing these polarization state parameters with preset reference polarization state parameters, these minute polarization anomalies can be accurately identified, thereby determining the presence of defects. This method avoids the strict dependence on the spectral purity of the light source and instead utilizes the polarization state change caused by the interaction between light and the defect structure, thereby overcoming the limitations of traditional methods in detecting subsurface defects in novel materials and improving the sensitivity and reliability of defect detection.

[0021] In some embodiments described above, the method of controlling the emitted incident light and its polarization state is proposed, so that the incident light illuminates the detection area of ​​the wafer containing integrated circuits with a preset polarization direction. However, in actual implementation, how to accurately and flexibly control the polarization state of the incident light to adapt to different detection needs or optimize detection results is a problem that needs further consideration. If only a simple polarization control method is used, it may be difficult to achieve high-precision polarization direction adjustment, or it may be inconvenient to operate in automated detection. In this regard, this application further proposes a more specific and controllable incident light polarization state control scheme, which achieves precise control of the incident light polarization state by having the incident light sequentially pass through a linear polarizer and a polarization rotation device to illuminate the detection area.

[0022] The concept of "establishing a correlation model or empirical relationship between a reference region and a complex region" described in this application has very practical and important application value in the field of semiconductor manufacturing. The following will illustrate its application in detail with a specific semiconductor wafer inspection scenario.

[0023] Application Scenario: Suppose we are manufacturing advanced integrated circuit wafers containing numerous complex structures at the micrometer or even nanometer scale, such as transistor gates, interconnect layers, and multilayer dielectric film stacks. On these wafers, there are two types of regions: 1. Reference polarization state parameter region (reference area): This is typically a region on the wafer that has not undergone complex patterning, such as a test structure in a scribe line, or a specially designed test area with a simple and uniform thin film layer. The optical properties of these regions are relatively simple, and their polarization state parameters (such as average phase delay Δ and fast axis azimuth Ψ) are easy to measure and interpret accurately, and they are sensitive to changes in process parameters (such as film thickness, refractive index, and uniformity).

[0024] 2. Polarization State Parameter Region (Complex Region): This refers to the actual device region on the wafer, such as logic circuits or memory cells containing dense transistor arrays, complex wiring, and multi-layer dielectric stacks. These regions have highly complex structures, including various materials, irregular geometries, etching morphologies, and localized stress distributions. Therefore, their polarization state parameters exhibit extremely complex spatial distribution patterns, and their response to minute process changes (such as etching depth, sidewall angles, thin film interface quality, and localized stress) is nonlinear and variable, making it difficult to directly determine their compliance using simple thresholds.

[0025] The process of establishing association models or empirical relationships: 1. Data accumulation and physical modeling: Experimental data accumulation: During the process development or early production phases, we select a large number of known qualified "gold wafers". These wafers undergo rigorous electrical testing and reliability verification to confirm that their device performance fully meets design requirements.

[0026] For each "gold wafer", we simultaneously use high-precision polarization measurement equipment (such as ellipsometers or polarization microscopes) to perform detailed measurements on its reference area and complex areas.

[0027] In the reference region, we can measure its average phase delay Δ_ref and fast axis azimuth Ψ_ref.

[0028] In complex regions, due to their complexity, we may need to obtain high-resolution spatial distribution maps of polarization state parameters, namely Δ_complex(x,y) and Ψ_complex(x,y). From these complex spatial maps, we can extract a series of parameters that can characterize the key properties of the device, such as the average Δ and Ψ of a specific sub-region, the spatial standard deviation of Δ and Ψ, the Δ or Ψ gradient at the edge of a specific structure, or the Fourier transform characteristics associated with periodic structures.

[0029] Physically based optical modeling: We can utilize tools such as thin-film optics theory and electromagnetic field simulations (e.g., finite-difference time-domain (FDTD) and rigorous coupled-wave analysis (RCWA)) to model the theoretical optical response of complex regions. For example, by inputting parameters such as different thin-film thicknesses, etching depths, and material compositions, we can predict the theoretical distribution of polarization state parameters. These physical models can help us understand the physical mechanisms of polarization response in complex regions and provide guidance and validation for the establishment of empirical models.

[0030] 2. Establish mapping relationships (association model): By pairing a large number of reference area measurements from "gold wafers" with corresponding complex region feature parameters, we can establish an accurate correlation model or empirical relationship. This model is not a simple equality relationship, but a complex mapping relationship.

[0031] Example mapping: Suppose that under certain process conditions, when the average phase delay Δ_ref of the reference region is 10 degrees and the fast axis azimuth Ψ_ref is 20 degrees (this may indicate the specific thickness and stress state of the thin film in the reference region).

[0032] Through the established correlation model, we found that for a qualified complex region (such as a specific memory cell array), the average phase delay Δ_gate_avg of the gate region of a key transistor inside it should be 35 degrees, and the phase delay gradient dΔ / dx at its gate edge should be between 0.5 degrees / nm and 0.7 degrees / nm. At the same time, the fast axis azimuth Ψ_array(x,y) of the entire array should exhibit a specific periodic symmetry pattern, and its spatial standard deviation should not exceed 3 degrees.

[0033] This model can describe how, when the reference region's Δ_ref and Ψ_ref undergo small changes (e.g., Δ_ref becomes 10.5 degrees due to process variations), the expected acceptable polarization state parameters (Δ_gate_avg, dΔ / dx, mode and standard deviation of Ψ_array(x,y)) in the complex region will also change predictively and non-linearly. For example, the model might predict that Δ_gate_avg will become 36 degrees, and dΔ / dx will become 0.6 degrees / nm.

[0034] Application of the model in actual production and qualification assessment: On a daily production line, after a new wafer completes specific process steps, we need to perform quality inspection on it: 1. Measurement of the reference region: First, measure the polarization state parameters of the reference region of the wafer to obtain Δ_ref' and Ψ_ref'.

[0035] 2. Model Prediction: Input Δ_ref' and Ψ_ref' into the pre-established correlation model. The model will predict the polarization state parameter characteristics that a qualified complex region on the wafer should exhibit based on the actual state of the current wafer reference area (e.g., the predicted mode and standard deviation of Δ_gate_avg', dΔ / dx', and Ψ_array(x,y)').

[0036] 3. Measuring complex regions: Next, the polarization state parameters of the actual complex regions on the wafer are measured, and the corresponding feature parameters are extracted (e.g., the mode and standard deviation of the actual measured Δ_gate_avg_actual, dΔ / dx_actual, and Ψ_array(x,y)_actual).

[0037] 4. Acceptance Judgment: Finally, the actual measured characteristic parameters of the complex region are compared with the acceptable characteristic parameters predicted by the model. If the two are highly consistent within the preset tolerance range, the complex region is deemed acceptable. Conversely, if there is a significant deviation, it indicates that the structure or material properties of the complex region are abnormal, and the wafer may be unacceptable or require further analysis.

[0038] In this way, even if the polarization state change patterns in complex regions are extremely complex, we can still indirectly, accurately, and efficiently judge the conformity of complex regions by using relatively simple measurements of the reference region and combining them with a precise correlation model, thereby greatly improving the quality control capabilities in the semiconductor manufacturing process.

[0039] In the aforementioned integrated circuit defect detection method using optical means, the incident light sequentially passes through a linear polarizer and a polarization rotation device to irradiate the detection area of ​​the wafer containing the integrated circuit. The steps of controlling the emission of the incident light and controlling the polarization state of the incident light include: Control the LED light source array to emit incident light, and make the incident light pass through a linear polarizer to generate linearly polarized light; Adjust the polarization rotation device to adjust the vibration direction of the linearly polarized light to a preset angle. The polarization rotation device is a rotatable half-wave plate or an electronically controlled solid-state polarization rotation device.

[0040] Specifically, before incident light illuminates the detection area of ​​the wafer, it is first guided through a linear polarizer. The function of this linear polarizer is to convert unpolarized or arbitrarily polarized light into linearly polarized light with a specific vibration direction. Subsequently, the linearly polarized light passes through a polarization rotation device. This polarization rotation device is configured to precisely change the vibration direction of the linearly polarized light to achieve the preset angle required for detection. The polarization rotation device can be understood as an optical element capable of changing the polarization direction of light; its specific implementation can be a rotatable half-wave plate or an electrically controlled solid-state polarization rotation device. A rotatable half-wave plate changes the polarization direction of the emitted light by mechanically rotating its crystal axis; while an electrically controlled solid-state polarization rotation device, such as a liquid crystal variable retarder or an electro-optic crystal, non-mechanically controls the rotation of the polarization direction by applying an electric field or voltage sequence, aiming to provide faster, more precise, and easily automated polarization direction adjustment capabilities.

[0041] The proposed solution converts the initial beam into linearly polarized light with a defined polarization direction by first passing the incident light through a linear polarizer, laying the foundation for subsequent precise polarization control. Subsequently, the linearly polarized light is guided through a polarization rotation device, which can precisely adjust the vibration direction of the linearly polarized light to the required angle according to preset detection requirements. It is precisely this two-stage polarization control mechanism that allows the polarization direction of the incident light to be systematically and repeatedly set, overcoming the potential inaccuracies or operational inconveniences of traditional single-polarization control methods. By precisely controlling the polarization direction of the incident light, the interaction between the light and wafer defects can be optimized, thereby improving the sensitivity and contrast of defect detection.

[0042] In some embodiments described above in this application, the polarization state of the incident light is controlled by a polarization rotation device. Specifically, the polarization rotation device can be an electrically controlled solid-state polarization rotation device.

[0043] Electro-controlled solid-state polarization rotation devices include liquid crystal variable retarders or electro-optic crystals. These devices control the rotation of the polarization direction by applying a series of voltages. Specifically, an electro-controlled solid-state polarization rotation device is a device that can adjust the polarization direction without mechanical rotation. Liquid crystal variable retarders utilize the property that the birefringence of liquid crystal materials changes under an applied electric field. By adjusting the applied voltage, the phase delay of light passing through is changed, thereby controlling the polarization direction. Electro-optic crystals utilize the electro-optic effect, that is, the phenomenon that the refractive index of a crystal changes under the influence of an electric field. By applying a voltage, the optical properties of the crystal are changed, thereby controlling the polarization direction of light. By applying a series of voltages, the polarization rotation angle of these devices can be adjusted precisely and rapidly.

[0044] This application's solution employs a liquid crystal variable retarder or an electro-optic crystal as an electrically controlled solid-state polarization rotation device, enabling precise, non-mechanical control of the polarization direction of incident light. When incident light emitted from an LED light source array passes through a linear polarizer to form linearly polarized light, this linearly polarized light enters the electrically controlled solid-state polarization rotation device. By applying a preset voltage sequence to this device, the optical properties of the liquid crystal variable retarder or electro-optic crystal can be altered as needed, thereby adjusting the vibration direction of the linearly polarized light to a preset angle. This electrical control method avoids the inertia problems and response speed limitations of traditional mechanical rotation devices, achieving rapid, precise, and flexible adjustment of the polarization direction.

[0045] Specifically, in the aforementioned integrated circuit defect detection method using optical means, the steps of collecting the reflected or scattered light rays from the detection area and calculating the polarization state parameters of the reflected light rays can be further refined.

[0046] The steps described above for collecting the returned light reflected or scattered from the detection area and calculating the polarization state parameters of the returned light include: The polarization analyzer acquires multiple images of the reflected or scattered light rays from the detection area to obtain multiple images of different polarization directions, including the intensity information of the returned light rays in different polarization directions; Based on multiple images with different polarization directions, the polarization state parameters of each pixel are calculated.

[0047] A polarization analyzer can be understood as an optical device used to analyze the polarization state of light. Specifically, it can consist of a rotatable polarizer and a photodetector, or a beam splitter and an array of detectors with fixed polarization directions. Its purpose is to decompose the returning light into components with different polarization directions and measure the intensity of these components. In practical applications, acquiring returning light using a polarization analyzer means obtaining images of the detection area one by one or simultaneously at multiple preset polarization angles.

[0048] Furthermore, calculating the polarization state parameters of each pixel based on multiple images with different polarization directions involves using the intensity information collected in these different polarization directions to determine the degree of polarization and polarization angle of each pixel through a specific mathematical model or algorithm. For example, the Stokes vector method can be used to calculate the Stokes vector by combining the light intensity values ​​in different polarization directions, and then extract the degree of polarization and polarization angle from the Stokes vector.

[0049] This application's solution, by introducing a polarization analyzer, can systematically and comprehensively acquire intensity information of reflected or scattered light from the detection area in different polarization directions. By acquiring multiple images with different polarization directions, richer polarization characteristics of the returned light can be captured, not just total light intensity information. Based on this detailed intensity information, the polarization state parameters of each pixel can be accurately calculated, thus providing a more accurate and comprehensive data foundation for subsequent defect judgment. This method can analyze the interaction between light and wafer surface or subsurface defects more precisely, because defects often cause small but detectable changes in the polarization state of reflected or scattered light.

[0050] Specifically, in some of the above embodiments, the step of calculating the polarization state parameter of each pixel based on multiple images with different polarization directions can be further refined as follows.

[0051] The steps for calculating the polarization state parameters of each pixel based on multiple images with different polarization directions include: Based on the light intensity value of each pixel in multiple images with different polarization directions, the Stokes vector is calculated. The Stokes vector includes the total light intensity, the difference between the horizontal and vertical polarization components, the difference between the 45-degree and 135-degree polarization components, and the intensity difference between the right-hand and left-hand circularly polarized light components. The degree of polarization and polarization angle of each pixel are calculated based on Stokes vectors.

[0052] Specifically, the Stokes vector is a mathematical tool used to describe the polarization state of light waves. It consists of four parameters: R0, R1, R2, and R3. R0 represents the total light intensity, i.e., the total energy or brightness of the light wave; R1 represents the intensity difference between horizontally and vertically polarized light, reflecting the dominance of horizontal or vertical polarization; R2 represents the intensity difference between 45-degree and 135-degree polarized light, reflecting the dominance of diagonal polarization; and R3 describes circular polarization, representing the intensity difference between right-handed and left-handed circularly polarized light. In practical applications, multiple images with different polarization directions acquired by a polarization analyzer can obtain light intensity information in different polarization directions. For example, images with polarization directions of 0 degrees, 90 degrees, 45 degrees, and 135 degrees can be acquired. Based on these intensity values, the Stokes vector for each pixel can be calculated using a standard formula.

[0053] Furthermore, once the Stokes vector for each pixel is obtained, the degree of polarization and the polarization angle can be derived from it. The degree of polarization (DOP) represents the proportion of polarized light in a light wave, with a value between 0 and 1, where 0 represents completely unpolarized light and 1 represents completely polarized light. The angle of polarization (AOP) represents the angle of vibration direction of linearly polarized light relative to a reference axis. The calculation of these parameters typically involves the components of the Stokes vector; for example, the degree of polarization can be calculated from R0, R1, and R2, and the angle of polarization can be calculated from R1 and R2.

[0054] The proposed method calculates the Stokes vector based on the light intensity value of each pixel in multiple images with different polarization directions, enabling a comprehensive and quantitative description of the polarization state of the returning light. As a standardized polarization description method, the Stokes vector integrates light intensity information from different polarization directions to form a complete polarization state description. Therefore, the degree of polarization and polarization angle can be accurately calculated using the Stokes vector, providing precise polarization state parameters for subsequent defect assessment. This method avoids the errors and uncertainties that may exist in directly measuring polarization parameters, ensuring the accuracy and reliability of the polarization state parameter calculation.

[0055] Traditional integrated circuit defect detection methods require comparing the polarization state parameters of the detection area with reference polarization state parameters of a pre-defined defect-free area when determining defects. However, in practice, accurately and reliably obtaining these reference polarization state parameters of the pre-defined defect-free area is crucial for ensuring detection accuracy and reducing false positives. Failure to address this issue may result in inaccurate reference parameters, thereby affecting the reliability of defect determination. To address this, this application proposes an optimized method for obtaining reference polarization state parameters by scanning a specific region at the wafer edge before wafer inspection begins.

[0056] In this regard, this application further proposes that the reference polarization state parameters of the aforementioned preset defect-free region are obtained through the following steps: Before wafer inspection begins, the test structure region or blank region at the edge of the wafer is scanned, and the average polarization degree and average polarization angle of the test structure region or blank region are calculated as reference polarization state parameters.

[0057] Specifically, "test structure regions or blank regions at the wafer edge" refer to structures specifically designed for process monitoring and testing, or blank areas not covered by circuit patterns, typically placed at the edge of the wafer during wafer manufacturing. These regions are generally considered defect-free, or their defect distribution is predictable, making them ideal for establishing a defect-free reference benchmark. By scanning these regions, polarization information representing the overall defect-free state of the wafer can be obtained. "Calculating the average degree of polarization and average polarization angle of the test structure region or blank region" refers to collecting the polarization state parameters, including the degree of polarization and polarization angle, of the returned light from each pixel within the selected test structure region or blank region after optical scanning. To obtain a stable and representative reference value, the degree of polarization and polarization angle of these pixels are averaged. Averaging helps eliminate local noise and minor fluctuations, resulting in more reliable reference polarization state parameters. The purpose is to provide a stable and accurate defect-free benchmark for subsequent defect assessment.

[0058] This application's solution effectively solves the problem of inaccurate or unstable reference parameter acquisition in traditional methods by using the test structure region or blank region at the wafer edge as a defect-free reference region before wafer inspection begins, and calculating its average polarization degree and average polarization angle as reference polarization state parameters. Specifically, the test structure region or blank region at the wafer edge is chosen because these regions are generally considered defect-free, or their characteristics are known and stable, providing a reliable defect-free benchmark. By scanning these regions and calculating the average polarization state parameters, errors caused by local process fluctuations or measurement noise can be minimized, thereby ensuring that the obtained reference polarization state parameters have high accuracy and representativeness. It is precisely because of this pre-established and systematic reference parameter acquisition mechanism that subsequent comparisons of the polarization state parameters of the inspection area with the reference polarization state parameters can more accurately identify polarization state changes caused by defects.

[0059] In some of the embodiments described above in this application, although it is proposed to compare the polarization state parameters with the reference polarization state parameters of a preset defect-free region to determine whether there are defects on the wafer with integrated circuits, in the implementation process, if there is a lack of clear quantitative standards and judgment logic, it may lead to insufficient accuracy and consistency in defect determination, and make it difficult to effectively distinguish between normal fluctuations and actual defects.

[0060] In response, this application further proposes a step of comparing polarization state parameters with reference polarization state parameters of a preset defect-free region to determine whether defects exist on a wafer containing integrated circuits, including: For each pixel in the detection area, calculate the difference index between its polarization state parameters and the reference polarization state parameters; If the difference index is greater than the preset tolerance threshold, it is determined that the corresponding region of the pixel on the wafer with integrated circuit has a potential defect and it is marked as a potential defect pixel; if the difference index is less than or equal to the preset tolerance threshold, it is determined that the corresponding region has no defect.

[0061] Specifically, the difference index can be understood as the quantitative difference between the polarization state parameters (including polarization degree and polarization angle) of each pixel in the detection area and the reference polarization state parameters of the preset defect-free area. For example, this difference index can be expressed by calculating the absolute value of the difference in polarization degree, the absolute value of the difference in polarization angle, or a comprehensive index combining both, such as Euclidean distance or weighted distance. Its purpose is to provide an objective numerical value to measure the degree of deviation in polarization characteristics between the detection area and the defect-free area. The preset tolerance threshold is a key boundary used to distinguish between normal fluctuations and potential defects. When the difference index exceeds this threshold, it indicates that the polarization characteristics of the pixel have significantly deviated from the normal defect-free state, and thus it is considered to have a potential defect. In practical applications, this threshold can be calibrated and set according to factors such as wafer material, process fluctuations, defect type, and detection sensitivity requirements. Therefore, pixels determined to have potential defects are considered potential defect pixels, and they will be of important concern in subsequent defect analysis and classification.

[0062] This application's solution concretizes the abstract "comparison" process into a quantified numerical judgment by introducing a difference index and a preset tolerance threshold. Specifically, the difference between the polarization state parameters of each pixel and the reference polarization state parameters is quantified as a difference index, which intuitively reflects the degree of abnormality in the polarization characteristics of that pixel. By comparing this difference index with the preset tolerance threshold, a clear judgment standard can be established. When the difference index exceeds the preset tolerance threshold, it indicates that the polarization characteristics of that region have changed significantly. This change is usually caused by subsurface defects, material anomalies, or structural inhomogeneities in integrated circuits. This quantitative judgment mechanism avoids the ambiguity of subjective judgment, making defect identification more objective and accurate.

[0063] In some embodiments described above in this application, the presence of a potential defect is determined by comparing the polarization state parameters of the detection area with the reference polarization state parameters of a preset defect-free area and based on whether the difference index exceeds a preset tolerance threshold. However, the setting of the preset tolerance threshold is crucial to the accuracy and reliability of defect detection. If the preset tolerance threshold is set improperly, such as being too high or too low, it may lead to missed detections or false alarms, thereby affecting the reliability of integrated circuit defect detection.

[0064] In this regard, this application further proposes the following steps for determining the aforementioned preset tolerance threshold: By performing polarized light detection on wafers with integrated circuits containing subsurface defects of different types and sizes, a large amount of actual defect sample data on the degree of polarization and polarization angle of defect regions was collected. Analyze actual defect sample data to determine the typical range of polarization degree and polarization angle changes caused by different defect types, and determine the preset tolerance threshold based on the typical range.

[0065] Specifically, "wafers with integrated circuits containing subsurface defects of different types and sizes" refers to test wafers that are artificially manufactured or naturally formed, and whose defect types (e.g., scratches, particles, voids, lattice defects, etc.) and sizes (e.g., nanometer-scale, micrometer-scale) are known. Polarization testing of these wafers aims to obtain polarization response data under conditions where real defects exist. "Collecting a large amount of actual defect sample data on the degree of polarization and polarization angle of polarization in defect areas" refers to accurately measuring and recording the polarization state parameters, including the degree of polarization and polarization angle, of reflected or scattered light for known defect areas during polarization testing of the aforementioned defective wafers. Collecting a large amount of sample data helps cover various defect conditions and improves data representativeness. In practical applications, "analyzing actual defect sample data to determine the typical range of polarization degree and polarization angle changes caused by different defect types" refers to performing statistical analysis on the collected massive amounts of polarization data, such as through data clustering, pattern recognition, or machine learning algorithms, to identify the unique ranges of polarization degree and polarization angle changes for different defect types. These typical ranges reflect the characteristics of the influence of specific defects on the polarization state of light. Therefore, "determining a preset tolerance threshold based on typical ranges" refers to setting one or a set of thresholds based on the typical ranges of polarization degree and polarization angle changes for different defect types obtained from analysis. This threshold aims to effectively distinguish between normal polarization state fluctuations in defect-free regions and significant polarization state changes caused by defective regions. For example, it can be set as a boundary value of the typical defect range, or a value that statistically differs significantly from the polarization state of defect-free regions.

[0066] This application's solution effectively addresses the issue of potentially inaccurate threshold settings in defect assessment by introducing a preset tolerance threshold determination process based on actual defect sample data. Specifically, by performing comprehensive polarization light detection on wafers with known defects, data on polarization state parameter changes caused by various defect types and sizes can be systematically collected. These actual sample data reflect the influence characteristics of defects on polarized light in the real world. Furthermore, through in-depth analysis of these actual defect sample data, the typical ranges of polarization degree and polarization angle changes caused by different defect types can be accurately identified. Based on these verified typical ranges, the determined preset tolerance threshold can more accurately capture polarization state anomalies caused by defects, thereby avoiding missed detections or false alarms caused by empirical or arbitrary threshold settings, significantly improving the reliability and accuracy of defect assessment.

[0067] In some embodiments described above, this application proposes determining the presence of potential defective pixels on a wafer by comparing polarization state parameters with reference polarization state parameters of a preset defect-free region and calculating a difference index. However, simply identifying potential defective pixels may not provide comprehensive defect information; for example, it may be unable to distinguish between different types of defects (such as scratches, particles, voids, etc.), which could lead to a lack of focus in subsequent defect analysis and process optimization. Without addressing these issues, it may be difficult to effectively guide defect tracing and yield improvement during the production process. Therefore, this application further proposes a method that, after identifying potential defective pixels, also includes clustering these pixels into candidate defect regions and extracting polarization state features and classifying defect types within these regions.

[0068] The method also includes: Adjacent potential defect pixels are clustered using image processing techniques to form at least one candidate defect region; For each candidate defect region, the polarization state features of the internal pixels are extracted based on the degree of polarization and polarization angle of the internal pixels. The extracted polarization state features are matched with a preset set of classification rules, and the candidate defect regions are classified according to the defect type based on the matching results.

[0069] Specifically, clustering adjacent potential defect pixels using image processing techniques involves utilizing various image processing algorithms, such as connected component analysis, morphological operations (e.g., dilation, erosion), or density-based clustering algorithms (e.g., DBSCAN), to combine spatially close pixels marked as potential defects into larger, physically meaningful regions. The aim is to integrate discrete potential defect pixels into identifiable defect entities, facilitating subsequent feature extraction and classification.

[0070] Specifically, for each candidate defect region, polarization state features of its internal pixels are extracted based on their polarization degree and polarization angle. This can be understood as comprehensively analyzing the polarization degree, polarization angle, and distribution characteristics of all potential defect pixels within each clustered candidate defect region to form a numerical or vector representation of the defect characteristics of that region. For example, average polarization degree, standard deviation of polarization degree, average polarization angle, polarization angle distribution range, and spatial gradient of polarization degree and polarization angle can be extracted as polarization state features. These features aim to capture the unique patterns of polarization response in different defect types.

[0071] In practical applications, the extracted polarization state features are matched with a pre-set classification rule set. Based on the matching results, candidate defect regions are classified into defect types. Specifically, the extracted polarization state features are input into a pre-trained classification model or rule set. This classification rule set can be constructed based on machine learning algorithms (such as Support Vector Machines (SVM), neural networks, decision trees, etc.) or based on threshold rules set by expert experience. By comparing the polarization state features of candidate defect regions with the feature patterns of known defect types in the rule set, the defect type of the candidate defect region is determined, such as scratches, particles, voids, subsurface defects, etc. The aim is to achieve automated and refined defect classification, providing precise guidance for subsequent defect analysis and process improvement.

[0072] This application's solution effectively overcomes the limitation of simply identifying potential defect pixels without providing defect type information by introducing clustering, feature extraction, and defect classification steps after identifying potential defect pixels. Specifically, firstly, image processing techniques are used to cluster adjacent potential defect pixels, integrating discrete pixels into candidate defect regions with actual physical meaning. This helps to aggregate minute, scattered polarization anomalies into identifiable defect entities, avoiding isolated judgments of individual pixels and improving the robustness of defect identification. Secondly, for each formed candidate defect region, polarization state features are extracted based on the polarization degree and polarization angle of its internal pixels. This fully utilizes the rich information provided by polarized light detection to capture subtle differences in polarization response among different defect types. Because different defect types have unique polarization characteristics in their reflection or scattering behavior of incident light, extracting these features can form a basis for distinguishing different defect types. Finally, matching these extracted polarization state features with a preset classification rule set enables automated defect type classification of candidate defect regions. This classification mechanism enables the system not only to determine "whether there is a defect" but also "what kind of defect it is," thereby providing more specific and valuable information for subsequent defect tracing, process optimization, and yield improvement.

[0073] In some of the embodiments described above in this application, an integrated circuit defect detection method using optical means is proposed. However, in its implementation, the lack of a structured and automated execution platform may lead to low detection efficiency, complex operation, and difficulty in ensuring consistent results. Especially in high-throughput production environments, manually or semi-automatically executing complex polarization control, data acquisition and analysis, and defect judgment steps will greatly limit its application value.

[0074] Secondly, see Figure 2This application proposes an integrated circuit defect detection system using optical means, for performing the aforementioned integrated circuit defect detection method using optical means. The system includes: The polarized light generation module 210 is used to control the emission of incident light and control the polarization state of the incident light so that the incident light irradiates the detection area of ​​the wafer with integrated circuits in a preset polarization direction. The polarization parameter calculation module 220 is used to collect the returned light reflected or scattered from the detection area and calculate the polarization state parameters of the returned light, including the degree of polarization and the polarization angle. The defect judgment module 230 is used to compare the polarization state parameters with the reference polarization state parameters of the preset defect-free region to determine whether there are defects on the wafer with integrated circuits.

[0075] 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 method for detecting defects in integrated circuits using optical means, characterized in that, include: The incident light is controlled to be emitted, and the polarization state of the incident light is controlled so that the incident light irradiates the detection area of ​​the wafer with integrated circuits in a preset polarization direction. Collect the reflected or scattered light rays from the detection area and calculate the polarization state parameters of the reflected light rays, including the degree of polarization and the polarization angle; The polarization state parameters are compared with the reference polarization state parameters of a preset defect-free region to determine whether there are defects on the wafer with integrated circuits. The step of comparing the polarization state parameter with the reference polarization state parameter of a preset defect-free region to determine whether there are defects on the wafer with the integrated circuit includes: For each pixel in the detection area, calculate the difference index between its polarization state parameter and the reference polarization state parameter; If the difference index is greater than a preset tolerance threshold, it is determined that the pixel has a potential defect in the corresponding area on the wafer with integrated circuits, and it is marked as a potential defect pixel. If the difference index is less than or equal to the preset tolerance threshold, then the corresponding area is determined to be defect-free. The preset tolerance threshold is determined through the following steps: By performing polarized light detection on wafers with integrated circuits containing subsurface defects of different types and sizes, a large amount of actual defect sample data on the degree of polarization and polarization angle of defect regions was collected. Analyze the actual defect sample data to determine the typical range of polarization degree and polarization angle changes caused by different defect types, and determine the preset tolerance threshold based on the typical range.

2. The integrated circuit defect detection method using optical means according to claim 1, characterized in that, The incident light sequentially passes through a linear polarizer and a polarization rotation device to irradiate the detection area of ​​the wafer with the integrated circuit. The steps of controlling the emission of the incident light and controlling the polarization state of the incident light include: Control the LED light source array to emit incident light, and make the incident light pass through the linear polarizer to generate linearly polarized light; The polarization rotation device is adjusted to adjust the vibration direction of the linearly polarized light to a preset angle, wherein the polarization rotation device is a rotatable half-wave plate or an electrically controlled solid-state polarization rotation device.

3. The integrated circuit defect detection method using optical means according to claim 2, characterized in that, The electrically controlled solid-state polarization rotation device includes a liquid crystal variable delay unit or an electro-optic crystal, and the electrically controlled solid-state polarization rotation device controls the rotation of the polarization direction by applying a voltage sequence.

4. The integrated circuit defect detection method using optical means according to claim 1, characterized in that, The step of collecting the reflected or scattered light rays from the detection area and calculating the polarization state parameters of the reflected light rays includes: A polarization analyzer is used to collect the reflected or scattered light rays from the detection area to obtain multiple images with different polarization directions. The multiple images with different polarization directions include the intensity information of the reflected light rays in different polarization directions. Based on the multiple images with different polarization directions, the polarization state parameters of each pixel are calculated.

5. The integrated circuit defect detection method using optical means according to claim 4, characterized in that, The step of calculating the polarization state parameter of each pixel based on the multiple images with different polarization directions includes: Based on the light intensity value of each pixel in the multiple images with different polarization directions, the Stokes vector is calculated respectively. The Stokes vector includes the total light intensity, the difference between the horizontal and vertical polarization components, and the difference between the 45-degree and 135-degree polarization components. The polarization degree and polarization angle of each pixel are calculated based on the Stokes vector.

6. The integrated circuit defect detection method using optical means according to claim 1, characterized in that, The reference polarization state parameters of the preset defect-free region are obtained through the following steps: Before wafer inspection begins, a preset defect-free reference area on the wafer is scanned, and the average polarization degree and average polarization angle of the defect-free reference area are calculated as the reference polarization state parameters. The defect-free reference area is the test structure area, blank area, or other designated area with comparable optical response to the inspection area at the edge of the wafer.

7. The integrated circuit defect detection method using optical means according to claim 1, characterized in that, The method further includes: The adjacent potential defect pixels are clustered using image processing techniques to form at least one candidate defect region; For each candidate defect region, the polarization state features of the internal pixels are extracted based on the polarization degree and polarization angle of the internal pixels. The extracted polarization state features are matched with a preset set of classification rules, and the candidate defect regions are classified according to the defect type based on the matching results.

8. An integrated circuit defect detection system using optical means, for performing the integrated circuit defect detection method using optical means as described in any one of claims 1 to 7, characterized in that, The system includes: A polarized light generation module is used to control the emission of incident light and control the polarization state of the incident light so that the incident light irradiates the detection area of ​​the wafer with integrated circuits in a preset polarization direction. A polarization parameter calculation module is used to collect the returned light reflected or scattered from the detection area and calculate the polarization state parameters of the returned light, including the degree of polarization and the polarization angle. The defect determination module is used to compare the polarization state parameters with the reference polarization state parameters of a preset defect-free region to determine whether there are defects on the wafer with integrated circuits.

Images