An ultrasonic wafer defect detection method
By setting decision lines and decision points at the interface wave, the problems of high computational load and missed detection in the inspection of patternless bonded wafers and patternless SOI wafers are solved, and efficient defect detection is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI YOU RUIPU SEMICON EQUIP CO LTD
- Filing Date
- 2026-04-21
- Publication Date
- 2026-06-12
AI Technical Summary
Existing defect detection methods for patternless bonded wafers and patternless SOI wafers are computationally intensive and slow. Furthermore, when imaging small defects, the grayscale values are low, which can easily lead to missed detections or confusion with noise.
By setting multiple judgment lines at the interface wave, the defect location is determined by comparing the vertical coordinate of the judgment point with the threshold, and white or black pixels are output, reducing the amount of data processing and improving detection efficiency.
Significantly reduces data processing time, increases detection speed, reduces missed detections and noise interference, and improves equipment efficiency.
Smart Images
Figure CN122193429A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wafer defect detection technology, and in particular to an ultrasonic wafer defect detection method. Background Technology
[0002] The main objects detected by ultrasonic defect inspection equipment for semiconductor bonded wafers can be divided into three categories: patterned bonded wafers, patternless bonded wafers, and patternless SOI wafers.
[0003] During operation, the ultrasonic transducer is positioned above the wafer under test. Since ultrasonic waves cannot propagate in air, pure water is injected between the transducer and the wafer as a coupling medium. The ultrasonic probe emits ultrasonic signals, which are driven by a drive shaft to reciprocate across the wafer surface, receiving echo signals reflected at bonding interfaces or SOI stack interfaces. The system analyzes the echo waveforms, detects peak values, maps these peak values to grayscale values, generates an image, and ultimately achieves defect detection through image recognition.
[0004] Existing defect detection methods for patternless bonded wafers and patternless SOI wafers are computationally intensive and slow. Moreover, for small defects with a defect area smaller than the acoustic spot area, the grayscale value is low during imaging, resulting in darker output pixels. This can easily lead to missed detections or confusion with noise during image defect recognition. Summary of the Invention
[0005] To address at least some of the problems mentioned above in the prior art, the present invention provides an ultrasonic wafer defect detection method, comprising the following steps: Ultrasonic waves are emitted by an ultrasonic transducer. During their propagation within the wafer, some of the ultrasonic waves are reflected and received, forming waveform signals. These waveform signals include surface waves, interface waves, and bottom surface waves. Analyze the waveform signal and set multiple decision lines at the interface wave. The intersection of the decision lines and the waveform is the decision point. Compare the ordinate of all decision points with the ordinate threshold of the waveform. If the ordinate of one or more decision points is greater than the ordinate threshold of the waveform, the corresponding detection position is determined to be a defect and a white pixel is output. Otherwise, the corresponding detection position is determined not to be a defect and a black pixel is output.
[0006] Furthermore, the wafer is a patternless bonded wafer, a patternless SOI wafer, or a patterned bonded wafer.
[0007] Furthermore, the horizontal axis of the waveform signal represents the propagation time, and the vertical axis represents the voltage amplitude or relative amplitude. The interface wave is located based on the propagation time, and multiple decision lines perpendicular to the horizontal axis are set in the propagation time interval corresponding to the interface wave. The spacing between the multiple decision lines needs to ensure that the vertical coordinate of the decision point corresponding to at least one decision line is greater than 1 / 2 of the peak value.
[0008] Furthermore, the distance between two adjacent decision lines is less than or equal to 1 / 4 wavelength.
[0009] Furthermore, when the corresponding detection location is determined to be a defect, a white pixel with the first grayscale value is output; When the ordinate of all judgment points is lower than the waveform ordinate threshold, it is determined that the detection position is not a defect, and a black pixel with a second gray value is output, where the second gray value is less than the first gray value.
[0010] Furthermore, it also includes: after the ultrasonic transducer scans the entire wafer, all the obtained pixels are stitched together to obtain a wafer image.
[0011] Furthermore, each white or black pixel is distributed according to its corresponding coordinates to output a wafer image.
[0012] Furthermore, it also includes: defect identification based on the grayscale values of pixels in the wafer image, and determining white pixel areas with grayscale values greater than or equal to the defect grayscale value as defects, including: Based on the grayscale threshold for defect determination, the grayscale value corresponding to each pixel of the wafer image is compared with the defect grayscale threshold. Pixels with grayscale values greater than or equal to the defect grayscale threshold are defect pixels, and the area composed of defect pixels is the defect. The defect area, number of defects, and coordinate position of the defects are output.
[0013] Furthermore, the defect grayscale threshold is less than the first grayscale value.
[0014] Furthermore, it also includes: In the software program of the ultrasonic defect detection equipment, select to detect either patternless bonded wafers or patternless SOI wafers. The ultrasonic transducer of the ultrasonic defect detection equipment moves above the wafer to be inspected.
[0015] The present invention has at least the following beneficial effects: The ultrasonic wafer defect detection method of the present invention can significantly reduce the amount of data processing for patternless bonded wafers or patternless SOI wafers, which is only a few tens or even a few hundredths of the amount of data processing for traditional detection methods. This greatly reduces the data processing time, improves the equipment detection efficiency, and eliminates the bottleneck of equipment detection efficiency in terms of data processing speed. This invention determines whether a detection location is a defect based on whether the vertical coordinate value of the determination point is greater than the vertical coordinate threshold of the waveform. All defect pixels are assigned a uniform gray value. Small defects with a defect area smaller than the area of the sound spot are also assigned a high gray value. The pixel brightness is high, making it less likely to miss or be confused with noise. Compared with traditional methods that use the gray value of pixels in the image to determine whether it is a defect, this method is more reliable.
[0016] The ultrasonic wafer defect detection method of the present invention reduces the computational load of image defect recognition algorithms and further improves the working efficiency of the equipment. Attached Figure Description
[0017] To further illustrate the above and other advantages and features of the various embodiments of the present invention, a more specific description of the embodiments of the invention will be presented with reference to the accompanying drawings. It is to be understood that these drawings depict only typical embodiments of the invention and are therefore not intended to limit its scope. In the drawings, identical or corresponding parts will be indicated by identical or similar reference numerals for clarity.
[0018] Figure 1 The flowchart of an ultrasonic wafer defect detection method according to an embodiment of the present invention is shown.
[0019] Figure 2 A waveform signal diagram is shown when a patternless wafer has defects, according to an embodiment of the present invention.
[0020] Figure 3 A waveform signal diagram of a patternless wafer without defects is shown according to an embodiment of the present invention.
[0021] Figure 4 The waveform signal diagram of the prior art is shown. Detailed Implementation
[0022] It should be noted that the components in the accompanying drawings may be shown exaggerated for illustrative purposes and may not be to scale.
[0023] In this invention, the various embodiments are merely intended to illustrate the solutions of the invention and should not be construed as limiting.
[0024] In this invention, unless otherwise specified, the quantifiers “a” and “one” do not exclude scenarios involving multiple elements.
[0025] It should also be noted that, in the embodiments of the present invention, only a portion of the parts or components may be shown for clarity and simplicity. However, those skilled in the art will understand that, under the teachings of the present invention, the required parts or components can be added as needed for specific scenarios.
[0026] It should also be noted that within the scope of this invention, the terms "same", "equal", and "equal to" do not mean that the two values are absolutely equal, but allow for a certain reasonable error. In other words, the terms also cover "substantially the same", "substantially equal", and "substantially equal to".
[0027] It should also be noted that in the description of this invention, the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not explicitly or implicitly suggest that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0028] Furthermore, the embodiments of the present invention describe the process steps in a specific order. However, this is only for the convenience of distinguishing each step, and is not a limitation on the order of each step. In different embodiments of the present invention, the order of each step can be adjusted according to the process.
[0029] The existing implementation methods for waveform peak extraction and discrimination are as follows: 1. Signal Acquisition and Preprocessing: The ultrasonic transducer emits high-frequency ultrasonic waves, which penetrate the sample and are then used to receive echo signals from interfaces with different acoustic impedances. The original signal is a time-domain waveform (A-scan) containing multiple interface reflection peaks. The system improves the signal-to-noise ratio through dynamic filtering and amplification adjustment, providing high-quality input for subsequent peak detection.
[0030] 2. Data gating for target layer selection: To avoid interference from multiple reflections, the system is equipped with a data gate, which only collects echo signals within a specified depth range.
[0031] 3. Peak detection algorithm: Peak identification of echoes within a selected time window at each scan point is commonly achieved using methods including: 3.1 Envelope Peak Detection Method: Extract the envelope of the echo signal, and locate the maximum value point by traversing and comparing it as the effective reflection intensity; 3.2. Quadratic fitting for precise localization: Parabolic fitting is performed on the maximum value and its neighboring sampling points to estimate the peak position and amplitude of the sub-sampling level; 3.3 Peak Extraction Method: When displaying or storing, the maximum and minimum values are retained in each group of sampling points to ensure that key features of high-frequency signals are not lost.
[0032] 4. Grayscale mapping and image generation (C-scan): The peak amplitude of each scan point is converted into grayscale values to form a two-dimensional planar image (C-scan). No reflection corresponds to black (grayscale value 0), and strong reflection corresponds to white (high grayscale value), thus visually presenting the defect distribution. Different peak amplitudes have different grayscale values, and the corresponding defects have differences in brightness.
[0033] For complex images with multiple grayscale values, multiple image recognition algorithms are usually integrated for comprehensive judgment in order to reduce the false negative and false positive rates of defect identification. However, the introduction of multiple algorithms inevitably leads to a significant increase in computational load.
[0034] In the aforementioned peak detection process, it is necessary to traverse and calculate the waveform data within the data gate, resulting in a large data volume. Waveform peak detection thus becomes the main bottleneck in the data processing speed of ultrasonic defect detection equipment for semiconductor bonding wafers.
[0035] Current ultrasonic defect detection equipment for semiconductor bonding wafers largely originates from ultrasonic testing equipment used in other fields such as substrates and biological organisms. Influenced by the technological development path, its waveform analysis, peak grayscale conversion, and image defect recognition methods continue the principles and algorithms of the aforementioned fields.
[0036] In other fields, due to the diverse morphology, uneven composition, and varying defect depths and types of the tested samples, it is usually necessary to perform peak analysis on the waveforms obtained by ultrasonic waves reflected or transmitted at the defects, map the peak values to different gray values, and then convert them into images to obtain images with different gray value differences, thereby depicting both the sample morphology and the defects in detail.
[0037] When this type of equipment is used on patterned bonded wafers, it is also necessary to extract the peak values of waveform signals at different locations and map the peak values to grayscale values, so as to achieve defect imaging while also presenting wafer patterns.
[0038] For patternless bonded wafers and patternless SOI wafers, the composition of the upper and lower wafer layers is highly uniform. Current mainstream advanced chip manufacturing processes generally use silicon materials with a purity of 11 nines (99.999999999%), and in the research and development and production of advanced processes, the purity of silicon wafers can even reach 13 nines (99.99999999999%). The bonding defects of concern exist only at the wafer bonding interface and are mainly bubble defects. When using traditional detection methods, when the defect size is larger than the ultrasonic transducer spot, the bubble causes total reflection of the ultrasonic signal, producing a high-amplitude echo, which is mapped as a high grayscale value and appears as a bright white area in the image; when the defect size is close to or smaller than the spot, the reflected signal is weaker, the echo amplitude is lower, which is mapped as a low grayscale value and appears as a darker area in the image. In actual inspection, users only need to detect bubble defects and do not require a specific image brightness. Therefore, when performing defect detection on the two types of patternless wafers mentioned above, continuing to use the traditional waveform peak extraction method will introduce unnecessary data computation.
[0039] Therefore, this invention proposes an ultrasonic defect detection method for patternless wafers, which reduces computational load and increases detection speed.
[0040] Figure 1 The flowchart of an ultrasonic wafer defect detection method according to an embodiment of the present invention is shown. Figure 2 A waveform signal diagram is shown when a patternless wafer has defects, according to an embodiment of the present invention. Figure 3 A waveform signal diagram of a patternless wafer without defects is shown according to an embodiment of the present invention. Figure 4 The waveform signal diagram of the prior art is shown.
[0041] like Figure 1 As shown, an ultrasonic wafer defect detection method includes the following steps: Step 1: In the software program of the ultrasonic defect detection equipment, select the detection object as a patternless bonded wafer or a patternless SOI wafer.
[0042] The operator selects the type of object to be inspected in the software interface of the ultrasonic defect inspection equipment. This step is crucial, as the system will call different processing algorithm libraries based on the selection. When applied to defect inspection of patternless bonded wafers or patternless SOI wafers, the method of this invention can be used to significantly reduce the amount of data to be processed and improve the inspection speed.
[0043] Step 2: The ultrasonic transducer of the ultrasonic defect detection equipment moves above the wafer to be inspected. Pure water is present between the ultrasonic transducer and the wafer surface. The wafer to be inspected is a patternless bonded wafer or a patternless SOI wafer.
[0044] Step 3: The ultrasonic transducer emits ultrasonic waves, which are transmitted to the wafer through pure water. During the propagation of the ultrasonic waves in the wafer, some of them are reflected and received, forming waveform signals, including surface waves, interface waves and bottom waves.
[0045] After ultrasonic waves enter the wafer, they are reflected at the interfaces of different materials (such as silicon and oxide layers, bonding layers). When the ultrasonic waves propagate to the upper surface of the upper wafer, some of them are reflected, and the ultrasonic transducer receives the reflected waveform, forming surface waves. The remaining ultrasonic waves then propagate further through the upper wafer to the bonding surfaces of the bonding wafers or the interface surfaces of the SOI wafers, where some are reflected and the waveform is received, forming interface waves. The remaining ultrasonic waves continue to propagate downwards, and at the lower surface of the lower wafer, some are reflected and the waveform is received, forming bottom surface waves.
[0046] When there are no defects at the detection location, the ultrasonic waves are almost completely reflected at the bonding or interface and propagate further to the lower wafer, resulting in a smaller peak value for the interface wave. When there are bubble defects at the detection location, some of the ultrasonic waves are reflected and received by the ultrasonic transducer, forming an interface wave with a larger peak value.
[0047] Step 4: Analyze the waveform signal. Set multiple decision lines at the interface wave. The intersection of the decision lines and the waveform is the decision point. Compare the ordinate of all decision points with the waveform ordinate threshold. If the ordinate of one or more decision points is greater than the waveform ordinate threshold, the corresponding detection position is determined to be a defect and a white pixel is output. Otherwise, the corresponding detection position is determined not to be a defect and a black pixel is output.
[0048] When the corresponding detection location is determined to be a defect, a white pixel with a high grayscale value (e.g., 255) is output. When the ordinate of all detection points is lower than the waveform ordinate threshold, the detection location is determined not to be a defect, and a black pixel with a low grayscale value (e.g., 0) is output.
[0049] like Figure 2 and 3 As shown, the horizontal axis of the waveform signal represents the propagation time (μs, microseconds), indicating the time it takes for the ultrasonic wave to propagate through the material. Since the speed of sound is relatively constant in a specific medium, time can also indirectly reflect depth. The vertical axis of the waveform signal represents the voltage amplitude (V) or relative amplitude (dB), representing the intensity of the reflected echo and reflecting the magnitude of the difference in acoustic impedance at the interface.
[0050] In the data processing system, a threshold is set for the waveform's vertical axis for defects. This threshold can be an absolute threshold or a relative threshold to the defect-free area.
[0051] The interface wave is located based on its propagation time, and a decision line perpendicular to the horizontal axis is set within the corresponding propagation time interval of the interface wave. The rules for setting the decision line are as follows: The x-coordinate (propagation time) of the determination line is within the propagation time interval of the bonding surface of the bonding wafer or the bonding surface of the SOI wafer of the waveform signal. The spacing between multiple decision lines needs to ensure that the ordinate of the decision point corresponding to at least one decision line is greater than 1 / 2 of the peak value, such as when the spacing between decision lines is less than or equal to 1 / 4 of the wavelength.
[0052] Furthermore, in defect detection applications involving patterned bonded wafers, some clients prefer that the pattern be clearly displayed on the image during inspection to visualize the defect's location and provide a better visual experience. Other clients, however, are not concerned with the pattern and only focus on the defect itself. These clients require that the pattern on the patterned bonded wafer be displayed entirely in black, while defects are displayed in white, to improve defect identification. The method of this invention is equally applicable to these defect detection applications where pattern display is not a primary concern.
[0053] The existing solution is as follows: Figure 4 As shown, a "data gate" with a preset width spanning the interface defect waveform is set on the waveform signal. The data processing system performs peak finding processing on the received waveform signal within the preset data gate, maps the found peak value to the corresponding gray value according to the set rules, and outputs the pixel with the corresponding gray value.
[0054] Existing technologies require calculations on all waveform data points covered by the data gate to locate peak values, resulting in a large data processing volume and becoming a significant bottleneck for the detection speed of ultrasonic testing equipment. Furthermore, for small defects with an area smaller than the acoustic spot area, the grayscale value is low during imaging, leading to darker output pixels. This can easily result in missed detections or confusion with noise during image defect recognition. To reduce the missed and false detection rates, existing technologies typically require the integration of multiple image recognition algorithms for comprehensive judgment, significantly increasing the computational load.
[0055] The method of this invention only requires determining the magnitude of single-point data and threshold values for several decision points in the waveform, involving only a few calculation steps. The data processing volume is significantly reduced compared to existing technologies, amounting to only tens or even hundreds of percent of the data processed in existing technologies. For ultrasonic defect detection of patternless bonded wafers or patternless SOI wafers, data processing is no longer a bottleneck for equipment detection speed, greatly improving the equipment's detection speed. Furthermore, the method of determining whether a defect exists based on the ordinate of the decision point being greater than the threshold is more reliable than using the grayscale values of pixels in an image, while also reducing the computational load of image defect recognition algorithms, further improving the equipment's working efficiency.
[0056] This invention unifies the grayscale values corresponding to defects, so there is no difference in brightness between defects of different sizes.
[0057] Step 5: After the ultrasonic transducer scans the entire wafer, all the obtained pixels are stitched together to obtain a wafer image.
[0058] When scanning each detection location on the wafer, in addition to outputting pixels (white or black pixels), the coordinates of the detection location are also recorded. During the mosaicking process, the corresponding white or black pixels are distributed to the corresponding positions in the image to be generated. After all pixels have been distributed, the wafer image is output.
[0059] Optionally, in step 6, defect identification is performed based on the grayscale values of the pixels in the wafer image, and white pixel areas with grayscale values greater than or equal to the defect grayscale value are identified as defects.
[0060] In step 4, defects are determined using waveform signals, but the user needs to see the output image. Defects in the image are regions composed of several pixels. Defect recognition is required to obtain the pixel area of different defects and the coordinates of the defect region's center point. The output provides information such as defect area, number of defects, and defect coordinates.
[0061] The specific steps are as follows: The equipment's detection software includes defect judgment threshold parameters. For example, if the grayscale threshold for defect judgment (defect grayscale threshold) is set to 200, then pixels with a grayscale value above 200 are judged as defect pixels. The region composed of defect pixels is then determined as a defect. Furthermore, the detection software can color the defect area or draw its outline, generating information such as the area and coordinate position of each defect, and the total number of defects.
[0062] The detection software colors the defective area (replacing white with other colors) or draws an outline, indicating that the algorithm has identified the corresponding area as a defect. This allows users to confirm the detection results and check for any missed or misjudged areas.
[0063] While some embodiments of the present invention have been described in this application, those skilled in the art will understand that these embodiments are merely illustrative. Numerous variations, alternatives, and improvements will arise in those skilled in the art under the teachings of this invention without departing from its scope. The appended claims are intended to define the scope of the invention and thereby cover methods and structures within the scope of the claims themselves and their equivalents.
Claims
1. A method for ultrasonic wafer defect detection, characterized in that, Includes the following steps: Ultrasonic waves are emitted by an ultrasonic transducer. During the propagation of the ultrasonic waves in the wafer, some are reflected and received, forming waveform signals, including surface waves, interface waves, and bottom waves. The waveform signals are analyzed by setting multiple decision lines at the interface waves. The intersection of the decision lines and the waveform is the decision point. The ordinate of all decision points is compared with the waveform ordinate threshold. If the ordinate of one or more decision points is greater than the waveform ordinate threshold, the corresponding detection position is determined to be a defect and a white pixel is output. Otherwise, the corresponding detection position is determined not to be a defect and a black pixel is output.
2. The ultrasonic wafer defect detection method according to claim 1, characterized in that, The wafer is a patternless bonded wafer, a patternless SOI wafer, or a patterned bonded wafer.
3. The ultrasonic wafer defect detection method according to claim 1, characterized in that, The horizontal axis of the waveform signal represents the propagation time, and the vertical axis represents the voltage amplitude or relative amplitude. The interface wave is located based on the propagation time, and multiple decision lines perpendicular to the horizontal axis are set in the propagation time interval corresponding to the interface wave. The spacing between the multiple decision lines needs to ensure that the vertical coordinate of the decision point corresponding to at least one decision line is greater than 1 / 2 of the peak value.
4. The ultrasonic wafer defect detection method according to claim 3, characterized in that, The distance between two adjacent decision lines is less than or equal to 1 / 4 wavelength.
5. The ultrasonic wafer defect detection method according to claim 1, characterized in that, When the corresponding detection location is determined to be a defect, a white pixel with the first grayscale value is output. When the ordinate of all judgment points is lower than the waveform ordinate threshold, it is determined that the detection position is not a defect, and a black pixel with a second gray value is output, where the second gray value is less than the first gray value.
6. The ultrasonic wafer defect detection method according to claim 5, characterized in that, Also includes: After the ultrasonic transducer scans the entire wafer, all the obtained pixels are stitched together to obtain a wafer image.
7. The ultrasonic wafer defect detection method according to claim 6, characterized in that, Distribute each white or black pixel according to its corresponding coordinates to output a wafer image.
8. The ultrasonic wafer defect detection method according to claim 6, characterized in that, Also includes: Defect identification is performed based on the grayscale values of pixels in the wafer image. White pixel areas with grayscale values greater than or equal to the defect grayscale value are identified as defects, including: Based on the grayscale threshold for defect determination, the grayscale value corresponding to each pixel of the wafer image is compared with the defect grayscale threshold. Pixels with grayscale values greater than or equal to the defect grayscale threshold are defect pixels, and the area composed of defect pixels is the defect. The defect area, number of defects, and coordinate position of the defects are output.
9. The ultrasonic wafer defect detection method according to claim 8, characterized in that, The defect grayscale threshold is less than the first grayscale value.
10. The ultrasonic wafer defect detection method according to claim 1, characterized in that, Also includes: In the software program of the ultrasonic defect detection equipment, select to detect either patternless bonded wafers or patternless SOI wafers. The ultrasonic transducer of the ultrasonic defect detection equipment moves above the wafer to be inspected.