Wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration point array combined constraint

Abstract

The application provides a wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration point array combined constraint, comprising: setting a first calibration point array on an upper wafer to be bonded, and setting a second calibration point array on a lower wafer to be bonded; collecting a visible light image of the upper wafer and an infrared image of the lower wafer; establishing a coordinate correspondence relationship between a visible light image coordinate system and an infrared image coordinate system, and converting the coordinates of the visible light image and the infrared image to a unified coordinate system; establishing a calibration point correspondence relationship, and determining an alignment deviation parameter; generating a compensation control instruction to drive an alignment execution mechanism to perform compensation adjustment; returning to execute the above steps after each round of compensation adjustment until the alignment error is less than a preset threshold, so that the application adopts a combination of visible light detection and infrared detection to avoid the fuzzy superposition error caused by the fact that the upper and lower calibration points in the traditional scheme both depend on the infrared image.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing and packaging technology, and specifically to a closed-loop compensation method and system for wafer bonding alignment deviation based on visual-infrared fusion and calibration dot matrix joint constraints. Background Technology

[0002] Wafer bonding is a crucial process in semiconductor manufacturing and advanced packaging. It refers to the precise manufacturing technology that uses chemical and physical interactions to tightly bond two or more wafers of homogeneous or heterogeneous materials with clean, atomically polished surfaces. This bonding allows the interface atoms to form stable chemical bonds, metallic bonds, or dielectric interconnects, ultimately creating a single integrated structure from multiple wafers. Wafer bonding is widely used in 3D integrated circuits, hybrid bonding, MEMS hermetically sealed packaging, SOI wafer fabrication, heterogeneous integration, and wafer-level advanced packaging.

[0003] In hybrid bonding or wafer-level 3D integration, the metal pads, dielectric structures, via structures, or functional patterns on the upper and lower wafers need to be aligned with high precision before bonding. Existing technologies typically design alignment marks on each of the two wafers and utilize the characteristic that infrared light can see through silicon wafers to simultaneously observe the positions of the marks on the upper and lower wafers using near-infrared imaging, thereby calculating the relative deviation between the two wafers.

[0004] However, traditional transmission-type infrared wafer alignment methods suffer from problems such as limited infrared imaging resolution, easy overlap of infrared signals at upper and lower calibration points, potential local thermal deformation caused by infrared heating or response enhancement, superposition of registration errors in dual infrared blurred images, coordinate reference transmission errors in independent observation systems, and difficulty in further reducing steady-state errors in closed-loop compensation.

[0005] Furthermore, existing technologies employ top-view, bottom-view, or multi-camera independent observation structures to acquire images of different wafers or different markers, followed by coordinate conversion and motion compensation. However, independent observation structures are susceptible to differences in camera field of view, mechanical movement reference drift, optical axis inconsistency, and coordinate transfer errors. Without coaxial visual-infrared imaging and unified coordinate mapping constraints, it is difficult to stably solve for the precise alignment deviation of the upper and lower wafers under the same reference datum. Summary of the Invention

[0006] This invention is made to solve the above-mentioned problems, and aims to provide a closed-loop compensation method and system for wafer bonding alignment deviation based on visual-infrared fusion and calibration lattice joint constraints.

[0007] This invention provides a closed-loop compensation method for wafer bonding alignment deviation based on visual-infrared fusion and joint constraints of calibration lattice. The method includes the following steps: Step S1, setting a first calibration lattice on the upper wafer to be bonded and a second calibration lattice on the lower wafer to be bonded; Step S2, acquiring visible light images of the upper wafer and infrared images of the lower wafer from the same observation area via a coaxial optical path, wherein the visible light image is used to identify the first calibration lattice and the infrared image is used to identify the second calibration lattice; Step S3, establishing a coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transforming the coordinates of the visible light image and the infrared image to a unified coordinate system based on this correspondence. Coordinate system; Step S4: Under a unified coordinate system, establish the correspondence between the calibration points of the upper wafer and the lower wafer based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship, and geometric topology consistency relationship between the first calibration point matrix and the second calibration point matrix, and determine the alignment deviation parameters between the upper wafer and the lower wafer based on the calibration point correspondence; Step S5: Generate compensation control commands based on the alignment deviation parameters to drive the alignment execution mechanism to compensate and adjust the position or orientation of at least one wafer; Step S6: After each round of compensation and adjustment, return to steps S2 to S5, re-acquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold.

[0008] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following feature: wherein the material of the first calibration lattice is at least one of dielectric material, doped dielectric material, composite thin film material or surface modified material.

[0009] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following feature: wherein the material of the second calibration lattice is at least one of copper, aluminum, tungsten, titanium, nickel, gold, chromium, or a metal composite thereof.

[0010] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following features: wherein the coaxial optical path is realized by a beam splitter, a beam combiner, a dichroic mirror, a filter, an objective lens, and a coaxial illumination assembly.

[0011] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following feature: wherein the coordinate correspondence is realized by at least one of calibration plate calibration, magnification correction, distortion correction, coordinate lookup table correction or camera mounting position correction.

[0012] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints provided by the present invention may also have the following features: wherein, before step S4, the visible light image and the infrared image are preprocessed respectively, and the preprocessing includes at least one of grayscale conversion, filtering, threshold segmentation, local background subtraction, connected component extraction, edge extraction, grayscale centroid calculation or feature point center localization.

[0013] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints provided by the present invention may also have the following features: wherein the alignment deviation parameters may include horizontal translation, vertical translation, in-plane rotation and attitude correction.

[0014] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following feature: wherein the preset threshold may be the calibration point average alignment error threshold, the calibration point root mean square alignment error threshold, the horizontal translation error threshold, the vertical translation error threshold, the in-plane rotation error threshold, or a combination thereof.

[0015] The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraint provided by the present invention may also have the following feature: wherein the alignment actuator includes at least one of a translation stage, a rotary stage, a piezoelectric displacement stage, a linear motor platform, a vacuum adsorption platform, or a multi-degree-of-freedom motion platform.

[0016] This invention also provides a closed-loop compensation system for wafer bonding alignment deviation based on visual-infrared fusion and joint constraints of calibration dot matrix, characterized by: a calibration dot matrix module for setting a first calibration dot matrix on the upper wafer to be bonded and a second calibration dot matrix on the lower wafer to be bonded; an image acquisition module for acquiring visible light images of the upper wafer and infrared images of the lower wafer respectively through a coaxial optical path on the same wafer observation area, wherein the visible light image is used to identify the first calibration dot matrix and the infrared image is used to identify the second calibration dot matrix; and a coordinate calibration module for establishing a coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transforming the coordinates of the visible light image and the infrared image to [the appropriate coordinate system]. A unified coordinate system; a deviation determination module, used to establish the correspondence between calibration points of the upper and lower wafers under the unified coordinate system based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship, and geometric topology consistency relationship between the first and second calibration point matrices, and to determine the alignment deviation parameters between the upper and lower wafers based on the calibration point correspondence; a compensation adjustment module, used to generate compensation control commands based on the alignment deviation parameters, driving the alignment actuator to compensate and adjust the position or orientation of at least one wafer; a closed-loop compensation control module, used to repeat the above modules after each round of compensation adjustment, re-acquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold.

[0017] The role and effect of invention

[0018] The wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration lattice joint constraints according to the present invention includes: Step S1, setting a first calibration lattice on the upper wafer to be bonded and setting a second calibration lattice on the lower wafer to be bonded; Step S2, acquiring visible light images of the upper wafer and infrared images of the lower wafer respectively through a coaxial optical path to the same wafer observation area, wherein the visible light image is used to identify the first calibration lattice and the infrared image is used to identify the second calibration lattice; Step S3, establishing a coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transforming the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence; Step S4, under the unified coordinate system, establishing the upper wafer alignment deviation closed-loop compensation method and system based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topological consistency relationship between the first calibration lattice and the second calibration lattice. The alignment deviation parameters between the upper and lower wafers are determined based on the correspondence between the calibration points of the upper and lower wafers. Step S5: A compensation control command is generated based on the alignment deviation parameters to drive the alignment execution mechanism to compensate and adjust the position or orientation of at least one wafer. Step S6: After each round of compensation and adjustment, the process returns to steps S2 to S5 to reacquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold. Therefore, the wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration point array joint constraints of the present invention adopts a combination of visible light detection of the upper wafer and infrared detection of the lower wafer, avoiding the blurring and superposition error caused by the reliance of both upper and lower calibration points on infrared images in the traditional dual infrared scheme. Furthermore, the sensing channel is separated by the upper thermal conductivity matching visual imaging calibration point array and the lower infrared enhancement calibration point array, reducing the effects of infrared signal aliasing and thermal deformation. Attached Figure Description

[0019] Figure 1 This is a schematic flowchart of a wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the calibration dot matrix in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of the wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of the imaging of the first calibration array of the upper wafer and the second calibration array of the lower wafer in an embodiment of the present invention.

[0023] Figure 5This is a schematic diagram of infrared imaging of the second calibration array of the lower wafer after closed-loop compensation in an embodiment of the present invention.

[0024] Figure 6 This is a schematic diagram of the convergence curve of the alignment error as a function of the number of closed-loop iterations in the simulation experiment of an embodiment of the present invention.

[0025] Figure 7 This is a schematic diagram of a closed-loop compensation system for wafer bonding alignment deviation based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention. Detailed Implementation

[0026] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the following embodiments, in conjunction with the accompanying drawings, specifically illustrate the wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration dot matrix joint constraints of the present invention.

[0027] Example

[0028] Figure 1 This is a schematic flowchart of a wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention.

[0029] like Figure 1 As shown, this embodiment provides a closed-loop compensation method for wafer bonding alignment deviation based on visual-infrared fusion and calibration lattice joint constraints, including:

[0030] Step S1: Set a first calibration dot matrix on the upper wafer to be bonded, and set a second calibration dot matrix on the lower wafer to be bonded.

[0031] Figure 2 This is a schematic diagram of the calibration dot matrix in an embodiment of the present invention.

[0032] like Figure 2 As shown, the material of the first calibration lattice is at least one of a dielectric material, a doped dielectric material, a composite thin film material, or a surface-modified material. The first calibration lattice is formed of a material that is recognizable by visible light imaging and has a thermal conductivity close to that of the upper wafer substrate material, so that it has high contrast in visible light images, while reducing local thermal stress and micro-deformation caused by infrared illumination, infrared heating, or temperature gradients.

[0033] The material of the second calibration dot array is at least one of copper, aluminum, tungsten, titanium, nickel, gold, chromium, or a composite of their metals. The second calibration dot array is used to enhance infrared absorption, infrared reflection, or infrared thermal response, enabling an infrared camera to identify the calibration points of the lower wafer through the upper wafer.

[0034] Step S2 involves acquiring visible light images of the upper wafer and infrared images of the lower wafer from the same wafer observation area via a coaxial optical path. The visible light image is used to identify the first calibration matrix, and the infrared image is used to identify the second calibration matrix. The coaxial optical path ensures that the visible light image imaging unit and the infrared image imaging unit correspond to the same or approximately the same wafer observation area, achieved through a beam splitter, beam combiner, dichroic mirror, filter, objective lens, and coaxial illumination assembly.

[0035] Step S3: Establish the coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transform the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence. The coordinate correspondence is achieved through at least one of the following methods: calibration plate calibration, magnification correction, distortion correction, coordinate lookup table correction, or camera mounting position correction. Transforming the image coordinates of the first and second calibration point matrices to a unified coordinate system allows the upper wafer calibration points and the lower wafer calibration points to perform alignment deviation analysis under the same reference datum.

[0036] In this embodiment, the visible light image and the infrared image can also be preprocessed separately. The preprocessing includes at least one of grayscale conversion, filtering, threshold segmentation, local background subtraction, connected component extraction, edge extraction, grayscale centroid calculation, or feature point center localization.

[0037] Step S4: Under a unified coordinate system, establish the correspondence between the calibration points of the upper wafer and the lower wafer based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topology consistency relationship between the first calibration point matrix and the second calibration point matrix, and determine the alignment deviation parameters between the upper wafer and the lower wafer based on the correspondence between the calibration points.

[0038] The alignment deviation parameters may include horizontal translation, vertical translation, in-plane rotation, and attitude correction. The alignment deviation parameters are determined by the positional deviations of multiple corresponding calibration points to reduce the impact of defects in individual calibration points, infrared image noise, local blurring, or local recognition errors on the compensation results.

[0039] In determining alignment deviations, the positional correspondence between calibration points, the geometric relationship of the calibration point array itself, the coordinate unification relationship between the visible light image and the infrared image, and the motion continuity during the closed-loop compensation process can all be used as constraints. Specifically, the positional correspondence between calibration points is used to reflect the relative positional deviation between the upper and lower wafers. The geometric relationship of the calibration point array itself is used to maintain the relative distance, angle, ring distribution, array distribution, adjacency relationship, or numbering order between calibration points. The coordinate unification relationship between the visible light image and the infrared image is used to ensure that the visible light image and the infrared image are fused under the same reference. The motion continuity during the closed-loop compensation process is used to suppress overcompensation or compensation oscillations caused by single-image noise, local mismatch, or actuator response errors.

[0040] The relative alignment deviation between upper and lower wafers can be described using a combination of translation, rotation, and attitude correction. Horizontal and vertical translations are used to compensate for positional deviations between the two wafers in the plane. In-plane rotation compensates for angular deviations around the normal. Attitude corrections compensate for deviations caused by wafer micro-tilts, optical axis angles, or projection differences. These methods of describing alignment deviations are only used to characterize the relative position and attitude relationship between wafers and do not limit the specific mathematical model or solution algorithm.

[0041] Step S5: Generate compensation control commands based on alignment deviation parameters to drive the alignment actuator to compensate and adjust the position or orientation of at least one wafer. The compensation control commands may include translation compensation commands, rotation compensation commands, orientation adjustment commands, or combinations thereof.

[0042] The alignment actuator includes at least one of a translation stage, a rotary stage, a piezoelectric displacement stage, a linear motor platform, a vacuum adsorption platform, or a multi-degree-of-freedom motion platform.

[0043] Step S6: After each round of compensation adjustment, return to steps S2 to S5 to reacquire images and redetermine the alignment deviation parameters until the alignment error is less than the preset threshold.

[0044] The preset threshold can be the average alignment error threshold of the calibration point, the root mean square alignment error threshold of the calibration point, the horizontal translation error threshold, the vertical translation error threshold, the in-plane rotation error threshold, or a combination thereof.

[0045] Figure 3 This is a schematic diagram of the wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention.

[0046] like Figure 3As shown, a visible light illumination source emits visible light, which is captured by a visible light camera to form the first calibration dot matrix image (visible light image) on the wafer surface. An infrared illumination source emits infrared light, which is captured by an infrared camera to capture the infrared transmission image inside the wafer.

[0047] The beam splitter / combiner merges visible and infrared light paths, allowing both beams to propagate along the same path. The combined beam then passes vertically downwards through the objective lens. The objective lens focuses the light onto the target below. The upper and lower wafers, located at the bottom, are the targets focused by the objective lens, and their surface or internal features are simultaneously captured.

[0048] Figure 4 This is a schematic diagram of the simulation experiment results in an embodiment of the present invention.

[0049] like Figure 4 As shown, the simulation experiment in this embodiment uses a two-dimensional image with a size of 600 pixels × 600 pixels, with the wafer center located at the center of the image. The outer contour radius parameter of the wafer is set to 240 pixels, and the simulation mainly renders a ring-shaped calibration point array. Both the upper and lower wafers are equipped with ring-shaped calibration point arrays, with a total of 8 calibration points located on a circle with a radius of 210 pixels. Each calibration point is simulated using a two-dimensional Gaussian bright spot, and the Gaussian standard deviation of the calibration point bright spot is set to 2.5 pixels to simulate the finite size and grayscale distribution of the actual calibration point in the imaging system.

[0050] In the simulation experiment, a known alignment deviation is pre-applied to the lower wafer relative to the upper wafer to simulate the initial misalignment between the upper and lower wafers before bonding. The known alignment deviation includes a horizontal translation deviation. Vertical translation deviation In-plane rotational deviation and attitude deviation coefficients caused by slight tilt or projection differences Specifically, the actual horizontal translation is set to 14.0 pixels, the actual vertical translation is set to 10.0 pixels, the actual in-plane rotation angle is set to 2.5°, and the attitude deviation coefficient β is set to 3.0×10^-4.

[0051] Attitude deviation coefficient This is used to characterize perspective deviation caused by a slight tilt of the lower wafer relative to the imaging optical axis, or by projection differences between the upper and lower wafers. In this simulation experiment, the attitude deviation coefficient... As a coefficient of the homogeneous coordinate term in perspective transformation, it participates in the generation of the lower-level wafer calibration point position. Its unit can be understood as one per pixel. It is used to simulate the effect of the wafer orientation not being completely parallel on the position of the calibration point image. The larger the value, the more obvious the perspective deviation caused by slight tilt or projection differences.

[0052] The offset image of the lower wafer is generated using a transformation matrix that includes translation, rotation, and perspective distortion. Let the image center translation matrix be... The perspective deviation matrix is The in-plane rotation matrix is The shift and translation matrices are: Then the image transformation matrix of the lower wafer relative to the upper wafer can be expressed as:

[0053]

[0054] in, Used to translate the origin of the image coordinate system to the center of the image. Used to simulate the attitude deviation coefficient The resulting perspective deviation Used to simulate in-plane rotation of a wafer. This matrix is ​​used to move the image coordinate origin back and superimpose the horizontal and vertical translation amounts. Using this matrix, the calibration point position and image state of the lower wafer can be generated under the combined effects of translation, rotation, and slight attitude deviations.

[0055] The perspective deviation matrix is The rotation matrix within the surface is They are represented as follows:

[0056]

[0057]

[0058] in, This is the in-plane rotation angle of the lower wafer relative to the upper wafer. This represents the attitude deviation coefficient. For any calibration point on the lower wafer, its offset image position can be obtained through the above transformation matrix. This matrix setting is only used to construct a controllable wafer misalignment state in simulation experiments and does not limit the specific deviation calculation method in practical applications of this invention.

[0059] The simulation experiment sets up two comparison methods. Traditional method A uses a transmission-type dual infrared alignment method, where both the upper and lower wafer calibration point arrays are acquired through infrared images. Method B uses the visual-infrared fusion method of this invention, where the first calibration point array of the upper wafer is acquired through a visible light image, and the second calibration point array of the lower wafer is acquired through an infrared image.

[0060] To simulate the imaging quality of different imaging channels, larger blur and noise parameters were used for infrared images, while smaller blur and noise parameters were used for visible light images. Specifically, the standard deviation of Gaussian blur for infrared images was set to 3.5 pixels, and the standard deviation of noise was set to 0.08; the standard deviation of Gaussian blur for visible light images was set to 0.4 pixels, and the standard deviation of noise was set to 0.008. These parameters characterize the cases where infrared imaging has lower resolution and higher noise, and the cases where visible light imaging has higher resolution and lower noise.

[0061] To ensure the comparability of the comparison results, both methods use the same wafer size, the same calibration dot array arrangement, the same true initial deviation, the same lower-layer infrared imaging conditions, and the same closed-loop compensation strategy. The difference lies in that, in traditional method A, the upper wafer calibration dot array also uses infrared image recognition, meaning that both the upper and lower layer calibration points are affected by infrared blurring and infrared noise; in method B of this invention, the upper wafer first calibration dot array uses visible light image recognition, and the lower wafer second calibration dot array uses infrared image recognition.

[0062] During the simulation experiment, the center of the calibration point in each image was extracted. Specifically, pre-smoothing, threshold segmentation, morphological opening operation, connected component filtering, local background subtraction, and gray-scale centroid calculation were used to obtain the center coordinates of the calibration point. The threshold was set to 0.15, the standard deviation of pre-smoothing was set to 0.8 pixels, the morphological opening operation used a 3×3 structuring element, the connected component area filtering range was set to 8 to 2000 pixels, and the local centroid calculation window radius was set to 12 pixels. Through these processes, the sub-pixel-level center localization of the calibration point in actual image processing can be simulated.

[0063] After obtaining the center coordinates of the calibration points on the upper and lower wafers, a correspondence is established using nearest neighbor matching or global minimum distance allocation. The current alignment deviation is then estimated based on the positional relationship between the corresponding calibration points. Closed-loop compensation is performed using a step-by-step compensation method. Each round of compensation generates a compensation amount based on the current estimated deviation and drives the alignment actuator to adjust according to the set compensation gain. In this simulation experiment, the compensation gain is set to 0.4, and the number of closed-loop iterations is set to 30.

[0064] To reduce the impact of random noise on the results of a single experiment, 20 independent random noise experiments were conducted on both traditional method A and the method of this invention B, and the average value of each experimental result was taken. After each iteration, the root mean square alignment error was calculated as an evaluation index based on the actual position of the corresponding calibration point in the simulation. The root mean square alignment error is used to characterize the overall alignment degree of the upper and lower wafer calibration point lattices in a unified coordinate system; the smaller the error, the better the alignment effect.

[0065] Figure 4This is a schematic diagram of the imaging of the first calibration array of the upper wafer and the second calibration array of the lower wafer in an embodiment of the present invention.

[0066] Figure 5 This is a schematic diagram of infrared imaging of the second calibration array of the lower wafer after closed-loop compensation in an embodiment of the present invention.

[0067] Figure 6 This is a schematic diagram of the convergence curve of the alignment error as a function of the number of closed-loop iterations in the simulation experiment of an embodiment of the present invention.

[0068] like Figure 4 , Figure 5 and Figure 6 As shown, both conventional method A and the present invention's method B can reduce alignment errors between upper and lower wafers through closed-loop compensation, but the present invention's method B performs better in terms of steady-state error. This is because, in conventional method A, both upper and lower calibration points rely on infrared image recognition, and both calibration point arrays are affected by infrared imaging blurring and noise; while in the present invention's method B, the first calibration point array on the upper wafer is recognized by visible light image recognition, whose blurring and noise levels are significantly lower than those of infrared images, thereby reducing the center positioning error of the upper calibration points and minimizing the impact of overlapping infrared signals between upper and lower calibration points on feature matching.

[0069] Under the simulation conditions described above, the root mean square alignment error of the last 10 iterations was statistically analyzed, and the average value was taken from 20 independent random noise experiments. The results show that the steady-state calibration point root mean square alignment error of the traditional dual-infrared method A is approximately 1.46 times that of the visual-infrared fusion method B of this invention. This result indicates that, under the same lower-layer infrared detection conditions, the same initial deviation, and the same closed-loop compensation strategy, replacing the upper-layer wafer measurement channel with a visible light image from the infrared image can effectively reduce the residual alignment error after closed-loop compensation and improve the wafer bonding alignment compensation accuracy.

[0070] As demonstrated by the simulation experiments above, this invention combines upper-layer visible light detection with lower-layer infrared detection, allowing the upper and lower calibration points to be acquired by different imaging channels. This reduces the errors caused by blurring and signal aliasing in traditional dual-infrared alignment. Furthermore, by incorporating the geometric correspondence of the calibration points and a closed-loop compensation process, this invention gradually converges the alignment error to below a preset threshold. It is suitable for alignment deviation compensation in wafer-to-wafer bonding, chip-to-wafer bonding, hybrid bonding, metal bonding, dielectric bonding, 3D integrated bonding, MEMS packaging bonding, or SOI wafer fabrication.

[0071] Figure 7 This is a schematic diagram of a closed-loop compensation system for wafer bonding alignment deviation based on visual-infrared fusion and calibration dot matrix joint constraints in an embodiment of the present invention.

[0072] like Figure 5As shown, this embodiment also provides a wafer bonding alignment deviation closed-loop compensation system 100 based on visual-infrared fusion and calibration dot matrix joint constraints, including: calibration dot matrix module 1, image acquisition module 2, coordinate calibration module 3, deviation determination module 4, compensation adjustment module 5, and closed-loop compensation control module 6.

[0073] The calibration dot matrix module 1 uses the above step S1 to set a first calibration dot matrix on the upper wafer to be bonded and a second calibration dot matrix on the lower wafer to be bonded.

[0074] The image acquisition module 2 uses the above step S2 to acquire visible light images of the upper wafer and infrared images of the lower wafer respectively through the coaxial optical path to observe the same wafer observation area. The visible light image is used to identify the first calibration matrix, and the infrared image is used to identify the second calibration matrix.

[0075] The coordinate calibration module 3 uses the above step S3 to establish the coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transforms the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence.

[0076] The deviation determination module 4 adopts the above step S4 to establish the calibration point correspondence between the upper wafer and the lower wafer under a unified coordinate system based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topology consistency relationship between the first calibration point matrix and the second calibration point matrix, and determines the alignment deviation parameters between the upper wafer and the lower wafer based on the calibration point correspondence.

[0077] The compensation adjustment module 5 uses the above step S5 to generate compensation control commands based on the alignment deviation parameters, and drives the alignment actuator to compensate and adjust the position or orientation of at least one wafer.

[0078] The closed-loop compensation control module 6 uses the above step S6 to repeat the above module after each round of compensation adjustment, re-acquire images and redetermine the alignment deviation parameters until the alignment error is less than the preset threshold.

[0079] The role and effect of the embodiments

[0080] According to the wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration dot matrix joint constraints involved in this embodiment, the method includes: Step S1, setting a first calibration dot matrix on the upper wafer to be bonded and setting a second calibration dot matrix on the lower wafer to be bonded; Step S2, acquiring visible light images of the upper wafer and infrared images of the lower wafer respectively through a coaxial optical path to the same wafer observation area, wherein the visible light image is used to identify the first calibration dot matrix and the infrared image is used to identify the second calibration dot matrix; Step S3, establishing a coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transforming the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence; Step S4, under the unified coordinate system, establishing a closed-loop compensation method for the upper wafer according to the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topology consistency relationship between the first calibration dot matrix and the second calibration dot matrix. The alignment deviation parameters between the upper and lower wafers are determined based on the correspondence between the calibration points of the upper and lower wafers. Step S5: A compensation control command is generated based on the alignment deviation parameters to drive the alignment execution mechanism to compensate and adjust the position or orientation of at least one wafer. Step S6: After each round of compensation and adjustment, the process returns to steps S2 to S5 to reacquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold. Therefore, the wafer bonding alignment deviation closed-loop compensation method and system based on visual-infrared fusion and calibration point array joint constraints of the present invention adopts a combination of visible light detection of the upper wafer and infrared detection of the lower wafer, avoiding the blurring and superposition error caused by the reliance of both upper and lower calibration points on infrared images in the traditional dual infrared scheme. Furthermore, the sensing channel is separated by the upper thermal conductivity matching visual imaging calibration point array and the lower infrared enhancement calibration point array, reducing the effects of infrared signal aliasing and thermal deformation.

[0081] This embodiment also uses a coaxial optical path and coordinate calibration method to place the two types of images under a unified reference standard, thereby reducing the coordinate error introduced by the difference in different imaging fields of view.

[0082] This embodiment also improves the stability and noise resistance of the alignment results by combining the positional relationship of calibration points, geometric topological relationship, unified relationship of dual-modal coordinates and closed-loop compensation state to determine the alignment deviation.

[0083] This embodiment can also be used in conjunction with translation stages, rotary stages, or multi-degree-of-freedom motion platforms in existing wafer bonding equipment.

[0084] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A closed-loop compensation method for wafer bonding alignment deviation based on visual-infrared fusion and calibration lattice joint constraints, characterized in that, include: Step S1: Set a first calibration matrix on the upper wafer to be bonded, and set a second calibration matrix on the lower wafer to be bonded. Step S2: The visible light image of the upper wafer and the infrared image of the lower wafer are acquired respectively through the coaxial optical path to the same wafer observation area. The visible light image is used to identify the first calibration matrix, and the infrared image is used to identify the second calibration matrix. Step S3: Establish the coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and transform the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence. Step S4: Under a unified coordinate system, based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topology consistency relationship between the first calibration point matrix and the second calibration point matrix, establish the calibration point correspondence relationship between the upper wafer and the lower wafer, and determine the alignment deviation parameter between the upper wafer and the lower wafer based on the calibration point correspondence relationship. Step S5: Generate a compensation control command based on the alignment deviation parameter to drive the alignment actuator to compensate and adjust the position or orientation of at least one wafer. Step S6: After each round of compensation adjustment, return to steps S2 to S5 to reacquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold.

2. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints as described in claim 1, characterized in that: in, The material of the first calibration lattice is at least one of dielectric material, doped dielectric material, composite thin film material, or surface modified material.

3. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints as described in claim 1, characterized in that: in, The material of the second calibration lattice is at least one of copper, aluminum, tungsten, titanium, nickel, gold, chromium, or a metal composite thereof.

4. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints as described in claim 1, characterized in that: in, The coaxial optical path is realized through a beam splitter, a beam combiner, a dichroic mirror, a filter, an objective lens, and a coaxial illumination assembly.

5. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints according to claim 1, characterized in that: in, The coordinate correspondence is achieved through at least one of the following methods: calibration plate calibration, magnification correction, distortion correction, coordinate lookup table correction, or camera installation position correction.

6. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints according to claim 1, characterized in that: in, Before step S4, the visible light image and the infrared image are preprocessed respectively. The preprocessing includes at least one of grayscale conversion, filtering, threshold segmentation, local background subtraction, connected component extraction, edge extraction, grayscale centroid calculation or feature point center localization.

7. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints according to claim 1, characterized in that: in, The alignment deviation parameters may include horizontal translation, vertical translation, in-plane rotation, and attitude correction.

8. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints as described in claim 1, characterized in that: in, The preset threshold may be the average alignment error threshold of the calibration point, the root mean square alignment error threshold of the calibration point, the horizontal translation error threshold, the vertical translation error threshold, the in-plane rotation error threshold, or a combination thereof.

9. The wafer bonding alignment deviation closed-loop compensation method based on visual-infrared fusion and calibration lattice joint constraints according to claim 1, characterized in that: in, The alignment actuator includes at least one of a translation stage, a rotary stage, a piezoelectric displacement stage, a linear motor platform, a vacuum adsorption platform, or a multi-degree-of-freedom motion platform.

10. A closed-loop compensation system for wafer bonding alignment deviation based on visual-infrared fusion and calibration lattice joint constraints, characterized in that, include: The calibration dot matrix module is used to set a first calibration dot matrix on the upper wafer to be bonded and a second calibration dot matrix on the lower wafer to be bonded. The image acquisition module is used to acquire visible light images of the upper wafer and infrared images of the lower wafer respectively from the same wafer observation area via a coaxial optical path. The visible light images are used to identify the first calibration matrix, and the infrared images are used to identify the second calibration matrix. The coordinate calibration module is used to establish the coordinate correspondence between the visible light image coordinate system and the infrared image coordinate system, and to transform the coordinates of the visible light image and the infrared image to a unified coordinate system based on the coordinate correspondence. The deviation determination module is used to establish the calibration point correspondence between the upper wafer and the lower wafer in a unified coordinate system based on the positional relationship, design number relationship, adjacency relationship, array arrangement relationship, ring arrangement relationship and geometric topology consistency relationship between the first calibration point matrix and the second calibration point matrix, and to determine the alignment deviation parameters between the upper wafer and the lower wafer based on the calibration point correspondence. The compensation adjustment module is used to generate compensation control commands based on the alignment deviation parameters, and drive the alignment actuator to compensate and adjust the position or orientation of at least one wafer. The closed-loop compensation control module is used to repeatedly execute the above module after each round of compensation adjustment, re-acquire images and redetermine the alignment deviation parameters until the alignment error is less than a preset threshold.

Images