A method for detecting and measuring chipping in a wafer dicing process

Abstract

The application discloses a kind of detection measurement methods of edge collapse in wafer cutting process, specifically related to edge collapse detection technical field, according to cutting load, material elasticity and thickness distribution to construct the deformation sensitive description quantity of cutting position, and according to this to the edge profile of cutting path, deformation component analysis and reference boundary compensation correction are carried out, by only to the local defect area that still exceeds real geometric boundary after compensation is carried out edge crack determination, and further calculate the edge collapse depth, extension length and effective edge crack area after rejecting the influence of elastic deformation, the authenticity and stability of edge collapse physical measurement result are significantly improved;Consistency check is carried out in combination with deformation sensitive description quantity, to avoid misjudgment to edge collapse grade in high deformation sensitive area, so as to realize the high consistency correlation of edge collapse detection result and cutting position and process parameters.

Description

Technical Field

[0001] This invention relates to the field of edge chipping detection technology, and more specifically, to a method for detecting and measuring edge chipping during wafer dicing. Background Technology

[0002] During ultrathin wafer dicing, as the wafer thickness decreases to 50–100 μm or even thinner, its overall bending stiffness drops significantly. Under the influence of cutting force and vibration energy, the wafer is highly susceptible to overall elastic flexural deformation. This deformation couples with the brittle fracture process induced by localized cutting, meaning the defect morphology at the diced edge is no longer solely determined by localized material fracture, but also by the geometric projection offset effect caused by transient wafer bending. In this situation, the diced edge profile obtained through imaging often contains significant deformation components, making it difficult to accurately reflect the actual geometric boundary of the wafer under no external force. This leads to systematic deviations in subsequent edge chipping identification and measurement results. Especially when measuring key parameters such as chipping depth, the detection results inevitably include spurious defect components caused by wafer elastic deformation, amplifying or misjudging the apparent degree of chipping. Existing technologies typically suppress wafer deformation by increasing clamping force, adding support, or introducing external correction methods. However, these methods themselves introduce new additional loads or residual stresses into the ultrathin wafer, altering the actual fracture state and making accurate measurement difficult without interfering with the physical process. Therefore, in ultrathin wafer dicing scenarios, effectively distinguishing between defects caused by genuine material fracture and pseudo-chipping caused by overall elastic deformation without introducing additional stress disturbances has become a key technical problem restricting the accuracy of wafer dicing chipping detection and measurement. Summary of the Invention

[0003] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method for detecting and measuring edge breakage during wafer dicing, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting and measuring edge chipping during wafer dicing includes the following steps: Obtain the wafer thickness distribution, cutting load history parameters, and material elastic parameters in the vicinity of the wafer dicing track; calculate the equivalent bending stiffness of the dicing area; and obtain the deformation-sensitive descriptor of the corresponding dicing position. Based on the deformation-sensitive descriptor, the deformation component of the imaging profile of the cutting path edge is analyzed, the mapping relationship between the edge profile normal offset and the overall elastic deflection response is established, the deformation compensation correction of the cutting path reference boundary is performed, and the true geometric boundary is obtained. Align and compare the real geometric boundary with the actual cutting edge contour, and extract the local defect area that still exceeds the real geometric boundary after deformation compensation as the real fracture candidate area. For the actual fracture candidate region, calculate the fracture depth, fracture extension length and effective fracture area after removing the influence of elastic deformation, and form a set of physical measurement parameters for the fracture edge; The consistency of the physical measurement parameter set of edge chipping with the deformation-sensitive descriptive quantity of the corresponding cutting position is verified, an edge chipping level evaluation result is generated, and the edge chipping detection measurement result associated with the cutting position and cutting process parameters is output.

[0005] In a preferred embodiment, the process of obtaining the wafer thickness distribution, historical cutting load parameters, and material elastic parameters in the vicinity of the wafer dicing track, calculating the equivalent bending stiffness of the dicing region, and obtaining the deformation-sensitive descriptor for the corresponding dicing position is as follows: In wafer dicing equipment, acquire thickness distribution data, historical cutting load parameters, and wafer material elastic parameters in the vicinity of the wafer dicing track. The thickness distribution data includes local thickness values ​​along the cutting track normal and the cutting direction; The historical parameters of the cutting load include the blade rotation speed, feed rate, cutting depth and corresponding cutting force; The material's elastic parameters include the material's elastic modulus and Poisson's ratio; The neighborhood of the cutting path is divided according to a preset spatial window length, resulting in multiple cutting position analysis windows. ; Analysis window for each cutting position The average thickness within the window is calculated based on the collected thickness distribution data. Calculate the equivalent bending stiffness of the cut region based on the average thickness, material elastic modulus, and Poisson's ratio. Based on the historical cutting load parameters, extract the cutting load corresponding to the cutting position analysis window i; The cutting load is correlated with the equivalent bending stiffness to calculate the deformation response of the cutting position analysis window i. Based on the deformation response, it is normalized to obtain the deformation sensitivity description of the cutting position analysis window i.

[0006] In a preferred embodiment, the process of analyzing the deformation components of the imaging profile of the cutting edge based on the deformation-sensitive descriptor, establishing the mapping relationship between the edge profile normal offset and the overall elastic deflection response, and performing deformation compensation correction on the cutting track reference boundary to obtain the true geometric boundary is as follows: Obtain the original imaging contour data of the cutting path edge, preprocess the original imaging contour data, and extract the initial contour point set of the cutting path edge; Based on the initial set of contour points, calculate the local tangent direction at each contour point along the edge of the cutting path, and determine the corresponding normal direction vector. Associate each contour point with its corresponding cutting position analysis window i, obtain the deformation-sensitive descriptive quantity corresponding to the window, and establish the correlation between the contour point normal displacement and the deformation-sensitive descriptive quantity. Based on the deformation-sensitive descriptor, a projection model of the overall elastic flexural response on the edge contour is constructed to obtain the normal deformation projection of the contour points. By introducing a correction term related to the cutting direction, the normal deformation projection is corrected by directional weighting to obtain the corrected normal deformation projection. Based on the normal deformation projection amount, the original contour point position is reverse-compensated along the normal direction of the contour point to obtain the deformed contour point. All deformation-compensated contour points are reconstructed according to the cutting path direction to form the deformation-compensated boundary curve of the cutting path, which serves as the true geometric boundary under conditions without external force interference.

[0007] In a preferred embodiment, the process of aligning and comparing the true geometric boundary with the actual cutting edge contour, and extracting the local defect areas that still exceed the true geometric boundary after deformation compensation as candidate areas for true fracture is as follows: Obtain the true geometric boundary curve of the cutting path and the actual cutting edge contour curve of the corresponding cutting path after deformation compensation; The true geometric boundary curve and the actual cutting edge contour curve are spatially aligned. After spatial alignment is completed, the real geometric boundary curve is parametrically represented along the cutting path direction to obtain a set of reference points arranged in the order of the cutting path. At the same time, the actual cutting edge contour curve is sampled at the same scale to obtain an actual contour point set that corresponds one-to-one with the reference point set. For each pair of corresponding reference points and actual contour points, calculate the normal offset distance of the actual contour points relative to the true geometric boundary. ; According to the normal offset distance The sign and numerical value are used to determine the defects of actual contour points: When the normal offset distance When the value is less than zero, it indicates that the actual contour point is located inside the true geometric boundary and belongs to the potential defect point that invades the interior of the wafer; When the normal offset distance When the value is greater than or equal to zero, it indicates that the contour points have not formed a valid defect; A minimum defect threshold is introduced among the potential defect points. Perform the screening; When satisfied When the actual contour point is determined to be the real defect point; Multiple real defect points that are continuously distributed in space are clustered according to the cutting direction to form several local defect areas; For each localized defect area, determine whether it retains stable intrusion characteristics after deformation compensation: If the local defect area satisfies the following at multiple adjacent sampling points If the defective area is identified, then the localized defective area is determined to be a candidate area for actual fracture.

[0008] In a preferred embodiment, the process of calculating the fracture depth, fracture extension length, and effective fracture area after removing the influence of elastic deformation for the actual fracture candidate region, and forming a set of physical measurement parameters for the fracture edge, is as follows: Obtain the set of defect contour points corresponding to each real fracture candidate region; For each real fracture candidate region, a local reference coordinate system for the real fracture candidate region is determined based on the real geometric boundary; under the local reference coordinate system, the normal intrusion distance of each defect contour point in the real fracture candidate region relative to the real geometric boundary is calculated. Calculate the edge depth of the actual fracture candidate region based on the normal inward intrusion distance; Based on the spatial distribution of the defect contour points in the tangential direction, calculate the fracture extension length of the actual fracture candidate region; The effective fracture area of ​​the real fracture candidate region is calculated based on the defect area enclosed between the real geometric boundary and the actual cutting edge contour. The fracture depth, fracture extension length, and effective fracture area corresponding to each real fracture candidate region are combined to form a set of physical measurement parameters for fracture.

[0009] In a preferred embodiment, the process of verifying the consistency between the set of physical measurement parameters for edge chipping and the deformation-sensitive descriptive quantities at the corresponding cutting position, generating an edge chipping level evaluation result, and outputting the edge chipping detection measurement result associated with the cutting position and cutting process parameters is as follows: Obtain the set of physical measurement parameters of the fracture edge corresponding to each real fracture candidate region, and obtain the deformation sensitive descriptive quantity corresponding to the cutting position analysis window to which the fracture candidate region belongs; Based on the set of physical measurement parameters of the chipped edge, a comprehensive chipped edge strength description quantity is constructed; After normalizing the comprehensive edge breakage strength description and the deformation sensitivity description, a consistency verification index is constructed for consistency verification to determine whether the current edge breakage conforms to the physical response characteristics of the material's actual fracture. The validity of edge collapse is determined based on the magnitude of the consistency verification index: When the consistency verification index is greater than or equal to the preset consistency threshold, the candidate region of the fracture is determined to be a physically consistent real fracture edge; when the consistency verification index is less than the consistency threshold, the candidate region of the fracture is determined to be an abnormal fracture edge suspected of being affected by deformation interference, and is marked as a low confidence result. The actual edge collapses that pass the consistency check will be determined according to the comprehensive edge collapse strength description. The numerical range generates the edge collapse level evaluation results; The chipping level evaluation results are associated with the corresponding cutting position, cutting path number, and cutting process parameters; The associated edge breakage detection measurement results are output and stored to form an edge breakage detection record that includes cutting location, deformation sensitivity description, edge breakage physical measurement parameter set, comprehensive edge breakage strength description, and edge breakage level evaluation results.

[0010] The technical effects and advantages of this invention are as follows: 1. This invention introduces a deformation-sensitive modeling and compensation measurement mechanism into the wafer dicing edge detection process, effectively separating the influence of the overall elastic deflection of the ultrathin wafer on the edge imaging profile from the crack identification process, and achieving high-precision identification and quantitative evaluation of real edge defects without applying additional clamping force or introducing additional load or residual stress interference.

[0011] 2. This invention constructs a deformation-sensitive descriptor for the cutting position based on cutting load, material elasticity, and thickness distribution. Based on this, it performs deformation component analysis and baseline boundary compensation correction on the cutting kerf edge contour, making the obtained true geometric boundary closer to the actual shape of the wafer under no external force. This eliminates the interference of pseudo-defects introduced by elastic deformation at the source. Simultaneously, by only judging the fracture in local defect areas that still exceed the true geometric boundary after compensation, and further calculating the fracture depth, extension length, and effective fracture area after removing the influence of elastic deformation, the authenticity and stability of the physical measurement results of the fracture edge are significantly improved. Consistency verification is performed using the deformation-sensitive descriptor to avoid misjudgment of the fracture level in highly deformation-sensitive areas. This achieves a high consistency correlation between the fracture edge detection results and the cutting position and process parameters, effectively solving the key technical problem of difficulty in distinguishing between actual fracture and overall elastic deformation coupling in ultra-thin wafer cutting, and significantly improving the accuracy and reliability of fracture edge detection and measurement. Attached Figure Description

[0012] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example: Figure 1 This invention provides a method for detecting and measuring edge chipping during wafer dicing, comprising the following steps: Obtain the wafer thickness distribution, cutting load history parameters, and material elastic parameters in the vicinity of the wafer dicing track; calculate the equivalent bending stiffness of the dicing area; and obtain the deformation-sensitive descriptor of the corresponding dicing position. Based on the deformation-sensitive descriptor, the deformation component of the imaging profile of the cutting path edge is analyzed, the mapping relationship between the edge profile normal offset and the overall elastic deflection response is established, the deformation compensation correction of the cutting path reference boundary is performed, and the true geometric boundary is obtained. Align and compare the real geometric boundary with the actual cutting edge contour, and extract the local defect area that still exceeds the real geometric boundary after deformation compensation as the real fracture candidate area. For the actual fracture candidate region, calculate the fracture depth, fracture extension length and effective fracture area after removing the influence of elastic deformation, and form a set of physical measurement parameters for the fracture edge; The consistency of the physical measurement parameter set of edge chipping with the deformation-sensitive descriptive quantity of the corresponding cutting position is verified, an edge chipping level evaluation result is generated, and the edge chipping detection measurement result associated with the cutting position and cutting process parameters is output.

[0015] In this embodiment of the invention, the process of obtaining the wafer thickness distribution, historical cutting load parameters, and material elastic parameters in the vicinity of the wafer dicing track, calculating the equivalent bending stiffness of the dicing region, and obtaining the deformation-sensitive descriptor of the corresponding dicing position is as follows: In wafer dicing equipment, acquire thickness distribution data, historical cutting load parameters, and wafer material elastic parameters in the vicinity of the wafer dicing track. The thickness distribution data includes local thickness values ​​along the cutting track normal and the cutting direction; The historical cutting load parameters include blade rotation speed, feed rate, cutting depth and corresponding cutting force, and load records; The material elastic parameters include the material elastic modulus. Compared with Poisson ; It should be noted that the material's elastic modulus Poisson's ratio is used to characterize the ability of wafer materials to resist elastic deformation under stress. Used to characterize the coupling relationship between lateral and longitudinal deformation of wafer materials under tension or compression; by simultaneously introducing the material's elastic modulus. Compared with Poisson This allows for a more realistic depiction of the overall bending response characteristics of ultrathin wafers under cutting loads, thereby avoiding the estimation bias of equivalent bending stiffness caused by relying solely on a single material parameter.

[0016] It should also be noted that the elastic modulus of the material Compared with Poisson These are the inherent mechanical parameters of the wafer material, which can be obtained directly by reading the wafer material specification parameter table or the material parameter database associated with wafer batch information; The neighborhood of the cutting path is divided according to a preset spatial window length, resulting in multiple cutting position analysis windows. ; It is understood that the length of the spatial window preferably covers a locally cut stable interval, which can reflect the local mechanical response characteristics and avoid parameter aliasing caused by crossing different support states.

[0017] Analysis window for each cutting position The average thickness within the window is calculated based on the collected thickness distribution data. ; The average thickness can be calculated using the average value of the thickness data within the window, and is used to reflect the most unfavorable bending stiffness state in this region.

[0018] According to the average thickness Material elastic modulus Compared with Poisson Calculate the equivalent bending stiffness of the cut region; For example, the equivalent bending stiffness corresponding to the cutting position analysis window i Calculate using the following formula: ; Understandably, equivalent bending stiffness is used to characterize the ability of an ultrathin wafer to resist elastic flexural deformation under a unit bending curvature, with an average thickness... The cubic relationship determines the significant amplification effect of small changes in thickness on the deformation response.

[0019] Based on the historical cutting load parameters, extract the cutting load corresponding to the cutting position analysis window i. ; For example, cutting load It can be obtained by weighted summation of measured cutting force, feed rate, cutting depth and insert speed; The cutting load With equivalent bending stiffness Perform correlation and calculate the deformation response of the cutting position analysis window i; For example, deformation response quantity It can be calculated as follows: ; Understandably, the deformation response amount It is used to characterize the relative strength of elastic deflection at the cut location due to insufficient bending stiffness under the same load conditions.

[0020] According to the deformation response amount After normalization, the deformation-sensitive descriptor of the cutting position analysis window i is obtained. ; For example, deformation-sensitive descriptor The calculation method is as follows: ,in, This represents the maximum value of the deformation response across all analysis windows within the same wafer or dicing batch.

[0021] Understandably, deformation-sensitive descriptive quantities The physical meaning of this parameter is that it comprehensively reflects the susceptibility of overall elastic deflection at the cutting position under a given cutting load due to the ultra-thin characteristics of the wafer; it is a deformation sensitivity descriptor. The larger the value, the more likely the location is to undergo significant elastic deformation during the cutting process, thus posing a greater risk of projection interference to the edge breakage detection results.

[0022] It should also be noted that by introducing a deformation-sensitive descriptor, the deformation risk at different cutting positions can be quantitatively distinguished without applying additional clamping force or external constraints, providing a physically consistent prior basis for the subsequent analysis of the deformation components of the cutting edge contour and the identification of actual fractures.

[0023] In this embodiment of the invention, based on the deformation-sensitive descriptor, the process of analyzing the deformation components of the imaging contour of the cutting edge, establishing the mapping relationship between the edge contour normal offset and the overall elastic deflection response, and performing deformation compensation correction on the cutting reference boundary to obtain the true geometric boundary is as follows: The original imaging contour data of the cutting path edge is obtained by an industrial camera or a line scan imaging device continuously acquiring data along the cutting path direction to form an edge contour curve composed of multiple sets of edge pixels. The original imaging contour data is preprocessed, including denoising, grayscale normalization, and sub-pixel edge localization, to extract the initial contour point set of the cutting edge. Each contour point This includes its spatial position in the imaging coordinate system and the corresponding cutting position analysis window identifier; According to the initial contour point set Calculate the local tangent direction at each contour point along the edge of the cutting path, and determine the corresponding normal direction vector. This is used to characterize the geometrically normal offset direction of the contour points: The initial contour point set Each contour point This indicates the spatial coordinates of the cutting edge in the imaging coordinate system; According to the extension direction of the cutting path, the initial contour point set is... Arrange them in an orderly manner so that adjacent contour points are distributed sequentially along the cutting path direction in the sequence; For any contour point in the contour point sequence Select its previous contour point in the cutting direction. With the next contour point Construct the local tangent vector at the contour point. ; For example, the local tangent vector The calculation method is as follows: ; It is understandable that by using the difference method between adjacent contour points, the local trend of the cutting edge at the contour point can be effectively characterized, and the influence of single-point noise on the direction calculation can be reduced.

[0024] For the local tangent vector After normalization, the unit tangent direction vector is obtained. The calculation method is as follows: ,in, This represents the magnitude of the local tangent vector.

[0025] According to the unit tangent direction vector Construct a normal direction vector orthogonal to it. The normal direction vector is used to characterize the geometrically normal offset direction of the cutting edge; For example, in a two-dimensional imaging coordinate system, the normal direction vector It can be calculated as follows: or ,in, and These represent the components of the unit tangent direction vector along the x-axis and y-axis, respectively.

[0026] It is understandable that the two normal direction vectors mentioned above correspond to the inner and outer normal directions of the dicing edge, respectively. In practical applications, the normal direction pointing to the inside of the wafer can be selected as the effective normal direction vector based on the positional relationship between the wafer dicing track and the chip area.

[0027] Each contour point Associate it with the cutting position analysis window i to which it belongs, and obtain the deformation sensitivity descriptor corresponding to that window. And establish the correlation between the normal displacement of the contour points and the deformation-sensitive descriptive quantities: Each contour point The contour points are mapped to the corresponding cutting position analysis window i according to their spatial position along the cutting path, establishing a one-to-one association between the contour points and the cutting position analysis windows; where cutting position analysis window i is a local analysis unit obtained based on the cutting path length direction, and each analysis window corresponds to a unique deformation-sensitive descriptor. This is used to characterize the susceptibility of the local cut area to overall elastic deflection under cutting load; According to the deformation sensitivity descriptor Construct a projection model of the overall elastic flexural response onto the edge contour to obtain the contour points. normal deformation projection amount ; It should be noted that the overall elastic flexural response refers to the bending deformation behavior of the wafer as a whole thin plate structure, dominated by elastic deformation, during the wafer dicing process, especially under ultra-thin wafer conditions, where the wafer, under the combined action of dicing load and dicing vibration, does not only experience material removal or fracture in the local area where the blade contacts, but also undergoes bending deformation. This response manifests as the overall flexural deformation of the wafer along the thickness direction during the transient or quasi-steady-state phases of dicing. Its deformation is continuous in space, recoverable in time, and can partially or completely spring back after the external load is removed.

[0028] From a physical perspective, the overall elastic flexural response originates from the sharp decrease in bending stiffness after the wafer thickness is significantly reduced. This causes the normal cutting force applied by the blade, cutting vibration, and changes in local support conditions during the cutting process to be transformed into elastic bending deformation across the vicinity of the cutting kerf and even a larger area. This flexural deformation is not accompanied by permanent damage to the material structure, and its nature is different from the actual fracture defects caused by brittle fracture.

[0029] At the imaging and measurement level, the overall elastic flexural response causes the edge of the cut track to produce a systematic normal displacement projection in the imaging coordinate system. This causes the acquired edge contour to shift or slowly change relative to the true geometric boundary of the wafer under no external force, thus visually presenting a pseudo-defect feature similar to edge chipping.

[0030] For example, contour points normal deformation projection amount It can be represented as: ,in, For deformation scale mapping coefficients, used to map dimensionless deformation-sensitive descriptors into normal displacement quantities with actual physical meaning; It is understandable that the deformation scale mapping coefficients It can be obtained from equipment calibration, historical cutting data, or experimental testing.

[0031] By introducing a correction term related to the cutting direction, the normal deformation projection is corrected by direction weighting to obtain the corrected normal deformation projection. ; For example, the corrected normal deformation projection amount It can be represented as: ,in, This indicates the angle between the normal direction at the contour point and the principal direction of the overall deflection.

[0032] According to the normal deformation projection amount The original contour point positions are compensated in reverse along the normal direction of the contour points to obtain the deformed contour points. The calculation method is as follows: ; All deformation-compensated contour points Reconstruct the cutting path according to the cutting direction to form the deformation compensation boundary curve of the cutting path, which serves as the true geometric boundary under the condition of no external force interference. It is understandable that the real geometric boundary is systematically compensated for the normal projection offset caused by the overall elastic deflection, so that the subsequent edge breakage identification and measurement process can be carried out based on the geometric boundary that is close to the actual fracture state of the material. It should also be noted that by segmenting and compensating the edge contour based on the deformation-sensitive descriptor, this method avoids the additional stress interference caused by adding clamping or external constraints to suppress wafer deformation, thereby improving the physical consistency and stability of the cutting edge detection measurement without changing the actual physical response of the wafer.

[0033] In this embodiment of the invention, the process of aligning and comparing the true geometric boundary with the actual cutting edge contour, and extracting the local defect areas that still exceed the true geometric boundary after deformation compensation as candidate areas for true fracture is as follows: After deformation compensation, the true geometric boundary curve of the cutting path and the actual cutting edge contour curve of the corresponding cutting path are obtained. The actual cutting edge contour curve is directly extracted from the edge imaging data after cutting. The real geometric boundary curve and the actual cutting edge contour curve are spatially aligned so that they have a consistent cutting direction reference and position reference in the same coordinate system. The alignment process includes translation correction based on the center line of the cutting path and angle correction based on the cutting direction to eliminate the influence of imaging installation error and overall pose deviation on the results. After spatial alignment is completed, the actual geometric boundary curve is parameterized along the cutting path direction to obtain a set of reference points arranged in the order of the cutting path. Simultaneously, the actual cutting edge contour curve is sampled at the same scale to obtain an actual contour point set that corresponds one-to-one with the reference point set. ; For each pair of corresponding reference points With actual contour points Calculate the normal offset distance of the actual contour points relative to the true geometric boundary. ; For example, normal offset distance The calculation method is as follows: ,in, This indicates that the true geometric boundary is at the reference point. The outward normal direction vector at that location.

[0034] According to the normal offset distance The sign and numerical value are used to determine the defects of actual contour points: When the normal offset distance When the value is less than zero, it indicates that the actual contour point is located inside the true geometric boundary and belongs to the potential defect point that invades the interior of the wafer; When the normal offset distance When the value is greater than or equal to zero, it indicates that no effective defect has been formed at that point; A minimum defect threshold is introduced among the potential defect points. Perform the screening; When satisfied When this occurs, the actual contour point is determined to be a true defect point; It is understood that the minimum defect threshold is used to filter out minor pseudo-offsets caused by imaging noise and sub-pixel positioning errors, and is a preset value.

[0035] Multiple real defect points that are continuously distributed in space are clustered according to the cutting direction to form several local defect regions; each local defect region consists of a set of contour points that are spatially adjacent and meet the criteria for determining real defects. For each localized defect area, determine whether it retains stable intrusion characteristics after deformation compensation: If the local defect area satisfies the following at multiple adjacent sampling points If so, the local defect area is determined to be a real fracture candidate area; It should be noted that multiple adjacent sampling points refer to two or more adjacent sampling points; Understandably, by comparing the normal offset based on the real geometric boundary and introducing the dual constraints of deformation compensation and minimum defect threshold, this process effectively eliminates the false edge breakage caused by the overall elastic flexural projection, imaging error or edge noise, and retains only the real fracture candidate region formed by the local brittle fracture of the material. It should also be noted that the extraction results of the actual fracture candidate region provide a reliable geometric basis for the subsequent physical measurement of fracture depth, extension length and effective fracture area, avoiding the secondary superposition of deformation factors in the measurement stage.

[0036] In this embodiment of the invention, the process of calculating the fracture depth, fracture extension length, and effective fracture area after removing the influence of elastic deformation for the actual fracture candidate region, and forming a set of physical measurement parameters for the fracture edge, is as follows: Obtain the set of defect contour points corresponding to each real fracture candidate region. The set of defect contour points consists of multiple actual cutting edge contour points that are continuously distributed inside the real geometric boundary. For each real fracture candidate region, a local reference coordinate system for the real fracture candidate region is determined based on the real geometric boundary; wherein, the tangential direction of the local reference coordinate system is set along the cutting path direction, and the normal direction is set along the outward normal direction of the real geometric boundary, which is used to unify the calculation basis of the fracture edge geometric parameters. In the local reference coordinate system, calculate the normal intrusion distance of each defect contour point within the actual fracture candidate region relative to the actual geometric boundary. ; For example, for the first in the real fracture candidate region For each missing contour point, its normal intrusion distance The calculation method is as follows: ,in, Indicates the actual cutting edge contour points. This represents the corresponding real geometric boundary reference point. This represents the outward normal direction vector at the reference point; It is understood that the normal inward distance has already undergone deformation compensation in the aforementioned steps, and therefore no longer includes the projection component introduced by the overall elastic deflection.

[0037] Calculate the edge depth of the actual fracture candidate region based on the normal inward intrusion distance; For example, the depth of the chipped edge Defined as the maximum value of all normal intrusion distances within the actual fracture candidate region, its calculation formula is: ; Based on the spatial distribution of the defect contour points in the tangential direction, calculate the fracture extension length of the actual fracture candidate region; For example, fracture extension length Defined as the projected length of the actual fracture candidate region along the cutting direction, it is calculated as follows: ,in, Indicates the first The arc length coordinates of each missing contour point in the cutting direction.

[0038] The effective fracture area of ​​the real fracture candidate region is calculated based on the defect area enclosed between the real geometric boundary and the actual cutting edge contour. For example, effective fracture area This is obtained by calculating the area integral of the closed region between the real geometric boundary and the actual cutting edge contour, and can be exemplarily represented as follows: ,in, This represents the normal intrusion distance function corresponding to the parameter s along the cutting direction.

[0039] It should be noted that s represents the tangential position parameter after parameterizing the real geometric boundary along the cutting path direction, that is, a two-dimensional coordinate position parameter, which is used to characterize the spatial order of different cutting positions in the cutting direction; This represents the intrusion distance function of the actual cutting edge contour relative to the real geometric boundary in the normal direction at the cutting position corresponding to parameter s; It is understood that the chipping depth, chipping extension length, and effective chipping area are all physical measurement parameters obtained after eliminating the influence of overall elastic deformation, which can truly reflect the degree of edge damage caused by local brittle fracture of the material.

[0040] The collapse edge depth corresponding to each real collapse candidate region , fracture extension length and effective fracture area Combine them to form a set of physical measurement parameters for edge collapse: ; It should also be noted that by constructing a set of physical measurement parameters for edge chipping, this method avoids the problem that a single geometric index is insufficient to characterize complex fracture morphologies, and makes the edge chipping detection results more comparable and stable under different wafer thicknesses and different cutting processes.

[0041] In this embodiment of the invention, the process of performing consistency verification between the set of physical measurement parameters for edge chipping and the deformation-sensitive descriptive quantities at the corresponding cutting position, generating an edge chipping level evaluation result, and outputting edge chipping detection measurement results associated with the cutting position and cutting process parameters is as follows: Obtain the set of physical measurement parameters of the fracture edge corresponding to each real fracture candidate region. And obtain the deformation-sensitive descriptor corresponding to the cutting position analysis window to which the candidate fracture region belongs. ; Based on the aforementioned set of physical measurement parameters for edge chipping, a comprehensive descriptive quantity for edge chipping strength is constructed. This is used to characterize the overall fracture severity of the actual fracture candidate region; For example, the comprehensive edge breakage strength descriptor is calculated after normalizing the edge breakage depth, fracture extension length, and effective fracture area. It can be calculated as follows: ,in, , , Preset weighting coefficients for chipping depth, chipping extension length, and effective chipping area; Understandably, the weighting coefficients can be set based on process experience, historical data, or quality specifications.

[0042] The comprehensive edge breakage strength descriptor Deformation-sensitive descriptor After normalization, a consistency verification index is constructed for consistency verification to determine whether the current edge breakage conforms to the physical response characteristics of the actual fracture of the material. For example, consistency check index It can be represented as: ,in, To prevent extremely small positive numbers with a denominator of zero (generally taken as...) ).

[0043] According to the consistency verification index The validity of edge collapse is determined by the numerical value: When the consistency verification index is greater than or equal to the preset consistency threshold, the candidate region of the fracture is determined to be a physically consistent real fracture edge; when the consistency verification index is less than the consistency threshold, the candidate region of the fracture is determined to be an abnormal fracture edge suspected of being affected by deformation, and can be marked as a low confidence result. The actual edge collapses that pass the consistency check will be determined according to the comprehensive edge collapse strength description. The numerical range generates the edge collapse level evaluation results; For example, edge collapse levels can be classified as follows: ,in, , These are the preset first threshold and second threshold for edge collapse level, respectively.

[0044] The chipping level evaluation results are associated with the corresponding cutting position, cutting path number, and cutting process parameters; wherein, the cutting process parameters include at least the blade rotation speed, feed rate, cutting depth, and cutting sequence number; The associated chipping detection measurement results are output and stored to form a chipping detection record that includes cutting location, deformation sensitivity description, chipping physical measurement parameter set, comprehensive chipping strength description, and chipping grade evaluation results. This record is used for subsequent cutting process optimization, anomaly tracing analysis, and quality statistical evaluation.

[0045] Understandably, by verifying the consistency between the physical measurement parameters of edge chipping and the deformation-sensitive descriptor, this method can effectively avoid misjudging minor fractures in high deformation-sensitive areas, while suppressing false severe edge chipping judgments caused by overall elastic deflection amplification, thereby significantly improving the reliability and engineering applicability of edge chipping detection results in ultra-thin wafer dicing scenarios.

[0046] This invention introduces a deformation-sensitive modeling and compensation measurement mechanism into the wafer dicing edge detection process, effectively separating the influence of the overall elastic deflection of the ultrathin wafer on the edge imaging profile from the crack identification process. This achieves high-precision identification and quantitative evaluation of real edge defects without applying additional clamping force or introducing additional loads or residual stress interference.

[0047] This invention constructs a deformation-sensitive descriptor for the cutting position based on cutting load, material elasticity, and thickness distribution. Based on this, it performs deformation component analysis and baseline boundary compensation correction on the cutting kerf edge contour, making the obtained true geometric boundary closer to the actual shape of the wafer under no external force. This eliminates the interference of pseudo-defects introduced by elastic deformation at the source. Simultaneously, by only judging the fracture in local defect areas that still exceed the true geometric boundary after compensation, and further calculating the fracture depth, extension length, and effective fracture area after removing the influence of elastic deformation, the authenticity and stability of the physical measurement results of the fracture edge are significantly improved. Consistency verification is performed using the deformation-sensitive descriptor to avoid misjudgment of the fracture level in highly deformation-sensitive areas. This achieves a high consistency correlation between the fracture edge detection results and the cutting position and process parameters, effectively solving the key technical problem of difficulty in distinguishing between actual fracture and overall elastic deformation coupling in ultra-thin wafer cutting, and significantly improving the accuracy and reliability of fracture edge detection and measurement.

[0048] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0049] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0050] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting and measuring edge chipping during wafer dicing, characterized in that: Includes the following steps: Obtain the wafer thickness distribution, cutting load history parameters, and material elastic parameters in the vicinity of the wafer dicing track; calculate the equivalent bending stiffness of the dicing area; and obtain the deformation-sensitive descriptor of the corresponding dicing position. Based on the deformation-sensitive descriptor, the deformation component of the imaging profile of the cutting path edge is analyzed, the mapping relationship between the edge profile normal offset and the overall elastic deflection response is established, the deformation compensation correction of the cutting path reference boundary is performed, and the true geometric boundary is obtained. Align and compare the real geometric boundary with the actual cutting edge contour, and extract the local defect area that still exceeds the real geometric boundary after deformation compensation as the real fracture candidate area. For the actual fracture candidate region, calculate the fracture depth, fracture extension length and effective fracture area after removing the influence of elastic deformation, and form a set of physical measurement parameters for the fracture edge; The consistency of the physical measurement parameter set of edge chipping with the deformation-sensitive descriptive quantity of the corresponding cutting position is verified, an edge chipping level evaluation result is generated, and the edge chipping detection measurement result associated with the cutting position and cutting process parameters is output.

2. The method for detecting and measuring edge chipping during wafer dicing according to claim 1, characterized in that: The process of obtaining the wafer thickness distribution, historical dicing load parameters, and material elastic parameters in the vicinity of the wafer dicing kerf, calculating the equivalent bending stiffness of the dicing region, and obtaining the deformation-sensitive descriptor for the corresponding dicing location is as follows: In wafer dicing equipment, acquire thickness distribution data, historical cutting load parameters, and wafer material elastic parameters in the vicinity of the wafer dicing track. The thickness distribution data includes local thickness values ​​along the cutting track normal and the cutting direction; The historical parameters of the cutting load include the blade rotation speed, feed rate, cutting depth and corresponding cutting force; The material's elastic parameters include the material's elastic modulus and Poisson's ratio; The neighborhood of the cutting path is divided according to a preset spatial window length, resulting in multiple cutting position analysis windows. ; Analysis window for each cutting position The average thickness within the window is calculated based on the collected thickness distribution data. Calculate the equivalent bending stiffness of the cut region based on the average thickness, material elastic modulus, and Poisson's ratio. Based on the historical cutting load parameters, extract the cutting load corresponding to the cutting position analysis window i; The cutting load is correlated with the equivalent bending stiffness to calculate the deformation response of the cutting position analysis window i. Based on the deformation response, it is normalized to obtain the deformation sensitivity description of the cutting position analysis window i.

3. The method for detecting and measuring edge chipping during wafer dicing according to claim 2, characterized in that: Based on the aforementioned deformation-sensitive descriptor, the deformation component of the imaging profile at the edge of the cutting path is analyzed, a mapping relationship is established between the edge profile normal offset and the overall elastic deflection response, and deformation compensation correction is performed on the cutting path reference boundary to obtain the true geometric boundary. Obtain the original imaging contour data of the cutting path edge, preprocess the original imaging contour data, and extract the initial contour point set of the cutting path edge; Based on the initial set of contour points, calculate the local tangent direction at each contour point along the edge of the cutting path, and determine the corresponding normal direction vector. Associate each contour point with its corresponding cutting position analysis window i, obtain the deformation-sensitive descriptive quantity corresponding to the window, and establish the correlation between the contour point normal displacement and the deformation-sensitive descriptive quantity. Based on the deformation-sensitive descriptor, a projection model of the overall elastic flexural response on the edge contour is constructed to obtain the normal deformation projection of the contour points. By introducing a correction term related to the cutting direction, the normal deformation projection is corrected by directional weighting to obtain the corrected normal deformation projection. Based on the normal deformation projection amount, the original contour point position is reverse-compensated along the normal direction of the contour point to obtain the deformed contour point. All deformation-compensated contour points are reconstructed according to the cutting path direction to form the deformation-compensated boundary curve of the cutting path, which serves as the true geometric boundary under conditions without external force interference.

4. The method for detecting and measuring edge chipping during wafer dicing according to claim 3, characterized in that: The process of aligning and comparing the actual geometric boundary with the actual cutting edge contour, and extracting the local defect areas that still exceed the actual geometric boundary after deformation compensation as candidate areas for actual fracture is as follows: Obtain the true geometric boundary curve of the cutting path and the actual cutting edge contour curve of the corresponding cutting path after deformation compensation; The true geometric boundary curve and the actual cutting edge contour curve are spatially aligned. After spatial alignment is completed, the real geometric boundary curve is parametrically represented along the cutting path direction to obtain a set of reference points arranged in the order of the cutting path. At the same time, the actual cutting edge contour curve is sampled at the same scale to obtain an actual contour point set that corresponds one-to-one with the reference point set. For each pair of corresponding reference points and actual contour points, calculate the normal offset distance of the actual contour points relative to the true geometric boundary. ; According to the normal offset distance The sign and numerical value are used to determine the defects of actual contour points: When the normal offset distance When the value is less than zero, it indicates that the actual contour point is located inside the true geometric boundary and belongs to the potential defect point that invades the interior of the wafer; When the normal offset distance When the value is greater than or equal to zero, it indicates that the contour points have not formed a valid defect; A minimum defect threshold is introduced among the potential defect points. Perform the screening; When satisfied When the actual contour point is determined to be the real defect point; Multiple real defect points that are continuously distributed in space are clustered according to the cutting direction to form several local defect areas; For each localized defect area, determine whether it retains stable intrusion characteristics after deformation compensation: If the local defect area satisfies the following at multiple adjacent sampling points If the defective area is identified, then the localized defective area is determined to be a candidate area for actual fracture.

5. The method for detecting and measuring edge chipping during wafer dicing according to claim 4, characterized in that: The process of calculating the fracture depth, fracture extension length, and effective fracture area after removing the influence of elastic deformation for the actual fracture candidate region, and forming the set of physical measurement parameters for the fracture edge, is as follows: Obtain the set of defect contour points corresponding to each real fracture candidate region; For each real fracture candidate region, the local reference coordinate system of the real fracture candidate region is determined based on the real geometric boundary; In the local reference coordinate system, calculate the normal inward distance of each defect contour point in the real fracture candidate area relative to the real geometric boundary; Calculate the edge depth of the actual fracture candidate region based on the normal inward intrusion distance; Based on the spatial distribution of the defect contour points in the tangential direction, calculate the fracture extension length of the actual fracture candidate region; The effective fracture area of ​​the real fracture candidate region is calculated based on the defect area enclosed between the real geometric boundary and the actual cutting edge contour. The fracture depth, fracture extension length, and effective fracture area corresponding to each real fracture candidate region are combined to form a set of physical measurement parameters for fracture.

6. The method for detecting and measuring edge chipping during wafer dicing according to claim 5, characterized in that: The process of verifying the consistency between the physical measurement parameter set of edge chipping and the deformation-sensitive descriptive quantity of the corresponding cutting position, generating an edge chipping level evaluation result, and outputting the edge chipping detection measurement result associated with the cutting position and cutting process parameters is as follows: Obtain the set of physical measurement parameters of the fracture edge corresponding to each real fracture candidate region, and obtain the deformation sensitive descriptive quantity corresponding to the cutting position analysis window to which the fracture candidate region belongs; Based on the set of physical measurement parameters of the chipped edge, a comprehensive chipped edge strength description quantity is constructed; After normalizing the comprehensive edge breakage strength description and the deformation sensitivity description, a consistency verification index is constructed for consistency verification to determine whether the current edge breakage conforms to the physical response characteristics of the material's actual fracture. The validity of edge collapse is determined based on the magnitude of the consistency verification index: When the consistency verification index is greater than or equal to the preset consistency threshold, the candidate region of the collapse is determined to be a physically consistent real collapse edge. When the consistency verification index is less than the consistency threshold, the candidate region of the fracture is determined to be an abnormal fracture edge suspected of being affected by deformation, and is marked as a low confidence result. The actual edge collapses that pass the consistency check will be determined according to the comprehensive edge collapse strength description. The numerical range generates the edge collapse level evaluation results; The chipping level evaluation results are associated with the corresponding cutting position, cutting path number, and cutting process parameters; The associated edge breakage detection measurement results are output and stored to form an edge breakage detection record that includes cutting location, deformation sensitivity description, edge breakage physical measurement parameter set, comprehensive edge breakage strength description, and edge breakage level evaluation results.

Images