Method, device and method for detecting bubbles in wafer bonding process
By collecting and processing the vacuum value sequence during the wafer bonding process, bubbles can be identified, solving the problem of lack of bubble detection in existing technologies and improving the stability and production efficiency of the bonding process.
Patent Information
- Application Number
- CN202611544867.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-03
AI Technical Summary
The existing hybrid bonding process lacks a bubble detection mechanism during wafer bonding, which causes impurities on the bottom chuck to affect the bonding effect of multiple batches of products, and losses have already occurred when relying on subsequent process measurement stations for detection.
By collecting historical and current chuck adsorption vacuum values from wafer bonding processes, a vacuum value sequence is formed. Bubbles are identified through local regression and difference processing, including dividing the vacuum value change stages, weighted summation, and warming processing to improve detection accuracy.
This technology enables timely detection of bubbles during wafer bonding, reducing bonding defects caused by bubbles, improving production stability and efficiency, and minimizing losses.
Smart Images
Figure CN123035923A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of bonding, and particularly relates to a method, apparatus and bonding method for detecting bubbles in wafer bonding process. Background Technology
[0002] An image sensor (CMOS Image Sensor, CIS) is a photoelectric conversion device widely used in the consumer electronics field.
[0003] Existing image sensors typically employ a stacking process to fabricate pixel wafers and logic wafers separately. The pixel wafers are primarily responsible for light sensing, while the logic wafers are mainly responsible for image processing. The pixel and logic wafers are then hybrid-bonded to obtain the image sensor.
[0004] Existing hybrid bonding processes typically involve first securing the logic layer with a bottom chuck and the pixel layer with a top chuck, with the bottom chuck holding the logic layer in place via vacuum. Subsequently, the bonding surfaces of the pixel and logic layers are aligned, followed by room temperature pre-pressing and annealing to bond the pixel and logic layers together.
[0005] However, the hybrid bonding process has the following drawbacks: During bonding, the top chuck, after fixing the pixel layer, flips over to face the logic layer with the bonding surface. The bottom chuck is located below the top chuck. When the top chuck flips the pixel layer, impurity particles are easily thrown off and fall onto the bottom chuck. Additionally, because the bottom chuck is located below, wafer fragments and other impurities remaining from the previous hybrid bonding process can deposit on it. If impurities are present on the bottom chuck, bubbles can form between the logic layer and the pixel layer when the bottom chuck adsorbs the logic layer, negatively impacting the bonding effect between them. Furthermore, the hybrid bonding process lacks a bubble detection mechanism, relying on measurement stations in subsequent processes for detection. By the time these stations detect bubbles and notify the bonding process to polish and clean the chuck, the impurities on the chuck have already affected the bonding of multiple batches of products, resulting in significant losses. Summary of the Invention
[0006] Therefore, one objective of this application is to provide a method and apparatus for detecting bubbles during the bonding process, thereby identifying potential bubbles during wafer bonding. A second objective of this application is to provide a bonding method that applies the aforementioned bubble detection method during the bonding process, enabling timely detection of bubbles present during the bonding process.
[0007] In a first aspect, this application provides a method for detecting bubbles during wafer bonding, including: Collect the adsorption vacuum value of the chuck during the previous wafer bonding process to form a historical vacuum value sequence; the previous wafer bonding process is a wafer bonding process without bubble generation. Collect the adsorption vacuum value of the chuck during the current wafer bonding process to form a current vacuum value sequence; The first slow drift sequence of the historical vacuum value sequence is determined by local regression. The difference between the historical vacuum value sequence and the first slow drift sequence is used to obtain the corrected historical vacuum value sequence. The second slow drift sequence of the current vacuum value sequence is determined by local regression, and the difference between the current vacuum value sequence and the second slow drift sequence is used to obtain the corrected current vacuum value sequence. The difference between the current vacuum value sequence and the historical vacuum value sequence is calculated to obtain the inter-piece difference sequence; Based on the inter-wafer difference sequence, bubble detection is performed on the current wafer bonding process.
[0008] Optionally, the historical vacuum value sequence and the current vacuum value sequence are divided into several vacuum value change stages, which include the bottom wafer adsorption stage, the chamber actuation stage, the top wafer adsorption stage, and the bonding stage.
[0009] Optionally, the process of detecting bubbles in the current wafer bonding process based on the inter-wafer difference sequence includes: The mean inter-chip difference of the inter-chip difference sub-sequence corresponding to several vacuum value change stages is determined respectively; wherein, the several inter-chip difference sub-sequences are obtained by splitting the inter-chip difference sequence through several vacuum value change stages; The weighted summation of the mean inter-chip difference values corresponding to several vacuum value change stages yields the weighted inter-chip difference value. Bubble detection is performed based on the weighted value of the inter-piece difference.
[0010] Optionally, the process of bubble detection based on the weighted value of the inter-piece difference includes: Calculate the mean and standard deviation of the weighted values of the inter-slice differences; Early warning lines and / or critical lines are formed based on the mean and standard deviation of the weighted values of the inter-slice differences; Bubbles are detected based on the warning line and / or critical line.
[0011] Optionally, before using the inter-slice difference sequence for bubble detection, the method further includes: The inter-slice difference sequence is subjected to Winsorization.
[0012] Optionally, it also includes: Obtain maintenance information from the bonding machine and determine whether the current wafer bonding process is one of the previous wafer bonding processes after maintenance. If so, directly set the bubble detection result of the current wafer bonding process to bubble-free.
[0013] Secondly, this application provides a device for detecting bubbles during wafer bonding, comprising: The first acquisition module is used to acquire the adsorption vacuum value of the chuck during the previous wafer bonding process and form a historical vacuum value sequence. The second acquisition module is used to acquire the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence; The extraction module is used to extract the vacuum value difference between the historical vacuum value sequence and the current vacuum value sequence to form an inter-chip difference sequence; The detection module is used to detect bubbles in the current wafer bonding process based on the inter-wafer difference sequence.
[0014] Thirdly, this application provides an electronic device, including the bubble detection device described above during the wafer bonding process.
[0015] Fourthly, this application provides a computer-readable storage medium storing at least one piece of program code, which is executed by a processor to implement the method for detecting bubbles during wafer bonding as described in any of the preceding claims.
[0016] Fifthly, this application provides a bonding method, comprising: Preprocessing is performed on logic wafers and pixel wafers; The logic wafer is secured to the bottom chuck in the bonding chamber; The pixel wafer is fixed on the top chuck of the bonding chamber, and the logic wafer and the pixel wafer are aligned. Pre-lamination bonding is performed on logic wafers and pixel wafers; Annealing is performed on logic wafers and pixel wafers; The adsorption vacuum value of the chuck during the previous wafer bonding process is acquired to form a historical vacuum value sequence; Based on the collected adsorption vacuum values during the current wafer bonding process, a current vacuum value sequence is formed; The first slow drift sequence of the historical vacuum value sequence is determined by local regression. The difference between the historical vacuum value sequence and the first slow drift sequence is used to obtain the corrected historical vacuum value sequence. The second slow drift sequence of the current vacuum value sequence is determined by local regression, and the difference between the current vacuum value sequence and the second slow drift sequence is used to obtain the corrected current vacuum value sequence. The difference between the current vacuum value sequence and the historical vacuum value sequence is calculated to obtain the inter-piece difference sequence; Based on the inter-wafer difference sequence, bubble detection is performed on the current wafer bonding process.
[0017] The technical solution provided in this application has unexpected technical effects, including: (1) This application provides a method for detecting bubbles during wafer bonding. The method involves acquiring the adsorption vacuum value of the chuck during a bubble-free wafer bonding process to form a historical vacuum value sequence. Then, the real-time adsorption vacuum value of the chuck during the current bonding process is used to form a current vacuum value sequence. By subtracting the current vacuum value sequence from the historical vacuum value sequence, an inter-wafer difference sequence is obtained. This inter-wafer difference sequence can reflect the difference in vacuum value between the current wafer bonding process and a bubble-free wafer bonding process. This vacuum value difference can be used to identify whether bubbles are generated during the current wafer bonding process.
[0018] (2) As a further technical improvement, the adsorption vacuum value acquisition process is divided into several vacuum value change stages, and the current vacuum value sequence and the historical vacuum value sequence are divided into current vacuum value subsequence and historical vacuum value subsequence corresponding to several vacuum value change stages. When the difference between the current vacuum value sequence and the historical vacuum value sequence is obtained to obtain the inter-segment difference subsequence, the difference between the current vacuum value subsequence and the historical vacuum value subsequence corresponding to the same vacuum value change stage is obtained respectively. The inter-segment difference subsequence can be obtained by combining several inter-segment difference subsequences in time sequence. The adsorption vacuum value of the chuck varies at different stages. For example, it differs during the bottom wafer adsorption stage (the bottom chuck adsorbs logic wafers), the chamber actuation stage (a bonding machine typically includes several bonding chambers; the actuation of other chambers affects the adsorption vacuum value of the bottom chuck in the current chamber), the top wafer adsorption stage (the top chuck adsorbs pixel wafers), and the final wafer bonding stage. Therefore, in this application, by dividing the vacuum value into subsequences according to the vacuum value change stages, a final inter-wafer difference subsequence is formed. A weighted sum of the inter-wafer difference averages from several inter-wafer difference subsequences is obtained to obtain a weighted inter-wafer difference value. This can mitigate the impact of the inter-wafer difference at a certain vacuum value change stage on the bubble detection results, ensuring the stability and accuracy of the bubble detection results.
[0019] (3) As a further technical improvement, local regression is adopted to eliminate the slow drift data caused by the environment / air pressure in the historical vacuum value sequence and the current vacuum value sequence, so as to avoid the influence of the environment / air pressure on the bubble detection results.
[0020] (4) As a further technical improvement, the inter-slice difference sequence is subjected to Wunder processing to suppress the influence of extreme noise points on the bubble detection results. For example, it prevents noise factors such as power fluctuations and vibration fluctuations in the equipment production process from causing instantaneous changes in the adsorption vacuum value and interfering with the bubble detection results. Attached Figure Description
[0021] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 A flowchart illustrating the method for detecting bubbles during wafer bonding in the first embodiment of this application; Figure 2 This is a schematic diagram of a bonding cavity provided in an embodiment of this application; Figure 3 A schematic diagram illustrating different vacuum value changes during the bonding process, provided as an embodiment of this application; Figure 4 A schematic diagram illustrating discrete adsorption vacuum values obtained by sampling at different vacuum value change stages during the bonding process, as provided in an embodiment of this application; Figure 5 A schematic diagram illustrating the continuous adsorption vacuum value obtained by sampling at different vacuum value change stages during the bonding process, as provided in an embodiment of this application; Figure 6 This application provides a control chart based on the adsorption vacuum value collected during the wafer bonding process, as an embodiment of the present application. Figure 7 A control chart provided for one embodiment of this application, formed based on the inter-wafer difference between the previous wafer bonding process and the current wafer bonding process; Figure 8 A flowchart illustrating the method for detecting bubbles during wafer bonding in the second embodiment of this application; Figure 9 A flowchart illustrating a method for detecting bubbles during wafer bonding, as provided in the third embodiment of this application. Figure 10 A schematic diagram of the adsorption vacuum value of the bottom chuck after maintenance, provided in an embodiment of this application; Figure 11 This is a schematic diagram illustrating the bubble detection effect provided in one embodiment of this application; Figure 12 A structural block diagram of a bubble detection device during wafer bonding process provided in an embodiment of this application; Figure 13A structural block diagram of a bubble detection device during wafer bonding process provided in another embodiment of this application; Figure 14 This is a structural block diagram of an electronic device provided in an embodiment of this application; Figure 15 This is a flowchart of a bonding method provided in an embodiment of this application.
[0023] The attached figures are labeled as follows: 11: Bonding chamber; 12: Bottom chuck; 13: Top chuck; 14: Logic wafer; 15: Pixel wafer; 20: Detection device; 21: First acquisition module; 22: Second acquisition module; 23: Extraction module; 24: Detection module; 25: Warming treatment module; 30: Electronic device; 31: Processor; 32: Memory. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0025] The reasons why foreign objects cause bubbles in the wafer bonding process are as follows: When there is a foreign object in the bottom chuck, and the bottom chuck adsorbs the logic layer, since the logic layer is not a rigid body, the foreign object will push up the area where the logic layer contacts the foreign object, causing the logic layer to form a local bulge.
[0026] The mechanism of the hybrid bonding pre-compression bonding process is as follows: the bonding surfaces of both the logic layer and the pixel layer are hydrophilically activated SiO2, with their surfaces covered with -OH groups. When the bonding surfaces of the logic layer and the pixel layer are bonded, they spontaneously "attract" each other through hydrogen bonds and van der Waals forces between the -OH groups. This bonding process requires that the bonding surfaces of the logic layer and the pixel layer gradually come into full contact, starting from the center or edge. Gas at the bonding interface is gradually expelled, and ultimately, there is no gas between the bonding surfaces of the logic layer and the pixel layer, resulting in good contact.
[0027] If a localized protrusion forms in the logic layer, during pre-press bonding with the pixel layer, this protrusion will contact the pixel layer first, causing the pixel layer to also be lifted and form a localized protrusion (the pixel layer is not a rigid body). This will result in poor contact at the bonding surfaces of the logic layer and the pixel layer in this protruding area, forming a ring-shaped seam. Since both the bonding surfaces of the logic layer and the pixel layer are hydrophilic, if the area around the ring-shaped seam adheres first, the gas at the seam cannot escape, thus forming a bubble.
[0028] The purpose of this application is to develop a method for detecting bubbles, identifying the generation of bubbles during the bonding process, and preventing impurities on the bottom chuck from affecting the bonding of multiple batches of wafers.
[0029] Figure 1 This is a flowchart illustrating a method for detecting bubbles during wafer bonding, as provided in the first embodiment of this application. See also... Figure 1 The method and steps include: S101. Collect the adsorption vacuum value of the chuck during the previous wafer bonding process to form a historical vacuum value sequence; the previous wafer bonding process is a wafer bonding process without bubble generation.
[0030] S102. Collect the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence.
[0031] S103. Use local regression to determine the first slow drift sequence of the historical vacuum value sequence, and subtract the historical vacuum value sequence from the first slow drift sequence to obtain the corrected historical vacuum value sequence.
[0032] S104. Use local regression to determine the second slow drift sequence of the current vacuum value sequence, and subtract the current vacuum value sequence from the second slow drift sequence to obtain the corrected current vacuum value sequence.
[0033] S105. Subtract the current vacuum value sequence from the historical vacuum value sequence to obtain the inter-piece difference sequence.
[0034] S106. Based on the inter-wafer difference sequence, perform bubble detection on the current wafer bonding process.
[0035] This application provides a method for detecting bubbles during wafer bonding. The method involves acquiring the adsorption vacuum value of the chuck during a bubble-free wafer bonding process to form a historical vacuum value sequence. Then, the real-time adsorption vacuum value of the chuck during the current bonding process is used to form a current vacuum value sequence. By subtracting the current vacuum value sequence from the historical vacuum value sequence, an inter-wafer difference sequence is obtained. This inter-wafer difference sequence reflects the difference in vacuum value between the current wafer bonding process and a bubble-free wafer bonding process. This vacuum value difference can be used to identify whether bubbles have formed during the current wafer bonding process.
[0036] In some embodiments, steps S101 and S102 collect the adsorption vacuum value of the bottom chuck in the same bonding chamber of the same bonding machine.
[0037] Please see Figure 2 Specifically, the bonding chamber 11, bottom chuck 12, top chuck 13, logic wafer 14, and pixel wafer 15 are shown. The bonding chamber 11 includes a bottom chuck 12 and a top chuck 13, with the top chuck 13 located above the bottom chuck 12. The bottom chuck 12 is used to hold and fix the logic wafer 14, and the top chuck 13 is used to hold and fix the pixel wafer 15.
[0038] See Figure 3 In some embodiments, the process of collecting adsorption vacuum values is divided into several vacuum value change stages. The collected historical vacuum value sequence and the current vacuum value sequence are also divided into several vacuum value change stages, including the bottom wafer adsorption stage, the chamber actuation stage, the top wafer adsorption stage, and the bonding stage.
[0039] The adsorption vacuum value acquisition process is divided into several vacuum value change stages. The current vacuum value sequence and the historical vacuum value sequence are further divided into current vacuum value subsequences and historical vacuum value subsequences corresponding to these stages. When subtracting the current vacuum value sequence from the historical vacuum value sequence to obtain the inter-segment difference subsequence, the current vacuum value subsequence is subtracted from the historical vacuum value subsequence corresponding to the same vacuum value change stage to obtain the inter-segment difference subsequences corresponding to these stages. Combining these inter-segment difference subsequences in time sequence yields the final inter-segment difference sequence. The adsorption vacuum value of the chuck varies at different stages. For example, the adsorption vacuum value differs during the bottom wafer adsorption stage (the bottom chuck adsorbs logic wafers), the chamber actuation stage (a bonding machine typically includes several bonding chambers, and chamber actuation affects the adsorption vacuum value of the bottom chuck in the current chamber), the top wafer adsorption stage (the top chuck adsorbs pixel wafers), and the final wafer bonding stage. Therefore, in this application, by dividing the vacuum value into current vacuum value subsequences and historical vacuum value subsequences according to the vacuum value change stages, and subtracting them, an inter-wafer difference subsequence is obtained. The inter-wafer difference subsequences corresponding to different vacuum value change stages are combined to form the final inter-wafer difference sequence. By weighted summing of the mean inter-wafer difference values in several inter-wafer difference subsequences, a weighted inter-wafer difference value is obtained. This can reduce the influence of the inter-wafer difference at a certain vacuum value change stage on the bubble detection results to a certain extent, ensuring the stability and accuracy of the bubble detection results.
[0040] See Figure 4 This is a schematic diagram of discrete adsorption vacuum values collected at different vacuum value change stages according to an embodiment of this application. Figure 4 The horizontal axis represents the sampling time, and the sampling interval for the adsorption vacuum value is 1 second. For example, a sampling time of 0:17 on the horizontal axis means that the adsorption vacuum value was obtained after sampling from the sampling start point (0:0) for 0 minutes and 17 seconds. Figure 4 The vertical axis represents the adsorption vacuum value of the bottom chuck obtained from the data collection. Figure 4 In the illustrated embodiment, the adsorption vacuum value on the vertical axis is the adsorption vacuum value obtained by collecting data from Bottom Table BCCP 12_2 vacuum. Bottom Table refers to the bottom chuck, BCCP refers to Bottom Chuck Compensation / Control Parameter, which is the compensation / control parameter group of the bottom chuck, 12_2 refers to the second vacuum of the 12th Bottom Table, and vacuum refers to negative pressure adsorption.
[0041] See Figure 5This is a schematic diagram of continuous adsorption vacuum values collected at different vacuum value change stages, provided in an embodiment of this application. Figure 4 After fitting the discrete points in the data, we can obtain... Figure 5 The curve showing the continuous vacuum value variation is shown.
[0042] In some embodiments, steps S103 to S105 employ local regression to eliminate slow drift data caused by the environment / air pressure in the historical vacuum value sequence and the current vacuum value sequence, thereby avoiding the influence of the environment / air pressure on the bubble detection results and further improving the accuracy of the detection results.
[0043] In some embodiments, in step S103, if the value used in the local regression process is 0.1, the process of eliminating slow-drift data in the historical vacuum value sequence is as follows:
[0044] In the formula: Indicates the first The first wafer bonding process In the first vacuum value change stage The corrected historical vacuum value corresponding to each sampling time step Indicates the first The first wafer bonding process In the first vacuum value change stage The historical vacuum value corresponding to each sampling time step Indicates to In the slow-drift data sequence determined by local regression, and The corresponding slow drift data, This indicates local regression. The sampling time step represents the adsorption vacuum value, such as 1 second for each sampling time step.
[0045] In some embodiments, the first Historical vacuum values corresponding to each vacuum value change stage By combining them in chronological order, we can obtain the [number]. The historical vacuum value subsequence corresponding to each vacuum value change stage The historical vacuum value subsequences corresponding to several vacuum value change stages are... By combining them in chronological order, a historical vacuum value sequence can be obtained. .
[0046] In some embodiments, the first Corrected historical vacuum values corresponding to each vacuum value change stage By combining them in chronological order, we can obtain the [number]. The corrected historical vacuum value subsequence corresponding to each vacuum value change stage The corrected historical vacuum value subsequences corresponding to several vacuum value change stages are... By combining them in chronological order, the corrected historical vacuum value sequence can be obtained. .
[0047] In some embodiments, in step S105, the inter-slice difference sequence The calculation process for the inter-slice difference can be expressed as:
[0048] In the formula: Indicates the first The first wafer bonding process The inter-slice difference subsequence corresponding to each vacuum value change stage The Middle The inter-slice difference corresponding to each sampling time step.
[0049] The inter-chip differences corresponding to a certain vacuum value change stage By combining them in chronological order, the inter-piece difference subsequence corresponding to the vacuum value change stage can be obtained. The inter-slice difference subsequences corresponding to several vacuum value change stages are... After combining them in chronological order, the inter-slice difference sequence can be obtained. .
[0050] Specifically, no. The first wafer bonding process The inter-slice difference subsequence corresponding to each vacuum value change stage It can be represented as: .
[0051] in, This represents the total number of stages of vacuum value change.
[0052] Specifically, no. Inter-wafer difference sequence of the sub-wafer bonding process It can be represented as: .
[0053] In some embodiments, in step S104, if the value of span in the local regression process is 0.1, the process of eliminating slow-drift data in the current vacuum value sequence is as follows:
[0054] In the formula: Indicates the first The first wafer bonding process In the first vacuum value change stage The corrected current vacuum value corresponds to each sampling time step. Indicates the first The first wafer bonding process In the first vacuum value change stage The current vacuum value corresponding to each sampling time step Represents the current vacuum value subsequence The slow-drift data sequence determined after local regression is related to The corresponding slow drift data, This indicates local regression.
[0055] In some examples, This indicates the stage of vacuum value change during the wafer bonding process. , respectively corresponding Figures 3 to 5 The four vacuum value change stages are shown.
[0056] In some examples, The sampling time step represents the adsorption vacuum value, such as 1 second for each sampling time step.
[0057] In some embodiments, several current vacuum values corresponding to a certain vacuum value change stage are used. By combining them in chronological order, we can obtain the current vacuum value subsequence corresponding to the vacuum value change stage. The current vacuum value subsequence corresponding to several vacuum value change stages. By combining them in chronological order, the current vacuum value sequence can be obtained. .
[0058] In some embodiments, several corrections to the current vacuum value are made during a certain vacuum value change phase. By combining them in chronological order, the corrected current vacuum value subsequence corresponding to the vacuum value change stage can be obtained. The current vacuum value subsequence is corrected for several vacuum value change stages. By combining them in chronological order, the corrected current vacuum value sequence can be obtained. .
[0059] Specifically, no. Correction of the current vacuum value subsequence for each vacuum value change stage Represented as: .in Indicates the first The number of sampling time steps in the vacuum value change phase (also the number of sampling time steps in the first phase) (Number of current vacuum values sampled during each vacuum value change phase) Correct the current vacuum value sequence Represented as: .in This indicates the number of stages in which the vacuum value changes.
[0060] In some embodiments, the process of detecting bubbles in the current wafer bonding process based on the inter-wafer difference sequence in step S106 includes: S1061. Determine the mean value of the inter-slice difference sub-sequences corresponding to several vacuum value change stages; wherein, the several inter-slice difference sub-sequences are obtained by splitting the inter-slice difference sequence through several vacuum value change stages.
[0061] S1062. The mean values of inter-chip differences corresponding to several vacuum value change stages are weighted and summed to obtain the weighted values of inter-chip differences.
[0062] S1063. Perform bubble detection based on the weighted value of the inter-piece difference.
[0063] In some embodiments, a weighted inter-slice difference value is obtained by weighted summing of the mean inter-slice difference values in several subsequences of inter-slice difference values. This can reduce the influence of inter-slice difference values during a certain vacuum value change stage on the bubble detection results to a certain extent, ensuring the stability and accuracy of the bubble detection results.
[0064] In some embodiments, the mean inter-slice difference in the inter-slice difference subsequence of step S1061 It can be calculated using the following formula:
[0065] In the formula: Indicates the first The first wafer bonding process Mean value of inter-segment difference during each vacuum value change stage Indicates the first The first wafer bonding process The first inter-chip difference sequence in the vacuum value change stage The inter-slice difference corresponding to the next sampling time point Indicates the first The number of sampling time steps for each vacuum value change phase.
[0066] In some embodiments, the calculation process of the inter-chip difference weighted value in step S1062 is as follows:
[0067] In the formula: Indicates the first Weighted value of inter-wafer difference in the next wafer bonding process This indicates the number of stages in the vacuum value change. Indicates the first Weights for each vacuum value change stage.
[0068] In some embodiments, , , , The values are 0.2, 0.2, 0.3, and 0.3 respectively. Of course, the weight values here are only an example provided by this application. In actual implementation, different weight values can be set adaptively, and this application does not limit this.
[0069] In some embodiments, the process of performing bubble detection based on the inter-piece difference weighted value in step S1063 includes: Step 1: Calculate the weighted mean of the inter-slice differences. and standard deviation .
[0070] Step 2: Form warning lines and / or critical lines based on the mean and standard deviation of the weighted values of the inter-slice differences.
[0071] Step 3: Detect bubbles according to the warning line and / or critical line.
[0072] In some embodiments, in step 1, the mean and standard deviation of the weighted inter-wafer difference values for several wafer bonding processes under historical conditions are calculated. Each wafer bonding process yields a weighted inter-wafer difference value. The weighted inter-wafer difference values from several wafer bonding processes under historical conditions are summarized to obtain a historical weighted inter-wafer difference value sequence. By statistically analyzing the mean and standard deviation of all historical weighted inter-wafer difference values in the historical weighted inter-wafer difference value sequence, or the historical weighted inter-wafer difference values preceding the current bonding process, the magnitude of the historical weighted inter-wafer difference value for the current wafer bonding process is compared to determine whether bubbles exist.
[0073] In some embodiments, the warning line in step 2 is The critical line is Of course, those skilled in the art can set different warning lines and critical lines as needed.
[0074] In step 3, when the weighted value of the inter-wafer difference exceeds the critical line, the system will issue a real-time alarm, and the factory will shut down for maintenance and polish the bottom chuck. When the weighted value of the inter-wafer difference exceeds the warning line, the bonding machine will not stop, and the subsequent measurement machine will be accelerated for analysis.
[0075] See Figure 6 The control chart formed by this application based on the adsorption vacuum value collected during the wafer bonding process shows that if bubble detection is performed directly based on the control chart formed by the adsorption vacuum value of the wafer during the bonding process, the fluctuation of the adsorption vacuum value is large, and it is not possible to accurately identify the bubbles present in the bonding process.
[0076] See Figure 7 This is a control chart generated in this application based on the inter-wafer bonding difference between the previous wafer bonding process and the current wafer bonding process. It can be observed that the control chart generated by the inter-wafer difference can effectively reflect the presence of bubbles, as shown by the abnormal points highlighted in the blue box in the figure.
[0077] Figure 8 This is a flowchart illustrating a method for detecting bubbles during wafer bonding, provided in the second embodiment of this application. See also... Figure 8 ,include: S201. Collect the adsorption vacuum value of the chuck during the previous wafer bonding process to form a historical vacuum value sequence; the previous wafer bonding process is a wafer bonding process without bubble generation.
[0078] See step S101.
[0079] S202. Collect the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence.
[0080] See step S102.
[0081] S203. Use local regression to determine the first slow drift sequence of the historical vacuum value sequence, and subtract the historical vacuum value sequence from the first slow drift sequence to obtain the corrected historical vacuum value sequence.
[0082] See step S103.
[0083] S204. Use local regression to determine the second slow drift sequence of the current vacuum value sequence, and subtract the current vacuum value sequence from the second slow drift sequence to obtain the corrected current vacuum value sequence.
[0084] See step S104.
[0085] S205. Subtract the current vacuum value sequence from the historical vacuum value sequence to obtain the inter-chip difference sequence.
[0086] See step S105.
[0087] S206. Perform Winsorization on the inter-slice difference sequence to obtain the corrected inter-slice difference sequence.
[0088] In some examples provided in this application, the truncated thresholds used for Winsorization are 5% and 95%. The Winsorization process can then be represented by the following formula:
[0089] In the formula: This indicates the first after the Wenser treatment. The first wafer bonding process In the first vacuum value change stage The corrected inter-chip difference value corresponding to each sampling time step. Indicates the first The first wafer bonding process In the first vacuum value change stage The inter-slice difference corresponding to each sampling time step This indicates a Winsorization process.
[0090] In some embodiments, the inter-piece difference corresponding to a certain vacuum value change stage is used to correct several inter-piece differences. By combining them in chronological order, the corrected inter-piece difference subsequence can be obtained. The corrected inter-piece difference subsequences corresponding to several vacuum value change stages are used. By combining them in chronological order, the corrected inter-film difference sequence can be obtained. .
[0091] S207. Based on the corrected inter-wafer difference sequence, perform bubble detection on the current wafer bonding process.
[0092] In some embodiments, step S207 includes: S2071. Determine the mean value of the inter-chip difference of the corrected inter-chip difference sub-sequence corresponding to several vacuum value change stages; wherein, the several corrected inter-chip difference sub-sequences are obtained by splitting the corrected inter-chip difference sequence through several vacuum value change stages.
[0093] S2072. The mean values of inter-chip differences corresponding to several vacuum value change stages are weighted and summed to obtain the weighted values of inter-chip differences.
[0094] S2073. Perform bubble detection based on the weighted value of the inter-piece difference.
[0095] In some embodiments, the calculation process of the mean inter-slice difference in step S2071 can be expressed by the following formula:
[0096] In the formula: Indicates the first The first wafer bonding process The average inter-chip difference corresponding to each vacuum value change stage Indicates the first The number of sampling time steps for each vacuum value change phase.
[0097] In some embodiments, the calculation process of the inter-chip difference weighted value in step S2072 is as follows:
[0098] In the formula: Indicates the first Weighted value of inter-wafer difference in the next wafer bonding process This indicates the number of stages in the vacuum value change. Indicates the first Weights for each vacuum value change stage.
[0099] In some embodiments, , , , The values are 0.2, 0.2, 0.3, and 0.3 respectively. Of course, the weight values here are only an example provided by this application. In actual implementation, different weight values can be set adaptively, and this application does not limit this.
[0100] In some embodiments, the process of performing bubble detection based on the inter-piece difference weighted value in step S2073 includes: Step 1: Calculate the weighted mean of the inter-slice differences. and standard deviation .
[0101] Step 2: Form warning lines and / or critical lines based on the mean and standard deviation of the weighted values of the inter-slice differences.
[0102] Step 3: Detect bubbles according to the warning line and / or critical line.
[0103] In some embodiments, the warning line in step 2 is The critical line is Of course, those skilled in the art can set different warning lines and critical lines as needed.
[0104] In step 3, when the weighted value of the inter-wafer difference exceeds the critical line, the system will issue a real-time alarm, and the factory will shut down for maintenance and polish the bottom chuck. When the weighted value of the inter-wafer difference exceeds the warning line, the bonding machine will not stop, and the subsequent measurement machine will be accelerated for analysis.
[0105] In this embodiment, in order to further suppress the influence of extreme noise points on the bubble detection results, such as preventing noise factors such as power fluctuations and vibration fluctuations in the equipment production process from causing instantaneous changes in the adsorption vacuum value and interfering with the bubble detection results, this embodiment further performs Wunder processing on the inter-slice difference sequence to further improve the accuracy of the bubble detection results.
[0106] Figure 9 A flowchart illustrating the method for detecting bubbles during wafer bonding according to the third embodiment of this application. See also... Figure 9 ,include: S301. Obtain the maintenance information of the bonding machine and determine whether the current wafer bonding process is one of the previous wafer bonding processes after maintenance. If so, directly set the bubble detection result of the current wafer bonding process to no bubbles. If not, proceed with the following steps.
[0107] In some embodiments, the first few wafer bonding processes after the bonding machine has been maintained can be assumed to be wafer bonding processes without the generation of bubbles. At this time, the bottom chuck in the bonding chamber is relatively clean because it has been maintained.
[0108] In some examples, the first few wafer bonding processes after maintenance can be the first, second, third, etc., after maintenance. The specifics depend on the bubble formation in historical wafer bonding processes after maintenance. If, historically, no bubbles were generated in the first 10 wafer bonding processes after maintenance, then these first 10 processes can be considered safe bonding processes, and the bubble detection results for these first 10 processes can be directly set to bubble-free.
[0109] Of course, this setup may also have certain risks. For example, if bubbles appear in the first 10 wafer bonding processes due to special circumstances, it may cause a large number of wafers to develop bubbles simultaneously during the bonding process. Therefore, to mitigate this risk, the first wafer bonding process after maintenance can be set to be bubble-free, while subsequent wafer bonding processes can be tested using the bubble detection method described in this application.
[0110] S302. Collect the adsorption vacuum value of the chuck during the previous wafer bonding process to form a historical vacuum value sequence; the previous wafer bonding process is a wafer bonding process without bubble generation.
[0111] See step S201.
[0112] S303. Collect the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence.
[0113] See step S202.
[0114] S304. Use local regression to determine the first slow drift sequence of the historical vacuum value sequence, and subtract the historical vacuum value sequence from the first slow drift sequence to obtain the corrected historical vacuum value sequence.
[0115] See step S203.
[0116] S305. Use local regression to determine the second slow drift sequence of the current vacuum value sequence, and subtract the current vacuum value sequence from the second slow drift sequence to obtain the corrected current vacuum value sequence.
[0117] See step S204.
[0118] S306. Subtract the current vacuum value sequence from the historical vacuum value sequence to obtain the inter-chip difference sequence.
[0119] See step S205.
[0120] S307. Perform Winsorization on the inter-slice difference sequence to obtain the corrected inter-slice difference sequence.
[0121] See step S206.
[0122] S308. Based on the corrected inter-wafer difference sequence, perform bubble detection on the current wafer bonding process.
[0123] See step S207.
[0124] See Figure 10 This diagram illustrates the adsorption vacuum value of the bottom chuck after maintenance, as provided in one embodiment of this application. It can be seen that after maintenance of the bonding machine, the adsorption vacuum value of the bottom chuck changes significantly compared to before maintenance. If the algorithm cannot identify that this fluctuation in adsorption vacuum value is caused by machine maintenance, it will falsely report the fluctuation, attributing it to air bubbles in the bottom chuck. This false reporting leads to a waste of manpower and bonding machine capacity. Therefore, to improve this problem, this embodiment further obtains machine maintenance information. After each maintenance and testing to ensure there are no foreign objects on the bottom chuck, the first wafer bonding process after machine maintenance confirms that there are no impurities on the bottom chuck.
[0125] See Figure 11 This diagram illustrates the bubble detection effect according to an embodiment of this application. When the detection method provided in this application is applied to a bonding machine, it can effectively detect the presence of bubbles during wafer bonding and provide an early warning. This bubble detection method can effectively prevent 98.5% of wafers from generating bubbles during bonding due to foreign objects in the bottom chuck.
[0126] Figure 12 This is a structural block diagram of a bubble detection device during wafer bonding process provided in one embodiment of this application. See also... Figure 12 The detection device 20 includes: The first acquisition module 21 is used to acquire the adsorption vacuum value of the chuck during the previous wafer bonding process and form a historical vacuum value sequence. The second acquisition module 22 is used to acquire the adsorption vacuum value of the chuck during the current wafer bonding process and form a current vacuum value sequence; Extraction module 23 is used to extract the vacuum value difference between the historical vacuum value sequence and the current vacuum value sequence to form an inter-chip difference sequence; The detection module 24 is used to perform bubble detection on the current wafer bonding process based on the inter-wafer difference sequence.
[0127] Figure 13 A structural block diagram of a bubble detection device during wafer bonding process provided in another embodiment of this application. See also Figure 13 The detection device 20 includes: The first acquisition module 21 is used to acquire the adsorption vacuum value of the chuck during the previous wafer bonding process and form a historical vacuum value sequence.
[0128] The second acquisition module 22 is used to acquire the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence.
[0129] Extraction module 23 is used to extract the vacuum value difference between the historical vacuum value sequence and the current vacuum value sequence to form an inter-chip difference sequence.
[0130] The Winsorization module 25 is used to perform Winsorization on the inter-slice difference sequence.
[0131] The detection module 24 is used to perform bubble detection on the current wafer bonding process based on the inter-wafer difference sequence.
[0132] Figure 14 This is a structural block diagram of an electronic device provided according to an embodiment of this application. See also... Figure 14 Electronic device 30 may include Figure 12 or Figure 13The device described above is for detecting bubbles during wafer bonding. Typically, the electronic device includes a processor 31 and a memory 32. The processor 31 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 31 may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 31 may also include a main processor and a coprocessor. The main processor is used to process data in the wake-up state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. The memory 32 may include one or more computer-readable storage media, which may be non-transitory. The memory 32 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage medium in memory 32 is used to store at least one instruction, which is executed by processor 31 to implement the method for detecting bubbles in the wafer bonding process performed by an electronic device, as provided in the method embodiments of this application.
[0133] Figure 15 This is a flowchart illustrating a bonding method according to an embodiment of this application. See also... Figure 15 Bonding methods include: S401. Perform preprocessing on the logic wafer and pixel wafer.
[0134] S402. Secure the logic wafer to the bottom chuck in the bonding chamber.
[0135] S403. Fix the pixel wafer on the top chuck of the bonding chamber and align the logic wafer and the pixel wafer.
[0136] S404. Pre-press bonding of logic wafers and pixel wafers.
[0137] S405, Anneal the logic wafer and pixel wafer.
[0138] S406. Obtain the adsorption vacuum value of the chuck during the previous wafer bonding process and form a historical vacuum value sequence.
[0139] See step S101.
[0140] S407. Based on the collected adsorption vacuum value during the current wafer bonding process, form a current vacuum value sequence.
[0141] See step S102.
[0142] S408. Use local regression to determine the first slow drift sequence of the historical vacuum value sequence, and subtract the historical vacuum value sequence from the first slow drift sequence to obtain the corrected historical vacuum value sequence.
[0143] See step S103.
[0144] S409. Use local regression to determine the second slow drift sequence of the current vacuum value sequence, and subtract the current vacuum value sequence from the second slow drift sequence to obtain the corrected current vacuum value sequence.
[0145] See step S104.
[0146] S410. Subtract the current vacuum value sequence from the historical vacuum value sequence to obtain the inter-chip difference sequence.
[0147] See step S105.
[0148] S411. Based on the inter-wafer difference sequence, perform bubble detection on the current wafer bonding process.
[0149] See step S106.
[0150] In this embodiment, the bubble detection method provided in this application is applied to the bonding process. After the current wafer bonding process, it can promptly provide feedback on whether there are bubbles in the current wafer bonding process, which can avoid the technical problem of multiple batches of wafers generating bubbles during the bonding process due to impurities in the bottom chuck.
[0151] In some embodiments, the pretreatment in step S401 includes: chemically mechanically polishing the bonding surfaces of the logic wafer and the pixel wafer, followed by plasma activation and cleaning of the bonding surfaces of the logic wafer and the pixel wafer.
[0152] In some embodiments, after chemical mechanical polishing, the roughness of the bonding surface is generally less than or equal to 0.4 nm. For example, the roughness is 0.2 nm.
[0153] In some embodiments, plasma activation can be achieved by bombarding the bonding surfaces with Ar plasma, causing the bonding surfaces to be covered with -OH hydroxyl groups, forming a hydrophilic surface. Subsequently, the logic wafers and pixel wafers are cleaned by megasonic cleaning to remove impurity particles.
[0154] In some embodiments, after the logic wafer is fixed on the bottom chuck in the bonding chamber in step S402, the adsorption vacuum value of the bottom chuck during the current wafer bonding process can be collected. Specifically, the adsorption vacuum value can be collected by a vacuum sensor.
[0155] In some embodiments, in step S404, a slight, uniform pressure is applied to the pixel wafer, causing the –OH hydroxyl groups on the bonding surfaces of the logic wafer and the pixel wafer to approach each other and undergo dehydration condensation to form Si–O–Si covalent bonds. The bonding surfaces of the logic wafer and the pixel wafer generally include a SiO2 bonding layer.
[0156] In some embodiments, in step S405, the logic wafer and pixel wafer are annealed at an annealing temperature of 200°C to 400°C. On the SiO2 bonding layer side, the Si–O–Si covalent bonds are further strengthened, and the bonding energy is >1.0 J / m. 2 .
[0157] In some embodiments, after annealing is completed in step S405, the current wafer bonding process is finished, and the collection of adsorption vacuum values during the current wafer bonding process can be terminated.
[0158] It should be noted that the technical solution provided in this application has unexpected technical effects, including: (1) This application provides a method for detecting bubbles during wafer bonding. The method involves acquiring the adsorption vacuum value of the chuck during a bubble-free wafer bonding process to form a historical vacuum value sequence. Then, the real-time adsorption vacuum value of the chuck during the current bonding process is used to form a current vacuum value sequence. By subtracting the current vacuum value sequence from the historical vacuum value sequence, an inter-wafer difference sequence is obtained. This inter-wafer difference sequence can reflect the difference in vacuum value between the current wafer bonding process and a bubble-free wafer bonding process. This vacuum value difference can be used to identify whether bubbles are generated during the current wafer bonding process.
[0159] (2) As a further technical improvement, the adsorption vacuum value acquisition process is divided into several vacuum value change stages, and the current vacuum value sequence and the historical vacuum value sequence are divided into current vacuum value subsequence and historical vacuum value subsequence corresponding to several vacuum value change stages. When the difference between the current vacuum value sequence and the historical vacuum value sequence is obtained to obtain the inter-segment difference subsequence, the difference between the current vacuum value subsequence and the historical vacuum value subsequence corresponding to the same vacuum value change stage is obtained respectively. The inter-segment difference subsequence can be obtained by combining several inter-segment difference subsequences in time sequence. The adsorption vacuum value of the chuck varies at different stages. For example, the adsorption vacuum value differs during the bottom wafer adsorption stage (the bottom chuck adsorbs logic wafers), the other chamber operation stages (a bonding machine typically includes several bonding chambers, and chamber operation affects the adsorption vacuum value of the bottom chuck in the current chamber), the top wafer adsorption stage (the top chuck adsorbs pixel wafers), and the final wafer bonding stage. Therefore, in this application, by dividing the vacuum value into subsequences according to the vacuum value change stages, a final inter-wafer difference subsequence is formed. The weighted sum of the average inter-wafer difference values from several subsequences is then used to obtain a weighted inter-wafer difference value. This can mitigate the impact of the inter-wafer difference value at a certain vacuum value change stage on the bubble detection results, ensuring the stability and accuracy of the bubble detection results.
[0160] (3) As a further technical improvement, local regression is adopted to eliminate the slow drift data caused by the environment / air pressure in the historical vacuum value sequence and the current vacuum value sequence, so as to avoid the influence of the environment / air pressure on the bubble detection results.
[0161] (4) As a further technical improvement, the inter-slice difference sequence is subjected to Wunder processing to suppress the influence of extreme noise points on the bubble detection results. For example, it prevents noise factors such as power fluctuations and vibration fluctuations during equipment production from causing instantaneous changes in the adsorption vacuum value and interfering with the bubble detection results.
[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for detecting bubbles during wafer bonding, characterized in that, include: Collect the adsorption vacuum value of the chuck during the previous wafer bonding process to form a historical vacuum value sequence; Collect the adsorption vacuum value of the chuck during the current wafer bonding process to form a current vacuum value sequence; The first slow drift sequence of the historical vacuum value sequence is determined by local regression. The difference between the historical vacuum value sequence and the first slow drift sequence is used to obtain the corrected historical vacuum value sequence. The second slow drift sequence of the current vacuum value sequence is determined by local regression, and the difference between the current vacuum value sequence and the second slow drift sequence is used to obtain the corrected current vacuum value sequence. The difference between the current vacuum value sequence and the historical vacuum value sequence is calculated to obtain the inter-piece difference sequence; Based on the inter-wafer difference sequence, bubble detection is performed on the current wafer bonding process.
2. The method for detecting bubbles during wafer bonding according to claim 1, characterized in that, The historical vacuum value sequence and the current vacuum value sequence are divided into several vacuum value change stages, which include the bottom wafer adsorption stage, the chamber actuation stage, the top wafer adsorption stage, and the bonding stage.
3. The method for detecting bubbles during wafer bonding according to claim 2, characterized in that, The process of detecting bubbles in the current wafer bonding process based on the inter-wafer difference sequence includes: The mean inter-chip difference of the inter-chip difference sub-sequence corresponding to several vacuum value change stages is determined respectively; wherein, the several inter-chip difference sub-sequences are obtained by splitting the inter-chip difference sequence through several vacuum value change stages; The weighted summation of the mean values of inter-chip differences corresponding to several vacuum value change stages yields the weighted value of inter-chip differences. Bubble detection is performed based on the weighted value of the inter-slice difference.
4. The method for detecting bubbles during wafer bonding according to claim 3, characterized in that, The process of bubble detection based on the weighted value of the inter-slice difference includes: Calculate the mean and standard deviation of the weighted values of the inter-slice differences; Early warning lines and / or critical lines are formed based on the mean and standard deviation of the weighted values of the inter-slice differences; Bubbles are detected based on the warning line and / or critical line.
5. The method for detecting bubbles during wafer bonding according to any one of claims 1 to 4, characterized in that, Before using inter-slice difference sequences for bubble detection, the following steps are also included: The inter-slice difference sequence is subjected to Winsorization.
6. The method for detecting bubbles during wafer bonding according to any one of claims 1 to 4, characterized in that, Also includes: Obtain maintenance information from the bonding machine and determine whether the current wafer bonding process is one of the previous wafer bonding processes after maintenance. If so, directly set the bubble detection result of the current wafer bonding process to bubble-free.
7. A device for detecting bubbles during wafer bonding, characterized in that, include: The first acquisition module is used to acquire the adsorption vacuum value of the chuck during the previous wafer bonding process and form a historical vacuum value sequence. The second acquisition module is used to acquire the adsorption vacuum value of the chuck during the current wafer bonding process and form the current vacuum value sequence; The extraction module is used to extract the vacuum value difference between the historical vacuum value sequence and the current vacuum value sequence to form an inter-chip difference sequence; The detection module is used to detect bubbles in the current wafer bonding process based on the inter-wafer difference sequence.
8. An electronic device, characterized in that, Includes the bubble detection device during wafer bonding as described in claim 7.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one piece of program code, which is executed by a processor to implement the method for detecting bubbles during wafer bonding as described in any one of claims 1 to 6.
10. A bonding method, characterized in that, include: Preprocessing is performed on logic wafers and pixel wafers; The logic wafer is secured to the bottom chuck in the bonding chamber; The pixel wafer is fixed on the top chuck of the bonding chamber, and the logic wafer and the pixel wafer are aligned. Pre-lamination bonding is performed on logic wafers and pixel wafers; Annealing is performed on logic wafers and pixel wafers; The adsorption vacuum value of the chuck during the previous wafer bonding process is acquired to form a historical vacuum value sequence; Based on the collected adsorption vacuum values during the current wafer bonding process, a current vacuum value sequence is formed; The first slow drift sequence of the historical vacuum value sequence is determined by local regression. The difference between the historical vacuum value sequence and the first slow drift sequence is used to obtain the corrected historical vacuum value sequence. The second slow drift sequence of the current vacuum value sequence is determined by local regression, and the difference between the current vacuum value sequence and the second slow drift sequence is used to obtain the corrected current vacuum value sequence. The difference between the current vacuum value sequence and the historical vacuum value sequence is calculated to obtain the inter-piece difference sequence; Based on the inter-wafer difference sequence, bubble detection is performed on the current wafer bonding process.