A defect detection method and system based on dual-wavelength coded illumination

Abstract

This application relates to a defect detection method and system based on dual-wavelength coded illumination. The defect detection method includes dividing an emitting end into multiple emitting modules along the circumference; adjusting the illumination mode of each emitting module; switching the illumination mode of each emitting module so that the illumination beam of each emitting module has different incident angles and azimuth angles relative to the test point on the wafer surface; acquiring a first light intensity and a second light intensity; acquiring compensation data G; acquiring a dual-wavelength dark-field image and correcting the dual-wavelength dark-field image according to the compensation data G; and determining the longitudinal depth of the defect based on the dual-wavelength dark-field image. The defect detection system includes a wafer stage, an emitting end including multiple emitting modules circumferentially surrounding the wafer stage; an imaging module; and a control system connected to the emitting end and the imaging module for setting and switching the illumination mode of each emitting module. This invention can utilize different wavelengths for defect detection and the illumination beam is adjustable.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to a defect detection method and system based on dual-wavelength coded illumination. Background Technology

[0002] In the semiconductor integrated circuit manufacturing process, accurate detection of wafer surface and subsurface defects (such as scratches, particles, thin film interlayer inclusions, and metal interlayer short-circuit bumps) is crucial for process yield control.

[0003] In existing technologies, most dark-field defect detection systems employ a fixed ring illumination layout. The incident angle and azimuth angle of the illumination beam are not adjustable after system assembly, making it difficult to dynamically match the optimal illumination configuration based on the differences in scattering characteristics of defects at different depths. Furthermore, existing systems typically acquire only a single wavelength or use a time-division multiplexing imaging method. The former cannot utilize the differences in penetration depth and scattering cross-section of different wavelengths to distinguish defect depths, while the latter introduces timing errors such as mechanical vibration and thermal drift due to time-division acquisition, resulting in distortion of dual-wavelength grayscale comparison.

[0004] Therefore, it is necessary to provide a new defect detection method and system based on dual-wavelength coded illumination to solve the above-mentioned problems in the prior art. Summary of the Invention

[0005] The technical problem to be solved by this application is how to provide a defect detection method and system based on dual-wavelength coded illumination that can perform defect detection using different wavelengths and has an adjustable illumination beam.

[0006] To address the aforementioned technical problems, according to embodiments of this application, a defect detection method based on dual-wavelength coded illumination is provided, comprising the following steps: dividing the transmitting end circumferentially into multiple transmitting modules, each transmitting module including an optical fiber sub-bundle and a collimating lens; the optical fiber sub-bundle is used to transmit illumination beams of a first wavelength and a second wavelength; adjusting the illumination mode of each transmitting module; the illumination mode includes a switch state, radial displacement, and circumferential misalignment angle; sequentially switching the illumination mode of each transmitting module according to a preset sequence, so that the illumination beams of each transmitting module have different incident angles and azimuth angles relative to the test point on the wafer surface; each After switching the illumination mode, the first light intensity at each collimating lens and the second light intensity at the fiber sub-bundle are collected; the center position (xc, yc) and beam deviation angle R of the light intensity distribution of the corresponding transmitting module are obtained based on the first light intensity; the end face normal vector deflection angle n of each transmitting module is obtained based on the second light intensity; compensation data G is obtained based on the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n; a dual-wavelength dark field image is collected, and the dual-wavelength dark field image is corrected based on the compensation data G; the longitudinal depth of the defect is determined based on the dual-wavelength dark field image.

[0007] According to an embodiment of this application, adjusting the illumination mode of each of the emitting modules includes: controlling the corresponding emitting module to move radially along the wafer, changing the radial distance of the emitting module relative to the geometric center of the emitting end to adjust the radial displacement; controlling the corresponding emitting module to deflect in the tangential direction of the circumference to adjust the circumferential misalignment angle; and controlling the on / off state of the corresponding emitting module to adjust the switching state.

[0008] According to an embodiment of this application, obtaining the center position (xc, yc) of the light intensity distribution of the corresponding emitting module and the beam deviation angle R based on the first light intensity includes: acquiring the first light intensity through a first sampling module, the first sampling module including a photodiode array; converting the first light intensity into a two-dimensional light intensity distribution through the photodiode array, and performing a weighted average of the two-dimensional light intensity distribution with pixel coordinates as weights to obtain the center position (xc, yc); determining the offset distance of the center position (xc, yc) relative to the geometric center (x0, y0) of the photodiode array; and obtaining the beam deviation angle R based on the offset distance and the focal length of the sampling lens.

[0009] According to an embodiment of this application, obtaining the end-face normal vector deflection angle n of each of the transmitting modules based on the second light intensity includes: calculating the ratio of the first wavelength light intensity to the second wavelength light intensity to obtain a dual-wavelength light intensity ratio; wherein the bending loss of the first wavelength is greater than the bending loss of the second wavelength, and when the fiber sub-bundle bends, the light intensity attenuation of the first wavelength is greater than the light intensity attenuation of the second wavelength; obtaining the fiber bending radius of each of the fiber sub-bundles based on the dual-wavelength light intensity ratio and a preset bending loss calibration curve; and obtaining the end-face normal vector deflection angle n of each of the transmitting modules based on the fiber bending radius.

[0010] According to an embodiment of this application, the step of obtaining compensation data G based on the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n includes: setting a deviation threshold; obtaining the beam offset of the illumination beam based on the center position (xc, yc) and the beam deviation angle R; calculating the difference between the beam offset and the end face normal vector deflection angle n and taking the absolute value to obtain the deviation amount; if the deviation amount is less than the deviation threshold, then fusing the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n to form compensation data G; if the deviation amount is greater than or equal to the deviation threshold, then re-acquiring the first light intensity and the second light intensity for calculation.

[0011] According to an embodiment of this application, the step of acquiring a dual-wavelength dark-field image and correcting the dual-wavelength dark-field image according to the compensation data G includes: determining the required reverse displacement amount and grayscale gain factor for each pixel according to the compensation data G; performing reverse displacement on the pixels of the dual-wavelength dark-field image according to the reverse displacement amount; and multiplying the pixels of the dual-wavelength dark-field image according to the grayscale gain factor to correct the dual-wavelength dark-field image.

[0012] According to an embodiment of this application, determining the longitudinal depth of the defect based on the dual-wavelength dark-field image includes: taking the gray-scale anomaly region in the corrected dual-wavelength dark-field image as the defect region; obtaining the first gray-scale value of the pixel in the defect region at the first wavelength and the second gray-scale value at the second wavelength; dividing multiple sampling points along the wafer axis; establishing the theoretical dual-wavelength gray-scale value h of each sampling point in each illumination mode based on the incident angle, azimuth angle, and the difference in scattering intensity of the defect to different wavelengths under each illumination mode; and comparing the first gray-scale value and the second gray-scale value with the theoretical dual-wavelength gray-scale value h to determine the depth sampling point where the defect is located.

[0013] According to an embodiment of this application, the defect detection method further includes: for each located defect region, dividing the first gray value by the second gray value to obtain a dual-wavelength ratio under each illumination mode; calculating the arithmetic mean of the dual-wavelength ratio under each illumination mode to obtain a dual-wavelength mean; performing linear fitting with the incident angle corresponding to each illumination mode as the abscissa and the dual-wavelength ratio as the ordinate to obtain a dual-wavelength slope; if the longitudinal depth of the defect is zero, it is determined to be a surface defect; if the longitudinal depth of the defect is greater than zero and both the mean and the slope are less than the corresponding set value, it is determined to be a subsurface defect; if the longitudinal depth of the defect is greater than zero and both the mean and the slope are greater than or equal to the corresponding set value, it is determined to be a subsurface strong reflection defect.

[0014] A defect detection system based on dual-wavelength coded illumination is provided for use in the aforementioned defect detection method. The defect detection system includes: a wafer stage for supporting a wafer under test; an emitting end including multiple emitting modules circumferentially surrounding the wafer stage, the emitting modules being used to emit illumination beams of a first wavelength and a second wavelength onto the wafer; an imaging module including an objective lens, a telescope lens, and a camera arranged sequentially along the optical axis to acquire dual-wavelength dark-field images; and a control system connected to the emitting end and the imaging module for setting and switching the illumination modes of each emitting module.

[0015] According to an embodiment of this application, the transmitting module includes an optical fiber sub-bundle, a collimating lens, a first sampling module, and a second sampling module; the optical fiber sub-bundle is used to transmit illumination beams of a first wavelength and a second wavelength, and the collimating lens is disposed at the output end of the optical fiber sub-bundle; the first sampling module is disposed on the side of the collimating lens away from the optical fiber sub-bundle to collect a first light intensity at the collimating lens; the second sampling module is disposed on the side of the collimating lens facing the optical fiber sub-bundle to collect a second light intensity at the optical fiber sub-bundle.

[0016] By adopting the above technical solution, the transmitter is divided into multiple transmitter modules along the circumference, and the switching state, radial displacement, and circumferential misalignment angle of each module are adjusted respectively. This allows the illumination beam to have different incident angles and azimuth angles relative to the test point on the wafer surface, thus transforming the illumination mode switching process into an active encoding and sampling process for defect depth information, breaking through the angular domain limitations of traditional fixed ring illumination. A multispectral camera is used to simultaneously acquire dual-wavelength dark-field images, avoiding errors caused by mechanical vibrations due to time-division switching of wavelengths, ensuring that the grayscale values ​​of the two wavelengths correspond at the same time. By setting the first sampling module and the second sampling module at the collimating lens and fiber sub-bundle respectively, direct and indirect detection are fused to form compensation data, and geometric offset correction and grayscale gain correction are performed on the dual-wavelength dark-field images, thereby eliminating errors introduced by the attitude offset of the transmitter modules. Furthermore, by utilizing the differences in dual-wavelength scattering characteristics under different illumination modes and comparing them with the theoretical dual-wavelength grayscale values ​​of each sampling point, the longitudinal depth of the defect is determined when the number of illumination modes is less than the number of depth sampling points. This extends dark-field detection from two-dimensional grayscale anomaly judgment to three-dimensional depth positioning, improving detection efficiency. Attached Figure Description

[0017] Figure 1 This is an optical path diagram of a defect detection system according to an embodiment of the present invention.

[0018] Figure 2 This is a top view showing the positional relationship between the transmitting module and the imaging module of a defect detection system according to an embodiment of the present invention.

[0019] Figure 3 This is a flowchart illustrating the steps of a defect detection method according to an embodiment of the present invention.

[0020] Figure label: 100. Wafer stage; 200. Transmitter; 210. Transmitter module; 211. Fiber sub-bundle; 212. Collimating lens; 213. First sampling module; 214. Second sampling module; 300. Imaging module; 310. Objective lens; 320. Tube lens; 330. Acquisition camera. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0022] The following is in conjunction with the appendix Figures 1-3 The specific embodiments of the present invention will be further described in detail below.

[0023] The embodiments of the present invention provide a defect detection method and a defect detection system based on dual-wavelength coded illumination. The defect detection system is used to implement the defect detection method. The defect detection system includes a wafer stage 100, an emitter 200, an imaging module 300, and a control system.

[0024] In some embodiments, the wafer stage 100 is used to support the wafer under test; specifically, the wafer stage 100 is located at the bottom layer of the entire optical system and includes a vacuum adsorption mechanism for stably adsorbing the wafer under test onto the stage surface, preventing wafer displacement due to airflow or mechanical vibration during the testing process. This is well known to those skilled in the art and will not be described in detail here. The transmitter 200 includes multiple transmitter modules 210 circumferentially surrounding the wafer stage 100. The transmitter modules 210 are used to emit illumination beams of a first wavelength and a second wavelength to the wafer; specifically, the transmitter 200 consists of multiple transmitter modules. Blocks 210 are spliced ​​together in a ring along the circumference, and there can be 8 to 16 blocks, preferably 12. The emitting ends of each emitting module 210 are all facing the wafer stage 100 so that the light can illuminate the wafer surface. At the same time, a mechanical gap of 0.5 mm to 1 mm is reserved between adjacent emitting modules 210 to avoid interference during the movement of each emitting module 210. It is worth noting that since each emitting module 210 is movable, it can be adjusted into a standard ring, an elliptical ring or other shapes according to the detection requirements, thereby improving the flexibility of the illumination azimuth angle and incident angle distribution.

[0025] In some embodiments, the imaging module 300 includes an objective lens 310, a tube lens 320, and an acquisition camera 330 arranged sequentially along the optical axis to acquire dual-wavelength dark-field images. The imaging module 300 is located above the wafer stage 100, with its optical axis parallel to the wafer's axial direction, and is used to receive dark-field light signals scattered from the wafer surface and subsurface defects. The emitting module 210 is arranged around the imaging module 300; light reflected from the wafer passes through the objective lens 310 and enters the tube lens, ultimately focusing onto the acquisition camera 330. The tube lens 320 focuses the parallel light emitted from the objective lens 310 onto the acquisition camera 330. The acquisition camera 330 is a multispectral camera to simultaneously acquire dark-field images of the first and second wavelengths, thereby avoiding errors caused by time-division switching of wavelengths. The control system is connected to the transmitter 200 and the imaging module 300, and is used to set and switch the illumination mode of each transmitter module 210. The control system includes a radial driver and a circumferential driver, and each transmitter module 210 corresponds to one radial driver and one circumferential driver. The radial driver is used to drive the transmitter module 210 to move linearly along the wafer radius, and the circumferential driver is used to control the deflection of the illumination angle of the transmitter module 210. The radial driver can be composed of a guide rail and a stepper motor. The circumferential driver can be multiple electric cylinders (for example, connected to the board through a universal joint, with the transmitter module 210 located at the bottom of the board, and multiple micro electric cylinders arranged circumferentially around the universal joint on the top of the board, the multiple micro electric cylinders working together to adjust the deflection angle of the board and the transmitter module 210, while the radial driver can drive the transmitter module 210, the board, the universal joint and other components to move synchronously). The selection and installation methods of the radial driver and the circumferential driver are known to those skilled in the art, and will not be described in detail here.

[0026] In some embodiments, the transmitting module 210 includes an optical fiber sub-bundle 211, a collimating lens 212, a first sampling module 213, and a second sampling module 214; in addition, each transmitting module 210 has a mounting frame for mounting the above components; the optical fiber sub-bundle 211 is used to transmit illumination beams of a first wavelength and a second wavelength, and the collimating lens 212 is disposed at the output end of the optical fiber sub-bundle 211; specifically, the input end of the optical fiber sub-bundle 211 is connected to a light source, and a dichroic beam splitter is provided between the two, the light source including a light source for emitting illumination light of the first wavelength. The system includes a first wavelength light source (e.g., 450nm blue light) and a second wavelength light source (e.g., 850nm near-infrared light) for emitting a second wavelength illumination light. The first wavelength light source is located on the transmission surface side of the dichroic beam splitter, and the second wavelength light source is located on the reflection surface side of the dichroic beam splitter. The dichroic beam splitter combines the first wavelength illumination light with the second wavelength illumination light. The combined illumination light enters the main fiber bundle, and the main fiber bundle is branched into multiple branches by fiber couplers to enter different transmitting modules 210 to form fiber sub-bundles 211. A collimating lens 212 is fixed to the mounting frame and placed at the output end of the fiber sub-bundle 211, thereby converting the light into collimated light. The first sampling module 213 is located on the side of the collimating lens 212 away from the fiber sub-bundle 211 to collect the first light intensity at the collimating lens 212. Specifically, the first sampling module 213 can be a photodiode array or other sampling device, which is fixed to the mounting frame and spaced from the end face of the collimating lens 212. The first sampling module 213 is offset from the principal optical axis of the collimating lens 212 by 5° to 10° to avoid blocking the illumination beam. It is worth noting that its position is close to the collimating lens 212, for example, 3mm away from the collimating lens 212, so as not to affect the light transmission and to detect the light intensity. Its detection signal is amplified by a signal amplifier and output after analog-to-digital conversion, for example, output to a central control computer, so as to reflect the light intensity. The specific details are not elaborated here. The second sampling module 214 is located on the side facing the fiber sub-bundle 211. To facilitate detection, a light leakage port is provided on the cladding of the fiber sub-bundle 211, which is well known to those skilled in the art and will not be described in detail here. The second sampling module 214 can be a photodiode or other sampling device, which is fixed to the mounting frame and placed on the side of the collimating lens 212 close to the fiber sub-bundle 211 to avoid blocking the illumination beam. Its detection signal is amplified by a signal amplifier and output after analog-to-digital conversion, for example, to a central control computer, so as to reflect the light intensity. The specific details will not be described in detail here.

[0027] The defect detection method provided in this application includes the following steps: S1. The transmitter 200 is divided into multiple transmitter modules 210 along the circumference. Each transmitter module 210 includes an optical fiber sub-bundle 211 and a collimating lens 212. The optical fiber sub-bundle 211 is used to transmit illumination beams of the first wavelength and the second wavelength. Specifically, through modular design, the transmitter 200 is transformed from a static state to a dynamic state, so that the illumination light field can switch between states such as standard circular ring, open ring, and elliptical ring according to the needs of defect detection, thereby improving the flexibility of the distribution of illumination azimuth and incident angle of the illumination light field.

[0028] S2. Adjust the illumination mode of each emitting module 210; the illumination mode includes the on / off state, radial displacement, and circumferential misalignment angle; specifically, the on / off state of each emitting module 210 is set to control the on / off state of the emitting module 210, the radial displacement is set to adjust the incident angle of the illumination beam relative to the test point on the wafer surface, and the circumferential misalignment angle is set to adjust the azimuth angle of the illumination beam relative to the test point on the wafer surface; thereby, by adjusting the illumination mode of each illumination module separately, multiple different illumination configurations can be formed, so that the incident angle and azimuth angle of each illumination mode meet the detection requirements, avoiding the problem of image blurring caused by the inability to adjust the light source angle.

[0029] S3. The illumination modes of each emission module 210 are switched sequentially according to a preset sequence, so that the illumination beams of each emission module 210 have different incident angles and azimuth angles relative to the test point on the wafer surface. By actively switching the illumination modes, different incident angles and azimuth angles are formed, thereby transforming the process of switching illumination modes into an active encoding and sampling process of defect depth information. The difference in scattering characteristics between the first wavelength and the second wavelength under different illumination modes is used to determine the defect depth.

[0030] S4. After each switching of illumination mode, the first light intensity at each collimating lens 212 and the second light intensity at the fiber sub-bundle 211 are collected; the center position (xc, yc) of the light intensity distribution of the corresponding transmitting module 210 and the beam deviation angle R are obtained based on the first light intensity; the deflection angle n of the end face normal vector of each transmitting module 210 is obtained based on the second light intensity; specifically, after each switching of illumination mode, the light intensity distribution at the collimating lens 212 of each transmitting module 210 is collected by the first sampling module 213 as the first light intensity, and the light intensity distribution between the fiber sub-bundle 211 and the collimating lens 212 of each transmitting module 210 is collected by the second sampling module 214 as the second light intensity; wherein... The first light intensity reflects the light intensity distribution at the emission surface of the transmitting module 210, and the second light intensity reflects the wavelength power attenuation state of the fiber sub-bundle 211 within the transmitting module 210. Furthermore, the first sampling module 213 is a photodiode array, and the second sampling module 214 consists of two photodiodes respectively set for different wavelengths. Therefore, the first light intensity reflects the offset of the illumination beam, which is a direct detection, while the second light intensity reflects the power of the fiber sub-bundle 211. Then, the offset of the collimating lens 212 is inverted based on the wavelength loss, which is an indirect detection. The combination of direct and indirect detection improves the accuracy of detection and avoids the situation where detection is impossible due to the failure of a single channel.

[0031] S5. Obtain compensation data G based on the center position (xc, yc), beam deviation angle R, and end face normal vector deflection angle n.

[0032] S6. Acquire dual-wavelength dark-field images and correct them according to the compensation data G. Specifically, after acquiring the compensation data G each time, correct the corresponding dual-wavelength dark-field image, that is, perform grayscale correction and geometric correction on the dual-wavelength dark-field image according to the compensation data G, so as to eliminate the error introduced by the light field distortion.

[0033] S7. Determine the longitudinal depth of the defect based on the dual-wavelength dark field image; specifically, based on the corrected dual-wavelength dark field image, the longitudinal depth of the defect is determined by utilizing the difference in scattering characteristics of the two wavelengths under different illumination modes, thereby transforming dark field detection from two-dimensional grayscale anomaly-based defect area determination to three-dimensional depth localization of the defect.

[0034] In some embodiments, adjusting the illumination mode of each emitting module 210 includes controlling the corresponding emitting module 210 to move radially along the wafer, changing the radial distance of the emitting module 210 relative to the geometric center of the emitting end 200 to adjust the radial displacement; by adjusting the radial displacement, each emitting module 210 can adjust its working distance within a certain range, thereby adjusting the incident angle to meet the illumination angle requirements of defects of different depths; controlling the corresponding emitting module 210 to deflect in the tangential direction of the circumference to adjust the circumferential misalignment angle; realizing illumination modes with different azimuth angles such as non-uniform distribution and spiral, solving the azimuth angle redundancy of traditional ring illumination; controlling the on / off state of the corresponding emitting module 210 to adjust the switching state, thereby realizing rapid switching of the switching state of the emitting module 210, and facilitating the adjustment of the illumination mode for forming an open, incomplete ring state.

[0035] In some embodiments, obtaining the center position (xc, yc) and beam deviation angle R of the light intensity distribution of the corresponding emitting module 210 based on the first light intensity includes: acquiring the first light intensity through a first sampling module 213, which includes a photodiode array and a sampling lens disposed between the photodiode array and the collimating lens 212; after passing through the sampling lens, the light illuminates the photodiode array; the photodiode array converts the first light intensity into a two-dimensional light intensity distribution; and a weighted average of the two-dimensional light intensity distribution is performed with pixel coordinates as weights to obtain the center position (xc, yc); specifically, the first light intensity is set as a two-dimensional light intensity distribution I(x, y), and a weighted average is performed with pixel coordinates as weights, which is known to those skilled in the art. The specifics are not detailed here. The center position (xc, yc) of the light intensity distribution is obtained. It is worth noting that in the initial system calibration state (i.e., the emitting module 210 has not shifted), the pixel coordinates of the principal optical axis on the photodiode array when each emitting module 210 has not shifted are recorded as the geometric center (x0, y0). The offset distance L is the Euclidean distance between the center position (xc, yc) and the geometric center (x0, y0). Here, the two-dimensional light intensity distribution I(x, y) is the light intensity value collected by the photodiode array at pixel coordinates (x, y). If the collimating lens 212 tilts, the center position (xc, yc) will shift along the tilt direction, thereby quantifying the lateral drift of the light field formed by the corresponding emitting module 210. The offset distance of the center position (xc, yc) relative to the geometric center (x0, y0) of the photodiode array is determined. Specifically, the offset distance L between the center position (xc, yc) and the geometric center (x0, y0) of the photodiode array is calculated. When the transmitting module 210 is not tilted, the center of the light spot coincides with the geometric center (x0, y0) of the array, and the offset distance L is 0. When the transmitting module 210 is tilted, the center of the light spot deviates from the geometric center (x0, y0), and the offset distance L is greater than 0. The offset distance L reflects the offset of the imaging light spot caused by the tilt of the collimating lens 212 of the transmitting module 210. The beam deviation angle R is obtained based on the offset distance and the focal length of the sampling lens, R = arctan(L*O / f), where f is the focal length of the sampling lens, the beam deviation angle R converts the lateral offset of the light spot formed by the illumination beam into the angular offset of the illumination beam, and O is the pixel size of the photodiode array.

[0036] In some embodiments, the deflection angle n of the end face normal vector of each transmitting module 210 is obtained based on the second light intensity, including calculating the ratio of the first wavelength light intensity to the second wavelength light intensity to obtain the dual-wavelength light intensity ratio; wherein, the bending loss of the first wavelength is greater than the bending loss of the second wavelength, and when the fiber sub-bundle 211 bends, the light intensity attenuation of the first wavelength is greater than the light intensity attenuation of the second wavelength; specifically, let the light intensity of the first wavelength be I1 and the light intensity of the second wavelength be I2, then the dual-wavelength light intensity ratio is I1 / I2; since the first wavelength is shorter than the second wavelength, the bending loss coefficient a1 of the first wavelength is greater than the bending loss coefficient a2 of the second wavelength. When the bending radius r of the fiber sub-bundle 211 decreases due to deformation, the attenuation of the first wavelength light intensity I1 is greater than that of the second wavelength light intensity I2. Therefore, the dual-wavelength light intensity ratio changes with the degree of bending, that is, the deformation of the fiber can be known without additional strain sensors. Based on the dual-wavelength intensity ratio and a preset bending loss calibration curve, the fiber bending radius of each fiber sub-bundle 211 is obtained. The bending loss calibration curve is obtained by winding the fiber sub-bundle 211 around a mold with a known bending radius. The radius of the mold covers all bending states that the transmitting module 210 may experience during actual adjustment, from a straight state to the minimum allowable bending radius. At each radius, the intensity of the first wavelength I1 and the intensity of the second wavelength I2 at the end of the fiber sub-bundle 211 are simultaneously acquired using a dual-channel photodiode, and their dual-wavelength intensity ratio is calculated. The dual-wavelength intensity ratios and their corresponding standard radii are recorded and fitted to form the bending loss calibration curve. During the detection process, the fiber bending radius of the fiber sub-bundle 211 is obtained based on the calculated dual-wavelength light intensity ratio, thus transforming the unmeasurable internal deformation of the fiber sub-bundle 211 into a calculable fiber bending radius. The deflection angle 'n' of the end face normal vector of each transmitting module 210 is obtained based on the fiber bending radius. Since the collimating lens 212 is located at the output end of the fiber sub-bundle 211, the bending of the fiber sub-bundle 211 directly causes a change in the spatial orientation of its output end. A fixed spatial angular relationship exists between the axial direction (tangential direction) of the fiber sub-bundle 211 and the end face of the collimating lens 212. For example... For example, let d be the arc length from the center of the output end face of the fiber sub-bundle 211 after bending to the position before bending, and let n satisfy n=d / r. Therefore, the collimating lens 212 and the fiber sub-bundle 211 will deflect synchronously. Based on this spatial angle relationship, the normal vector offset of the collimating lens 212 of each transmitting module 210 can be obtained, that is, the end face normal vector deflection angle n. The end face normal vector deflection angle n reflects the amount of deflection of the normal of the end face of the collimating lens 212 relative to the normal of the end face of the collimating lens 212 when the fiber sub-bundle 211 is not bent. It is an indirect parameter for quantifying the angle drift of the illumination beam.

[0037] In some embodiments, compensation data G is obtained based on the center position (xc, yc), beam deviation angle R, and end face normal vector deflection angle n. This includes setting a deviation threshold, with a deviation threshold of 5% selected; obtaining the beam offset of the illumination beam based on the center position (xc, yc) and beam deviation angle R. Here, the center position (xc, yc) reflects the lateral deviation of the illumination beam, and the beam deviation angle R reflects the angular deviation of the illumination beam. The two together constitute the beam offset of the illumination beam. The beam offset is used to characterize the drift state of the illumination light field of the emission module 210, and the lateral deviation and angular deviation are fused into a single offset, which is convenient for comparison with the end face normal vector deflection angle n in the same dimension. The difference between the beam offset and the end-face normal vector deflection angle n is calculated, and the absolute value is taken. This absolute value is then multiplied by 100% to obtain the deviation. It is worth noting that in the difference calculation, the beam deviation angle R in the beam offset is used to calculate the absolute value of the difference with the end-face normal vector deflection angle n. If the deviation is less than the deviation threshold, the center position (xc, yc), beam deviation angle R, and end-face normal vector deflection angle n are fused to form compensation data G. Specifically, the beam deviation angle R and end-face normal vector deflection angle n are weighted and fused to form an angle correction component. , ,in, and These are weighted coefficients, and their sum is 1. For example, we can take... The value is 0.3-0.7, preferably 0.5. The value is 0.3-0.7, preferably 0.5; based on the center position (xc, yc) and the geometric center of the photodiode array (x0, y0), the lateral correction component M of the illumination beam is obtained, M=(xc-x0, yc-y0), and the compensation data G is obtained from the angle correction component. Together with the lateral correction component M, it constitutes the subsequent geometric offset correction and grayscale gain correction for the dual-wavelength dark field image. If the deviation is greater than or equal to the deviation threshold, the first and second light intensities are re-acquired for calculation. Specifically, if the deviation is greater than or equal to the deviation threshold, it is determined that the transmitting module 210 has abnormal deformation, such as cracks in the optical fiber or contamination of the sampling module. In this case, the weighting coefficient is adjusted. and weighting coefficients The system re-collects the first and second light intensities and recalculates them until the deviation is less than the deviation threshold, thereby avoiding errors caused by abnormal deformation. If the deviation is still greater than or equal to the deviation threshold after three consecutive recalculations, the system determines that the transmitting module 210 has a hardware fault and stops the operation of the transmitting module 210.

[0038] In some embodiments, acquiring a dual-wavelength dark-field image and correcting the dual-wavelength dark-field image according to compensation data G includes determining the required reverse displacement and grayscale gain factor for each pixel based on the compensation data G; the reverse displacement is equal in magnitude and opposite in direction to the lateral correction component M reflected in the compensation data G; the grayscale gain factor is the reciprocal of the light intensity attenuation reflected in the compensation data G; specifically, the compensation data G includes an angle correction component. The system includes a lateral correction component M and a light intensity attenuation. The light intensity attenuation is determined by the ratio of the second light intensity after the control module 210 moves to the second light intensity when the control module 210 is stationary. The second light intensity when the control module 210 is stationary is set to I3, and the second light intensity after the control module 210 moves is set to I4. Therefore, the light intensity attenuation is I4 / I3. Simultaneously, when the illumination beam tilts, the incident angle of the scattered light returning to the imaging module 300 from each point on the wafer changes synchronously, causing a radial shift in the imaging position of each point on the image plane. The pixels of the dual-wavelength dark-field image are reverse-displaced according to the reverse displacement. For example, for a pixel in the image with coordinates... The pixels that are located from the image center The radial distance is The amount of subpixel displacement required for this pixel for,

[0039]

[0040] Wherein, P is the coefficient that directly converts the angle (radians) into pixel displacement, determined by the focal length of objective lens 310, the focal length of tube lens 320, and the pixel size of acquisition camera 330. In actual testing, it is determined by the actual focal lengths of objective lens 310 and tube lens 320, which will not be elaborated here. This is achieved through the angle correction component. It can eliminate image geometric distortion caused by beam drift. The dual-wavelength dark-field image is corrected by multiplying the pixels of the image by the grayscale gain factor.

[0041] In some embodiments, determining the longitudinal depth of a defect based on a dual-wavelength dark-field image includes: identifying gray-scale anomaly regions in the corrected dual-wavelength dark-field image as defect regions, wherein regions with gray levels greater than the average gray level of the background are identified as gray-scale anomaly regions; obtaining the first gray-scale value of the pixels in the defect region at a first wavelength and the second gray-scale value at a second wavelength; specifically, extracting the first gray-scale value and the second gray-scale value of the defect candidate region under each illumination mode; the first gray-scale value and the second gray-scale value respectively reflect the difference in the scattering response of the defect to short-wavelength and long-wavelength illumination light; short wavelengths are sensitive to surface defects, while long wavelengths have strong penetrating power for subsurface defects; multiple sampling points are divided along the axial direction of the wafer, and each sampling point corresponds to the depth position of a defect. Based on the incident angle, azimuth angle, and differences in scattering intensity of defects at different wavelengths under various illumination modes, the theoretical dual-wavelength grayscale value h of each sampling point under each illumination mode is established. Surface defects are dominated by single scattering, with a large scattering cross-section for short wavelengths and a small scattering cross-section for long wavelengths; subsurface defects are dominated by volume scattering. Limited by the Lambert-Beer law, short wavelengths are strongly absorbed during transmission, while long wavelengths have strong penetration ability. The theoretical dual-wavelength grayscale value h is...

[0042] in, For the first The depth coordinates of each sampling point; For wavelength (first wavelength or second wavelength); Angle of incidence; It is the azimuth angle; The surface scattering weighting coefficient has a value ranging from 0.1 to 1.0, preferably 0.9. This is the volume scattering weighting coefficient, with a value ranging from 0 to 1.0, preferably 0.1; denoted as Fresnel reflectance, the values ​​of which are well known to those skilled in the art and will not be elaborated here; The value is the volume scattering angle distribution function, ranging from 0.1 to 1.0, preferably 0.3; For the medium at wavelength The absorption coefficient at the first wavelength ranges from 0.5 to 5.0. Preferably 1.5 At the second wavelength, its value ranges from 0.001 to 0.1. The preferred value is 0.02. ; To indicate parameters, when (Surface) is 1, otherwise it is 0, and when At that time, the volume scattering weighting coefficient Set to 0; The attenuation term describes the attenuation of the longer wavelength (second wavelength) inside the medium; based on the theoretical dual-wavelength gray value h, the gray values ​​of defects at different depths under the first wavelength and the second wavelength can be obtained.

[0043] The first and second gray values ​​are compared with the theoretical dual-wavelength gray value h to determine the depth sampling point where the defect is located. Specifically, at the i-th sampling point, the first gray value is bi1, the second gray value is bi2, and the theoretical first gray value in the dual-wavelength gray value h is hi1, and the theoretical second gray value is hi2. For each sampling point, a separate criterion E1 for a single transmitting module 210 is calculated.

[0044] The arithmetic mean of the individual criteria E1 of all the emission modules 210 is used to obtain the joint criterion E. Among all the sampling points, the minimum joint criterion E is found to determine the location of the sampling point where the defect is located. Since defects in semiconductor wafers usually only exist on the surface or at the interface of a specific film layer, rather than being continuously distributed throughout the entire depth range, the discrete characteristics of the defect depth distribution can be used to determine the depth of the defect when the number of illumination modes is less than the number of depth sampling points, thus avoiding the problem of low detection efficiency of traditional layer-by-layer scanning.

[0045] In some embodiments, the defect detection method further includes: for each located defect region, dividing the first gray value by the second gray value to obtain the dual-wavelength ratio under each illumination mode; calculating the arithmetic mean of the dual-wavelength ratio under each illumination mode to obtain the dual-wavelength mean; performing linear fitting with the incident angle corresponding to each illumination mode as the abscissa and the dual-wavelength ratio as the ordinate to obtain the dual-wavelength slope; setting the mean threshold to 0.8-1.5, preferably 1.2, and setting the slope threshold to 0.03-0.08, preferably 0.04. If the longitudinal depth of the defect is zero, it is determined to be a surface defect, such as a surface scratch; if the longitudinal depth of the defect is greater than zero and both the mean and slope are less than the corresponding set values, it is determined to be a subsurface defect, such as interlayer particles in a thin film; if the longitudinal depth of the defect is greater than zero and both the mean and slope are greater than or equal to the corresponding set values, it is determined to be a subsurface strong reflection defect, such as a short-circuit protrusion between metal layers. Although both subsurface defects and subsurface strong reflection defects are located below the wafer surface, their scattering physical mechanisms differ. Subsurface defects are dominated by volume scattering in the medium, with a low dual-wavelength ratio that changes gradually with the incident angle, corresponding to thin film deposition defects. Subsurface strong reflection defects, on the other hand, are dominated by single scattering from the high reflectivity of the metal surface, with a high dual-wavelength ratio that changes drastically with the incident angle, corresponding to metal etching residues or interlayer short circuits. By independently identifying subsurface strong reflection defects from the broader category of subsurface defects, it is easier to adjust the parameters of the process equipment and avoid misdiagnosing metal process anomalies as thin film process anomalies.

[0046] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A defect detection method based on dual-wavelength coded illumination, characterized in that, Includes the following steps: The transmitter is divided into multiple transmitter modules along the circumference. Each transmitter module includes an optical fiber sub-bundle and a collimating lens. The optical fiber sub-bundle is used to transmit illumination beams of a first wavelength and a second wavelength. Adjust the illumination mode of each of the aforementioned transmitting modules; the illumination mode includes on / off state, radial displacement, and circumferential misalignment angle; The illumination modes of each of the emission modules are switched sequentially according to a preset sequence, so that the illumination beams of each emission module have different incident angles and azimuth angles relative to the test point on the wafer surface. After each switching of the illumination mode, the first light intensity at each collimating lens and the second light intensity at the fiber sub-bundle are collected; the center position (xc, yc) of the light intensity distribution of the corresponding transmitting module and the beam deviation angle R are obtained based on the first light intensity; the end face normal vector deflection angle n of each transmitting module is obtained based on the second light intensity. Compensation data G is obtained based on the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n. Acquire dual-wavelength dark-field images and correct the dual-wavelength dark-field images according to the compensation data G; The longitudinal depth of the defect is determined based on the dual-wavelength dark-field image.

2. The defect detection method according to claim 1, characterized in that, The lighting mode of each of the aforementioned transmitting modules is adjusted. include, The corresponding transmitter module is controlled to move radially along the wafer, changing the radial distance of the transmitter module relative to the geometric center of the transmitter end, so as to adjust the radial displacement. Control the corresponding transmitting module to deflect in the tangential direction of the circumference to adjust the circumferential misalignment angle; Control the on / off state of the corresponding transmitting module to adjust the switch state.

3. The defect detection method according to claim 1, characterized in that, The step of obtaining the center position (xc, yc) of the light intensity distribution of the corresponding emitting module and the beam deviation angle R based on the first light intensity includes, The first light intensity is acquired by a first sampling module, which includes a photodiode array. The first light intensity is converted into a two-dimensional light intensity distribution by the photodiode array, and the two-dimensional light intensity distribution is weighted by pixel coordinates to obtain the center position (xc, yc). Determine the offset distance of the center position (xc, yc) relative to the geometric center (x0, y0) of the photodiode array; The beam deviation angle R is obtained based on the offset distance and the focal length of the sampling lens.

4. The defect detection method according to claim 1, characterized in that, The deflection angle n of the end face normal vector of each of the transmitting modules is obtained based on the second light intensity. include, The ratio of the intensity of the first wavelength to the intensity of the second wavelength is calculated to obtain the dual-wavelength intensity ratio; wherein, the bending loss of the first wavelength is greater than that of the second wavelength, and when the fiber sub-bundle is bent, the intensity attenuation of the first wavelength is greater than that of the second wavelength. The fiber bending radius of each fiber sub-bundle is obtained based on the dual-wavelength light intensity ratio and the preset bending loss calibration curve. The deflection angle n of the end face normal vector of each transmitting module is obtained based on the bending radius of the optical fiber.

5. The defect detection method according to claim 1, characterized in that, The step of obtaining compensation data G based on the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n includes: Set a deviation threshold; The beam offset of the illumination beam is obtained based on the center position (xc, yc) and the beam deviation angle R. The difference between the beam offset and the deflection angle n of the end face normal vector is calculated and the absolute value is taken to obtain the deviation. If the deviation is less than the deviation threshold, the center position (xc, yc), the beam deviation angle R, and the end face normal vector deflection angle n are fused to form compensation data G; If the deviation is greater than or equal to the deviation threshold, the first light intensity and the second light intensity are re-acquired and recalculated.

6. The defect detection method according to claim 1, characterized in that, The acquisition of dual-wavelength dark-field images and the correction of the dual-wavelength dark-field images based on the compensation data G include, The required reverse displacement and grayscale gain factor for each pixel are determined based on the compensation data G. The pixels of the dual-wavelength dark field image are reverse-displaced according to the reverse displacement amount, and the pixels of the dual-wavelength dark field image are multiplied according to the grayscale gain factor to correct the dual-wavelength dark field image.

7. The defect detection method according to claim 6, characterized in that, The step of determining the longitudinal depth of the defect based on the dual-wavelength dark-field image includes, The gray-scale abnormal regions in the corrected dual-wavelength dark field image are considered as defect regions. Obtain the first grayscale value of the pixel in the defective region at the first wavelength and the second grayscale value at the second wavelength; Multiple sampling points are divided along the axial direction of the wafer; Based on the incident angle, azimuth angle and the difference in scattering intensity of defects to different wavelengths under each illumination mode, the theoretical dual-wavelength gray value h of each sampling point under each illumination mode is established. The first gray value and the second gray value are compared with the theoretical dual-wavelength gray value h to determine the depth sampling point where the defect is located.

8. The defect detection method according to claim 7, characterized in that, It also includes, For each located defect region, the first gray value is divided by the second gray value to obtain the dual-wavelength ratio under each illumination mode; Calculate the arithmetic mean of the dual-wavelength ratios for each of the lighting modes to obtain the dual-wavelength mean; A linear fit is performed with the incident angle corresponding to each illumination mode as the abscissa and the dual-wavelength ratio as the ordinate to obtain the dual-wavelength slope. If the longitudinal depth of the defect is zero, it is determined to be a surface defect; If the longitudinal depth of the defect is greater than zero and both the mean and the slope are less than the corresponding set value, it is determined to be a subsurface defect. If the longitudinal depth of the defect is greater than zero and both the mean and the slope are greater than or equal to the corresponding set values, it is determined to be a subsurface strong reflection defect.

9. A defect detection system based on dual-wavelength coded illumination, characterized in that, For implementing the defect detection method according to any one of claims 1-8, the defect detection system comprises, A wafer stage is used to hold the wafer to be tested. The emitting end includes multiple emitting modules circumferentially surrounding the wafer stage, the emitting modules being used to emit illumination beams of a first wavelength and a second wavelength toward the wafer; The imaging module includes an objective lens, a tube lens, and a camera arranged sequentially along the optical axis to acquire dual-wavelength dark-field images. A control system, connected to the transmitter and the imaging module, is used to set and switch the illumination modes of each transmitter module.

10. The defect detection system according to claim 9, characterized in that, The transmitting module includes an optical fiber sub-bundle, a collimating lens, a first sampling module, and a second sampling module; The fiber sub-bundle is used to transmit illumination beams of the first wavelength and the second wavelength, and the collimating lens is disposed at the output end of the fiber sub-bundle; The first sampling module is located on the side of the collimating lens away from the fiber sub-bundle, in order to collect the first light intensity at the collimating lens; The second sampling module is located on the side of the collimating lens facing the fiber sub-bundle to collect the second light intensity at the fiber sub-bundle.

Images