An auto-focusing system and method suitable for multi-surface patterned wafer defect detection

Abstract

This invention discloses an automatic focusing system and method for detecting defects in multi-surface patterned wafers, including a light source, an off-axis optical path, a detection optical path, a detector, a host computer, and a motion control module. The light output from the light source illuminates the wafer through the off-axis and detection optical paths, is reflected, and received by the detector. By adjusting the off-axis optical path in real time, the reflected light from wafer surfaces at different depths generates controllable lateral displacement on the detector. Combined with an adaptive signal extraction algorithm in the host computer, the light spots on each surface are accurately identified and separated. The defocus position of the target wafer surface is calculated based on the mapping relationship between the light spot displacement and the defocus amount. Automatic focusing of the wafer stage is achieved by controlling the motion control module. This invention aims to overcome the focusing misalignment problem caused by multi-layer surface reflection and etched circuit structures during patterned wafer defect detection.

Description

Technical Field

[0001] This invention belongs to the field of optical inspection technology, specifically relating to an automatic focusing system suitable for defect detection of multi-surface patterned wafers. Background Technology

[0002] Autofocus (AF) technology originated in commercial cameras and, with the rapid development of computer technology in the 21st century, has demonstrated immense application value in medicine, industry, and daily life. In industry, autofocus technology is widely used in Automated Optical Inspection (AOI), with common applications including welded component inspection and circuit board quality inspection. In the semiconductor industry, a major application of autofocus technology is the optical inspection and measurement of wafers used in chip manufacturing.

[0003] To meet yield requirements, most semiconductor inspection equipment currently employs active autofocus (AAF) systems. This involves adding an optical system to the system to convert the defocus amount into a detectable optical physical quantity, which is then calibrated using a motion mechanism. Currently, commercial microscopic focusing modules use a system based on a pupil segmentation scheme for defocus detection. Specifically, a beam of light emitted from an on-axis light source is blocked by an off-center aperture. The light from the unblocked optical axis plane is collimated and focused onto the wafer surface by the objective lens. This beam is then reflected by the wafer surface, propagates along the other side of the optical axis plane, and finally converges onto the sensor. The direction and amount of defocus are determined by analyzing the intensity distribution and grayscale centroid of the reflected beam on the detector.

[0004] In actual testing of patterned wafers, due to the complex structure of chip wafers composed of multiple layers of materials, different surfaces need to be placed on the working surface of the testing system depending on the target being tested. The different surface reflections of the semi-transparent wafer can significantly affect the focusing accuracy. At the same time, the detection light will cause interference and diffraction phenomena after being reflected by the periodic circuit patterns etched in the patterned wafer, making it difficult to calculate the defocusing amount.

[0005] Currently, companies such as WDI, MSG, Prior, Evident, and Olympus have developed commercial focusing modules for microscope tubes. The specific method involves an on-axis light source whose beam is blocked by an off-center aperture. The light from the unblocked optical axis plane is collimated and focused onto the wafer surface by the objective lens. This beam is then reflected by the wafer surface, propagates along the other side of the optical axis plane, and finally converges onto the sensor. The spot of the reflected beam on the sensor changes depending on the position of the objective lens focal plane on the wafer's reflective surface. For example, when the measured surface is on the objective lens focal plane, the spot on the sensor is a point. When the measured surface is above the objective lens focal plane, the spot on the sensor is a semi-circle located to the right of the sensor's central axis; when the measured surface is below the objective lens focal plane, the spot on the sensor is a semi-circle located to the left of the sensor's central axis.

[0006] To meet response speed requirements, different detector styles can be used in such devices. If a photodiode is used as the detector, the system detects the light intensity of the light spot in the left and right quadrants, calculates the difference between the signal values ​​to obtain a curve showing the change in the defocus amount, takes the linear portion as the working area, and uses the feedback signal for focusing.

[0007] If a sensor array, such as a camera, is used as a detector, the signal received by the sensor needs to be calculated to obtain the center of gravity (COG) corresponding to the surface signal spot. This COG is then correlated with the defocusing change to obtain a corresponding curve. The linear portion of this curve is taken as the working area, and the feedback signal is used for focusing. Compared to photodiodes, which only detect intensity signals, sensor arrays provide the position information of the returned spot and support image generation and adjustment of the signal spot.

[0008] When processing multi-layered surface reflective samples, light spots of different sizes reflected back from different surfaces are superimposed on the sensor, causing the solution used in this type of equipment to be unable to accurately determine the specific defocus position of the target surface.

[0009] Patent document CN 115390233 A discloses an autofocus system for an optical microscope, proposing a system for microscopic imaging of multi-layered samples. The system replaces the eccentric aperture with an off-axis annular aperture, enabling the separation and detection of light spots returned from different surfaces. However, the fixed-position eccentric aperture cannot adapt to differences in the thickness of transparent surfaces. If the two surfaces are too close, or if the refractive index of a certain layer changes significantly, the returned light spots may still overlap noticeably. Furthermore, when processing image signals using the annular aperture, difficulties in layering arise, requiring manual judgment or contour extraction using convolution algorithms, which is time-consuming. Finally, assuming each sample surface is smooth and flat, the system does not address the measurement inaccuracies caused by the modulation of the detection spot and the generation of interference diffraction patterns due to the complex patterns between different sample surfaces. Summary of the Invention

[0010] In view of the above, the purpose of this invention is to provide an automatic focusing system and method for detecting defects in multi-surface patterned wafers, aiming to overcome the focusing misalignment problem caused by multi-layer surface reflection and etched circuit structure during patterned wafer defect detection.

[0011] To achieve the above-mentioned objectives, the embodiment provides an automatic focusing system suitable for defect detection of multi-surface patterned wafers, including a light source, an off-axis optical path including an off-axis aperture, an aperture motion control module, a detection optical path, a detector, a host computer, and a stage motion control module. The light output from the light source illuminates the wafer through the off-axis optical path and the detection optical path, and is reflected and received by the detector. The off-axis optical path is adjusted in real time by the aperture motion control module so that the reflected light from the wafer surface at different depths produces a controllable lateral displacement on the detector. Combined with the adaptive signal extraction algorithm in the host computer, the light spots on each surface are accurately identified and separated. The defocus position of the target wafer surface is calculated by the mapping relationship between the light spot displacement and the defocus amount. The automatic focusing of the wafer stage is achieved by controlling the stage motion control module.

[0012] Preferably, the off-axis optical path includes a beam shaping module and an off-axis aperture. The beam shaping module includes a lens and a slit, and the light output from the light source passes through the lens, the slit, and the off-axis aperture in sequence to form an off-axis beam.

[0013] Preferably, the aperture motion control module includes a control board, a micro motor driver, a clamp rack, a gear, and a micro motor. The clamp rack holds the off-axis aperture and is restricted to moving only along the gear transmission direction in cooperation with the gear. The gear is fixed to the micro motor. The control board inputs a signal to control the micro motor driver to drive the micro motor to move the off-axis aperture laterally. The movement range is from the point where the light transmission center of the off-axis aperture coincides with the optical axis of the system to the point where the light transmission center of the off-axis aperture moves to the edge of the system aperture.

[0014] Preferably, the detection optical path includes a beam splitter, a collimating lens, and an objective lens. The off-axis beam output from the off-axis optical path is reflected by the beam splitter and then incident on the collimating lens. The collimated parallel detection beam is coupled into the imaging system and focused onto the surface of the patterned wafer by the objective lens. The detection beam is reflected by multiple surfaces of the patterned wafer, returns from the other side of the optical axis, and is collimated again by the objective lens. The collimated beam is focused by the incident lens, and the beam is focused onto the detector after passing through the beam splitter, forming an image of the slit in the off-axis optical path.

[0015] Preferably, the mapping relationship between the spot displacement and the defocusing amount is as follows: in, Let be the lateral displacement of the light spot corresponding to the nth surface of the patterned wafer on the detector. δ represents the distance from the target wafer surface to the objective lens object-side principal plane, and δ is the defocusing amount of the target wafer surface. For the top to bottom i The thickness of the material on the surface of the layer, Let be the refractive index of the medium on the surface of the i-th layer. , λ represents the focal length of the objective lens and the collimating lens, respectively, and y represents the position of the light transmission center of the off-axis aperture. d This is the distance from the off-axis aperture to the collimating lens.

[0016] Preferably, the adaptive signal extraction algorithm accurately identifies and separates each surface light spot, including: First, the light spot image detected by the detector is converted into a grayscale image. Then, noise cancellation and threshold filtering are performed to remove weak signals generated by high-order diffraction and surface reflection, thus achieving preprocessing. Then, the preprocessed image is filtered for leading clutter on each surface through a verification mechanism to identify the target surface signal. Specifically, the signal is defined and distinguished by setting variables and adjustable parameters. The variables include: the signal start position and end position, the cumulative rise and fall of the signal, the signal peak value and peak coordinates, the current peak-to-valley value, and the signal width. The adjustable parameters include: the amplitude critical parameter, correction coefficient 1, and correction coefficient 2.

[0017] Preferably, the signal is defined and distinguished by setting variables and adjustable parameters, including: A starting point is set, and a first-order forward differential scan is performed on the signal. When the differential value is non-positive, no record is made. When the differential value is positive, the starting position is marked, and the upward trend is tracked. When the differential value changes from positive to non-positive, it is considered to have reached a signal peak. The peak position and peak-to-trough amplitude are recorded. If a higher point appears later, the peak information is updated. Afterward, if the differential value is non-positive, the waveform shows a decline. The decline is accumulated, and when the scan reaches a positive differential position, a decision-making stage is entered: For signals whose amplitude exceeds the amplitude critical parameter, condition 1 is checked: whether the decline depth meets the correction requirement. The result of "coefficient 1 * peak-to-valley value" is considered a valid signal if it is true, otherwise it is not considered a valid signal, the cumulative decreasing value is cleared and the scan continues. For signals with insufficient amplitude to the amplitude threshold parameter, if condition 1 is met, an additional judgment is made on "condition 2: whether the amplitude-signal width relationship defined by correction coefficient 2 is satisfied". Signals that satisfy both conditions 1 and 2 are considered valid signals. If condition 2 is not satisfied, the signal is considered a stray signal and the scan start point is reset. After the conditions are met, the peak position of the start and end points of the surface signal and the intensity center (Center of Gravity, COG) are locked, and the target surface signal is obtained and stored. The amplitude-signal width relationship is as follows: in, Minimum signal width, The peak value of the signal. The correction factor is 2.

[0018] Preferably, separating the light spots on each surface includes: calculating the coordinates of the lateral displacement center of the light spot using the following formula, based on the start and end points, peak positions, and characteristic parameters contained in the identified target surface signal: in, and The coordinates of the start and end positions of the light spot on the target surface are given. for to The Middle i coordinates Δ i Signal strength at that location The coordinates are the center coordinates of the lateral displacement of the light spot.

[0019] Preferably, the automatic focusing of the wafer stage is achieved by controlling the stage motion control module, including: Based on the calculated defocus position, a PID control algorithm is activated, and the motion control module is used to achieve automatic focusing of the wafer stage.

[0020] To achieve the above-mentioned objectives, the embodiments also provide an autofocus method suitable for defect detection in multi-surface patterned wafers, wherein the method uses the above-mentioned autofocus system to achieve autofocus.

[0021] Compared with the prior art, the beneficial effects of the present invention include at least the following: This invention effectively overcomes the focusing misalignment problem caused by the superposition of reflected light spots from multiple surfaces and interference and diffraction effects caused by surface etching circuits in traditional active focusing systems when inspecting patterned wafers. By introducing a real-time adjustable off-axis optical path and a matching signal extraction algorithm, the invention achieves controllable lateral displacement of reflected light from surfaces at different depths on the detector. Combined with an adaptive signal extraction algorithm, it accurately identifies and separates the light spots from each surface. Furthermore, it precisely calculates the defocus position of the target surface through the mapping relationship between the light spot displacement and the defocus amount, and achieves rapid and precise focusing of the wafer stage based on digital PID control. This solution significantly improves the anti-interference capability and measurement accuracy of the focusing system in the inspection of multi-layer, patterned wafers, while reducing the manual judgment or complex image post-processing time required in traditional methods. It improves the automation level and overall efficiency of defect detection, making it suitable for the industrial demand for high-precision, high-efficiency optical inspection in semiconductor manufacturing. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the autofocus system for detecting defects in multi-surface patterned wafers, as shown in the embodiment. Figure 2 This is a schematic diagram of the adjustable off-axis aperture structure shown in the embodiment; Figure 3 This is a schematic diagram of a patterned wafer with multi-surface reflective light spots, as shown in the embodiment. Figure 4 This is a signal distribution diagram of the host computer after image preprocessing, as shown in the embodiment. Figure 5 This is a partial diagram of the algorithm evolution process shown in the embodiment; Figure 6 This is a schematic diagram illustrating the signal range interception strategy in an embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0025] like Figure 1 As shown, the automatic focusing system for defect detection of multi-surface patterned wafers provided in the embodiment includes a laser light source 101, an off-axis optical path formed by a lens 102, a slit 103 and an off-axis aperture 104, a detection optical path composed of a beam splitter 105, a collimating lens 106 and an objective lens 107, a patterned wafer 108, a detector 109, a host computer 110, an aperture motion control module formed by a control board 111, a micro motor driver 112, a fixture rack 201, a gear 202 and a micro motor 203, and a stage motion control module formed by the control board 111, a stage motor driver 113 and a wafer stage 114.

[0026] A laser light source 101 emits a diverging beam, which is focused by a lens 102. This beam has a small divergence angle and then passes through a slit 103 at the conjugate position of the light source. After passing through the slit 103, the beam emits a diverging beam with a small divergence angle through a small aperture. This slit position can be regarded as the emission point of a point light source. The beam emitted from the slit then passes through an off-axis aperture 104 controlled by a micro-motor to form an off-axis beam.

[0027] like Figure 2 As shown, the off-axis aperture 104 is installed in the optical path and fixed by a rack and pinion clamp 201. The rack and pinion clamp 201 is also restricted to moving only along the transmission direction of the gear 202. The gear 202 is fixed to a micro motor 203, and a signal input from the control board 111 controls the micro motor driver 112 to drive the micro motor 203, which in turn moves the off-axis aperture 104 laterally. The range of movement is from the point where the light transmission center 204 of the off-axis aperture coincides with the system optical axis, to the point where the light transmission center of the aperture moves to the edge of the system aperture. In actual use, a suitable off-axis aperture is replaced according to the surface properties and system aperture, and the aperture is moved to a suitable light-blocking position.

[0028] The off-axis beam, after being blocked by the aperture 104, passes through the 50:50 beam splitter 105, and the reflected portion is incident on the collimating lens 106. The distance between the aperture 104 and the collimating lens 106 is determined by the aperture of the objective lens 107. The collimated parallel beam is coupled into the imaging system and converged onto the surface of the patterned wafer 108 by the objective lens 107. The probe beam is reflected by multiple surfaces of the patterned wafer 108, returns from the other side of the optical axis, and is collimated again by the objective lens 107. The collimated beam is focused by the incident lens 106, and after passing through the beam splitter 105, the beam is focused onto the detector 109, forming the image of the slit 103. After receiving the signal, the detector 109 transmits the data to the host computer 110 for processing and analysis.

[0029] Laser light reflected from different surfaces of the patterned wafer 108 is imaged onto the detection surface of the detector 109 after passing through the objective lens 107 and the beam splitter 105. Because the center of the aperture does not coincide with the optical axis, the reflected light from surfaces at different depths exhibits different lateral displacements on the detector 109. For example... Figure 3 As shown, the probe light is emitted from objective lens 107 onto the surface of the patterned wafer 108 sample. It is reflected by multiple interfaces 301, 302, 303, and 304 from the surface to the bottom layer of the sample. The reflected light is collected again by objective lens 107 and finally received by detector 109. 305, 306, 307, and 308 are the reflected light spots corresponding to the surfaces 301, 302, 303, and 304, respectively.

[0030] The lateral displacement distance of the light spot changes with the axial position of the wafer sample, that is, there is a mapping relationship between the light spot displacement and the defocusing amount, which is expressed as: in, Let be the lateral displacement of the light spot corresponding to the nth surface of the patterned wafer on the detector. δ represents the distance from the target wafer surface to the objective lens object-side principal plane, and δ is the defocusing amount of the target wafer surface. For the top to bottom i The thickness of the material on the surface of the layer, Let be the refractive index of the medium on the surface of the i-th layer. , λ represents the focal length of the objective lens and the collimating lens, respectively, and y represents the position of the light transmission center of the off-axis aperture. d This represents the distance from the off-axis aperture to the collimating lens. By calculating the displacement of the reflected light spot position from the detector center at different surfaces, the defocus position of any surface can be calculated using a formula.

[0031] The detector can be a CMOS camera or a line scan camera. The host computer 110 converts the acquired spot images into grayscale images, selecting one row of data from the image, or summing the data column-wise to obtain one row representing the spot distribution on different surfaces. Gaussian filtering is used for noise reduction, and a threshold is set to filter out weak signals generated by higher-order diffraction and surface reflection. The preprocessed signal is as follows: Figure 4 As shown, 401, 402, 403, and 404 represent the signals generated by light spots 305, 306, 307, and 308, respectively. Currently, surface 303 is located at the focal plane of the objective lens, and its signal is located at the center of the sensor, exhibiting the narrowest linewidth and the highest peak value. The signals of the other light spots broaden to varying degrees on detector 109 with changes in defocusing amount, and also show corresponding positional distributions on the sensor according to the defocusing direction. Simultaneously, the surface signals inside the wafer are superimposed with coherent stray signals generated by the interaction between the circuit structure and the focusing probe light, which are received by the sensor in the far field. Therefore, the calculation of the surface signal spot displacement needs to remove the influence of such stray signals to accurately extract the target surface signal.

[0032] The adaptive signal extraction algorithm primarily filters leading clutter from each surface through a verification mechanism to identify signal characteristics of the target surface. The signal processing algorithm defines and distinguishes signals using several variables and adjustable parameters. Specific variables include: signal start and end positions, cumulative rise and fall amounts, peak value and peak coordinates, current peak-to-valley value, and signal width. Adjustable parameters include: amplitude threshold parameter, correction coefficient 1, and correction coefficient 2. The specific process includes: A starting point is set, and a first-order forward differential scan is performed on the signal. When the differential value is non-positive, no record is made. When the differential value is positive, the starting position is marked, and the upward trend is tracked. When the differential value changes from positive to non-positive, it is considered to have reached a signal peak, and the peak position and peak-to-trough amplitude are recorded. If a higher point appears later, the peak information is updated. Afterward, if the differential value is non-positive, the waveform shows a decline. The decline is accumulated, and when the scan reaches a positive differential position again, a decision-making stage is entered: For signals whose amplitude exceeds the amplitude critical parameter, a judgment is made on "Condition 1: Does the decline depth meet the correction coefficient 1 * peak-to-trough value?" The result is... If yes, it is confirmed as a valid signal; otherwise, it is not considered a valid signal. The cumulative value of the decline is cleared and the scanning continues. For signals with amplitudes below the amplitude threshold parameter, if condition 1 is met, an additional judgment is made on "condition 2: whether the amplitude-signal width relationship defined by correction coefficient 2 is satisfied". Signals that satisfy both conditions 1 and 2 are considered valid signals. If condition 2 is not satisfied, the signal is considered a spurious signal, and the scanning start point is reset. After the conditions are met, the start and end points and peak positions of the surface signal are locked, and the target surface signal is obtained and stored, for example, in a register, and the detection and extraction of the next surface signal begins.

[0033] The amplitude-signal width relationship is as follows: in, Minimum signal width, The peak value of the signal. The correction factor is 2. The value of k2 is determined by the number of surface layers to be detected and the global peak intensity received by the sensor, and needs to be set manually.

[0034] Without loss of generality, for Figure 4 The second surface signal 402 is used for defocus detection. The amplitude critical parameter is set to 100, the correction coefficient 1 is set to 0.5, and the correction coefficient 2 is set to 3000. The algorithm operation process is as follows: Figure 5 As shown. The forward differential result becomes positive after reaching point 501, so the starting point 501 is recorded and the scan continues. After reaching point 502, the differential result becomes negative, indicating a peak value has been reached. The current peak coordinates and values ​​are recorded, and the accumulating decrease value begins. At point 503, the differential value becomes positive again. Since the signal amplitude is less than 100 at this point, and the decrease is less than 0.5 times (correction factor 1) of the increase from the starting point 501 to the peak point 502, it is not considered a valid signal. The accumulated decrease value is cleared, and the scan continues. When scanning to point 504, the peak values ​​at 504 and 502 are compared. 504 becomes the new signal peak. The current peak coordinates and values ​​are recorded, and the accumulating decrease value begins. At point 505, the differential value becomes positive again. Since the signal amplitude is greater than 100 at this point, and the decrease exceeds 0.5 times (correction factor 1) of the increase from the starting point 501 to the peak point 504, it is considered a valid signal. At this point, 501 and 505 are the start and end positions of the second surface signal, and 504 is the signal peak position. Data can be recorded and subsequent calculations can be performed.

[0035] If focusing is required on the surface after the second surface, the algorithm continues scanning data backward after identifying the second surface, with position 505 as the new starting point. Scanning to position 506 is considered a new peak. However, since the signal amplitude formed by 505, 506, and 507 is less than 100, and it satisfies condition 1 (the decrease in amplitude from 506 to 507 is greater than the increase in amplitude from 505 to 506 * 0.5), but does not satisfy the width range of condition 2, it is considered a stray signal. The algorithm then scans backward from 507. Similarly, the signal formed by 507, 508, and 509 satisfies condition 1 but not condition 2. The algorithm continues scanning backward from 509 until a subsequent surface signal that satisfies the conditions is found.

[0036] After determining the start and end positions of the signal corresponding to target surface 302, it is necessary to calculate the center position of the signal on that surface to calculate the defocus amount corresponding to that surface. The signal center position is directly determined by the signal peak position 504, or it can be obtained by calculating the gray-level weighted average from the pixel coordinates. The algorithm provides three signal weighting ranges: 1. Compare the values ​​at point 501 and point 505, find point 601 on the side from the smaller point 501 to the peak 504 that is closest to the value of point 505, and perform a weighted average calculation from 601 to 505; 2. Compare the coordinate distances of point 501 and point 505 from the peak position 504, find point 602 on the side from the farther point 501 to the peak 504 that has the same coordinate distance as the coordinate distance between point 505 and the peak 504, and perform a weighted average calculation from 602 to 505; 3. Compare the values ​​at point 501 and point 505, extend the line from the larger point 505 to the peak 504 using the slope of the straight line from 504 to 505 or a polynomial fitting until its value is close to that of 501, obtaining point 603, and perform a weighted average calculation from 501 to 603. In actual use, the specific weighting range method should be selected according to the situation.

[0037] The specific formula for weighted calculation of the center coordinates of the lateral displacement of the light spot is as follows: in, and The coordinates of the start and end positions of the light spot on the target surface are given. for to The Middle i coordinates Δ i Signal strength at that location The coordinates (grayscale value) of the center of the lateral displacement of the light spot.

[0038] In the actual focusing process, the wafer surface position needs to be manually adjusted for calibration first to determine the coordinates of the center position of the surface signal at the focus position. During subsequent scanning, the center position of the surface spot is obtained by weighted averaging of the coordinates using an algorithm. The actual defocusing amount can then be calculated using the formula that the lateral displacement distance of the spot center changes with the axial position of the wafer sample. This data is then transmitted to the detection control board 111 for stage motion control. The control board 111 initiates a precision motion control sequence based on a digital PID control algorithm. The control board 111 uses the target position that the wafer stage needs to reach (i.e., the absolute coordinates obtained during defocusing calibration) as the setpoint and uses the lateral displacement coordinates of the spot output by the image processing algorithm after acquiring the image from the current focusing position as the feedback value. It calculates the positional deviation in real time and then executes a position-based PID algorithm, combining the immediate response of the proportional term to the current deviation and the integral term to the historical accumulated deviation. The ability to eliminate differences and the damping effect of the differential term on future trends are used to calculate an output quantity characterizing the strength of the control action. After limiting and conversion, this output quantity is precisely mapped to the control signal of the stage motor driver 113. The sign of the control signal determines the level state of the direction signal, while its absolute value determines the output frequency of the pulse signal, thereby realizing the synchronous control of the motor speed and direction. The stage motor driver 113 drives the motor in the wafer stage 114 to perform the corresponding angular displacement according to the received pulse sequence and direction signal, and converts it into the precise micro-displacement of the wafer stage through the mechanical transmission mechanism. During this process, the control board 111 updates the position feedback with a fixed sampling period and iteratively executes PID calculations to dynamically adjust the pulse output characteristics until the position error converges within the allowable tolerance range, so that the objective lens 107 is accurately aligned with the wafer plane, completing the autofocus process.

[0039] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic focusing system suitable for defect detection in multi-surface patterned wafers, characterized in that, It includes a light source, an off-axis optical path containing an off-axis aperture, an aperture motion control module, a detection optical path, a detector, a host computer, and a stage motion control module; The light output from the light source illuminates the wafer through the off-axis optical path and the detection optical path, and is reflected and received by the detector. The off-axis optical path is adjusted in real time by the aperture motion control module, so that the reflected light from the wafer surface at different depths produces a controllable lateral displacement on the detector. Combined with the adaptive signal extraction algorithm in the host computer, the light spots on each surface are accurately identified and separated. The defocus position of the target wafer surface is calculated by the mapping relationship between the light spot displacement and the defocus amount. The automatic focusing of the wafer stage is achieved by controlling the stage motion control module.

2. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 1, characterized in that, The off-axis optical path includes a beam shaping module and an off-axis aperture. The beam shaping module includes a lens and a slit. The light output from the light source passes through the lens, the slit, and the off-axis aperture in sequence to form an off-axis beam.

3. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 2, characterized in that, The aperture motion control module includes a control board, a micro motor driver, a clamp rack, a gear, and a micro motor. The clamp rack holds the off-axis aperture and is restricted to moving only along the gear transmission direction. The gear is fixed to the micro motor. The control board inputs a signal to control the micro motor driver to drive the micro motor to move the off-axis aperture laterally. The range of movement is from the point where the light transmission center of the off-axis aperture coincides with the optical axis of the system to the point where the light transmission center of the off-axis aperture moves to the edge of the system aperture.

4. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 1, characterized in that, The detection optical path includes a beam splitter, a collimating lens, and an objective lens. The off-axis beam output from the off-axis optical path is reflected by the beam splitter and then incident on the collimating lens. The collimated parallel detection beam is coupled into the imaging system and focused onto the surface of the patterned wafer by the objective lens. The detection beam is reflected by multiple surfaces of the patterned wafer, returns from the other side of the optical axis, and is collimated again by the objective lens. The collimated beam is focused by the incident lens, and the beam is focused onto the detector after passing through the beam splitter, forming an image of the slit in the off-axis optical path.

5. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 1, characterized in that, The mapping relationship between the spot displacement and the defocusing amount is as follows: in, Let be the lateral displacement of the light spot corresponding to the nth surface of the patterned wafer on the detector. δ represents the distance from the target wafer surface to the objective lens object-side principal plane, and δ is the defocusing amount of the target wafer surface. For the top to bottom i The thickness of the material on the surface of the layer, Let be the refractive index of the medium on the surface of the i-th layer. , λ represents the focal length of the objective lens and the collimating lens, respectively, and y represents the position of the light transmission center of the off-axis aperture. d This is the distance from the off-axis aperture to the collimating lens.

6. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 1, characterized in that, The adaptive signal extraction algorithm accurately identifies and separates the light spots on each surface, including: First, the light spot image detected by the detector is converted into a grayscale image. Then, noise cancellation and threshold filtering are performed to remove weak signals generated by high-order diffraction and surface reflection, thus achieving preprocessing. Then, the preprocessed image is filtered for leading clutter on each surface through a verification mechanism to identify the target surface signal. Specifically, the signal is defined and distinguished by setting variables and adjustable parameters. The variables include: the signal start position and end position, the cumulative rise and fall of the signal, the signal peak value and peak coordinates, the current peak-to-valley value, and the signal width. The adjustable parameters include: the amplitude critical parameter, correction coefficient 1, and correction coefficient 2.

7. The autofocusing system for defect detection in multi-surface patterned wafers according to claim 6, characterized in that, Signals are defined and differentiated by setting variables and adjustable parameters, including: Set a starting point and perform a first-order forward differential scan of the signal. When the differential value is non-positive, no record is made. When the differential value is positive, mark the starting position and track the upward trend. When the differential value changes from positive to non-positive, it is considered to have reached a signal peak. Record the peak position and peak-to-trough amplitude. If a higher point appears later, update the peak information. Afterward, if the differential value is non-positive, the waveform will show a decline. Accumulate the decline and enter the decision-making stage when the differential value is positive again: For signals whose amplitude exceeds the amplitude critical parameter, determine "Condition 1: Does the decline depth meet the correction coefficient 1*?" The peak-to-valley value is checked. If the result is positive, the signal is considered valid; otherwise, it is not considered valid. The cumulative value of the decline is cleared and the scan continues. For signals with insufficient amplitude to meet the amplitude threshold parameter, if condition 1 is met, an additional check is performed: "Condition 2: Does it meet the amplitude-signal width relationship defined by correction coefficient 2?" Signals that meet both conditions 1 and 2 are considered valid. If condition 2 is not met, the signal is considered a stray signal, and the scan start point is reset. After the conditions are met, the start and end points, peak position, and intensity center of the surface signal are locked, and the target surface signal is obtained and stored. The amplitude-signal width relationship is as follows: in, Minimum signal width, The peak value of the signal. The correction factor is 2.

8. The autofocusing system for defect detection in multi-surface patterned wafers according to claim 7, characterized in that, Separate the light spots from each surface, including: calculating the coordinates of the lateral displacement center of the light spot based on the start and end points, peak positions, and characteristic parameters of the identified target surface signal using the following formula: in, and The coordinates of the start and end positions of the light spot on the target surface are given. for to The Middle i coordinates Δ i Signal strength at that location The coordinates are the center coordinates of the lateral displacement of the light spot.

9. The automatic focusing system for defect detection of multi-surface patterned wafers according to claim 7, characterized in that, Automatic focusing of the wafer stage is achieved by controlling the stage motion control module, including: Based on the calculated defocus position, a PID control algorithm is activated, and the motion control module is used to achieve automatic focusing of the wafer stage.

10. An automatic focusing method suitable for defect detection in multi-surface patterned wafers, characterized in that, The method employs the autofocus system described in any one of claims 1-9 to achieve autofocus.

Images