A kind of ultra-thin wafer parameter test system and method based on topography detection

Abstract

This application discloses a method and system for testing ultrathin wafer parameters based on topography detection, relating to the field of wafer parameter testing technology. The method includes performing parameter testing on the ultrathin wafer under test, obtaining a topography mapping matrix and extracting irregular contour trajectory data; calculating the local curvature value of each sampling point, and using the coordinates of sampling points exceeding a preset curvature threshold as a set of high-curvature region coordinates; generating a non-periodic time jitter sequence and superimposing it onto a basic laser trigger frequency sequence to obtain a time-modulated trigger sequence; calculating a multi-focus spatial coordinate matrix, corresponding phase mask data, and excitation delay sequences between each focus point based on the trajectory tangent vector of each coordinate point; and then generating the wafer parameter test results for the ultrathin wafer under test. This method achieves topography parameter identification and high-curvature region localization for complex contour regions of ultrathin wafers, improving the accuracy of wafer parameter test results in characterizing edge topography changes and local abnormal regions.

Description

Technical Field

[0001] This application relates to the field of wafer parameter testing technology, and in particular to an ultrathin wafer parameter testing system and method based on morphology detection. Background Technology

[0002] Existing wafer profiling technologies, particularly for defects in conventional thicknesses or straight profiles, primarily focus on reducing the heat-affected zone (HAZ) of lasers or optimizing mechanical forces. However, when wafer thickness decreases to the micrometer level and the profile trajectory becomes a complex geometric curve, even with good thermal effect control, high-frequency, irregular edge defects, microcracks, or delayed cleavage cracks still appear at the bends of irregular trajectories, especially at inner radius corners. Traditional processing or testing systems often employ open-loop profiling methods or rely solely on two-dimensional visual alignment, lacking the capability for high-precision wafer parameter testing of the three-dimensional surface morphology and structural features of ultra-thin wafers. Without supporting data on the actual physical morphology of the wafer, the system cannot quantitatively analyze the material mechanical response, curvature changes, and potential defect regions along complex geometric trajectories, nor can it accurately obtain the parameter distribution characteristics corresponding to irregular profile regions. Summary of the Invention

[0003] To address the aforementioned shortcomings, this application provides an ultrathin wafer parameter testing system and method based on topology detection, which achieves acoustic topology protection for ultrathin wafers by actively intervening in the generation frequency, spatial propagation direction, and boundary dissipation characteristics of shock waves.

[0004] In a first aspect, this application provides a method for testing parameters of ultrathin wafers based on morphology detection, including:

[0005] Perform wafer parameter testing on the ultrathin wafer under test, obtain the topography mapping matrix of the surface of the ultrathin wafer under test, and extract the irregular contour trajectory data based on the topography mapping matrix;

[0006] Calculate the local curvature value of each sampling point in the irregular contour trajectory data to generate a trajectory curvature distribution sequence; extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold to obtain a set of high curvature region coordinates.

[0007] Based on the trajectory curvature distribution sequence and the high curvature region coordinate set, an aperiodic time jitter sequence is generated; the aperiodic time jitter sequence is superimposed on the basic trigger frequency sequence to obtain a time modulation trigger sequence.

[0008] Extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set. Based on the trajectory tangent vector, calculate and generate a multi-focus spatial coordinate matrix and the corresponding phase mask data, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix.

[0009] The topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data, and the excitation delay sequence are correlated to generate the wafer parameter test results of the ultrathin wafer under test.

[0010] Optionally, calculating the local curvature values ​​of each sampling point in the irregular contour trajectory data and generating a trajectory curvature distribution sequence includes:

[0011] Extract the coordinates of continuous sampling points from the irregular contour trajectory data and perform polynomial curve fitting to construct a smooth contour trajectory curve.

[0012] Calculate the first and second derivatives of the coordinates of each sampling point at the smooth contour trajectory curve, and use the first and second derivatives to calculate the local curvature value;

[0013] Arrange the local curvature values ​​of all sampling points in the sampling order to generate a trajectory curvature distribution sequence.

[0014] Optionally, the step of extracting the coordinates of sampling points whose absolute values ​​of the local curvature values ​​exceed a preset curvature threshold to obtain a set of coordinates for high curvature regions includes:

[0015] Calculate the absolute value of each local curvature value in the trajectory curvature distribution sequence;

[0016] The absolute value is calculated by comparing it with a preset curvature threshold, and the coordinates of the sampling points with a difference greater than zero are extracted as candidate coordinate points. The distance between adjacent paths between each candidate coordinate point is calculated, and the candidate coordinate points with adjacent path distances less than a preset merging spacing threshold are clustered and merged to generate a set of coordinates for high curvature regions.

[0017] Optionally, generating a non-periodic time jitter sequence based on the trajectory curvature distribution sequence and the set of coordinates of the high curvature region includes:

[0018] Extract the local curvature value of each candidate coordinate point in the high curvature region coordinate set in the trajectory curvature distribution sequence, and use it as the target local curvature value;

[0019] The target local curvature value is input into a preset pulse delay mapping function to calculate the pulse delay offset corresponding to each candidate coordinate point;

[0020] The pulse delay offset is converted into a time interval value according to the sampling order of each candidate coordinate point to generate an aperiodic time jitter sequence.

[0021] Optionally, the step of superimposing the aperiodic time jitter sequence onto the base trigger frequency sequence to obtain the time-modulated trigger sequence includes:

[0022] Obtain a preset fixed pulse time interval and generate a basic trigger frequency sequence;

[0023] On the same time axis, the pulse delay offset in the aperiodic time jitter sequence is superimposed with the fixed pulse time interval at the corresponding moment in the basic trigger frequency sequence;

[0024] The pulse timestamp sequence containing non-uniform time intervals generated after superposition is used as the time modulation trigger sequence.

[0025] Optionally, the step of calculating and generating a multi-focal spatial coordinate matrix based on the trajectory tangent vector includes:

[0026] Extract the trajectory tangent vector of each candidate coordinate point in the coordinate set of the high curvature region, and calculate the normal offset vector that is perpendicular to the trajectory tangent vector;

[0027] Obtain the preset number of focal points and the basic focal space spacing;

[0028] Using each candidate coordinate point as a reference, along the direction of the normal offset vector, calculate the three-dimensional coordinate data of the remaining related foci according to the basic focal spatial spacing, and combine them to generate a multi-foci spatial coordinate matrix.

[0029] Optionally, the corresponding phase mask data is calculated and generated, including:

[0030] The amplitude of each focus in the multifocal spatial coordinate matrix is ​​set to a constant value to construct the target light field distribution function;

[0031] Using the target light field distribution function as input parameters, an iterative phase retrieval algorithm is executed to perform forward and inverse Fourier transform operations;

[0032] Extract the frequency domain two-dimensional phase distribution data obtained after the convergence of the iterative phase retrieval algorithm, quantize the two-dimensional phase distribution data into a gray value matrix, and generate phase mask data.

[0033] Optionally, generating the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix includes:

[0034] Obtain the material sound velocity constant of the ultrathin wafer to be tested;

[0035] Calculate the actual spatial straight-line distance between adjacent foci in the multifoci spatial coordinate matrix;

[0036] Divide the actual spatial straight-line distance by the material's sound velocity constant to calculate the basic sound wave propagation time delay. Multiply the basic sound wave propagation time delay by a preset avoidance interference coefficient to generate an excitation time delay sequence.

[0037] Optionally, generating the wafer parameter test results of the ultrathin wafer under test includes:

[0038] The trajectory curvature distribution sequence, high curvature region coordinate set, time modulation trigger sequence, multi-focus spatial coordinate matrix, phase mask data and excitation delay sequence are transformed to the wafer coordinate system corresponding to the topography mapping matrix;

[0039] Based on the irregular contour trajectory data, the trajectory curvature distribution sequence, high curvature region coordinate set, time modulation trigger sequence, multi-focus spatial coordinate matrix, phase mask data and excitation delay sequence are indexed and matched to obtain the local curvature value, high curvature mark value, trigger time parameter, focus spatial coordinate parameter, phase mask parameter and excitation delay parameter corresponding to each sampling point;

[0040] The local curvature value, high curvature mark value, trigger time parameter, focal space coordinate parameter, phase mask parameter and excitation delay parameter are written into the corresponding wafer coordinate positions in the topography mapping matrix to generate a wafer parameter mapping table.

[0041] The wafer parameter test results of the ultrathin wafer to be tested are generated based on the wafer parameter mapping table.

[0042] Secondly, this application provides an ultrathin wafer parameter testing system based on morphology detection, comprising:

[0043] The parameter testing and analysis module is used to perform wafer parameter testing on the ultrathin wafer under test, obtain the topography mapping matrix of the surface of the ultrathin wafer under test, and extract irregular contour trajectory data based on the topography mapping matrix; calculate the local curvature value of each sampling point in the irregular contour trajectory data to generate a trajectory curvature distribution sequence; extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold to obtain a set of high curvature region coordinates;

[0044] The sequence generation module is used to generate an aperiodic time jitter sequence based on the trajectory curvature distribution sequence and the high curvature region coordinate set; and to superimpose the aperiodic time jitter sequence onto the base trigger frequency sequence to obtain a time modulation trigger sequence.

[0045] The parameter calculation module is used to extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set, calculate and generate a multi-focus spatial coordinate matrix and corresponding phase mask data based on the trajectory tangent vector, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix.

[0046] The test result generation module is used to associate the topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data, and the excitation delay sequence to generate the wafer parameter test results of the ultrathin wafer under test.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This application performs wafer parameter testing on the ultrathin wafer under test and introduces wafer morphology detection in the wafer parameter testing to obtain a fine morphology mapping matrix of the ultrathin wafer surface, providing basic data for the parametric characterization of wafer surface morphology, edge contours and local abnormal regions.

[0049] This application extracts irregular contour trajectory data based on the shape mapping matrix and calculates the local curvature value of each sampling point to generate a trajectory curvature distribution sequence; further, it extracts the coordinate set of high curvature region by preset curvature threshold, realizing the coordinate identification of complex contour regions, curvature change regions and edge morphology abnormal regions of ultrathin wafers, and improving the positioning accuracy of wafer morphology parameter test results for local abnormal regions.

[0050] This application also generates aperiodic time jitter sequences, time modulation trigger sequences, multi-focus spatial coordinate matrices, phase mask data, and excitation delay sequences based on the trajectory curvature distribution sequence and the coordinate set of high curvature regions. These parameters are then unified to the wafer coordinate system corresponding to the topography mapping matrix, forming a wafer parameter mapping table. This enables comprehensive testing and unified output of ultrathin wafer topography parameters, contour curvature parameters, high curvature region parameters, and related temporal control parameters, improving the completeness, consistency, and traceability of wafer parameter testing results. Attached Figure Description

[0051] Figure 1 A flowchart illustrating a method for testing parameters of an ultrathin wafer based on morphology detection, provided as an embodiment of this application;

[0052] Figure 2 A flowchart for generating aperiodic time jitter sequences provided in this application embodiment;

[0053] Figure 3 This is a schematic diagram of an ultrathin wafer parameter testing system based on morphology detection, provided in an embodiment of this application. Detailed Implementation

[0054] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0055] See Figure 1 The diagram illustrates a method for testing ultrathin wafer parameters based on morphology detection, as provided in this application embodiment. The method includes steps S101 to S105, wherein:

[0056] S101, Perform wafer parameter testing on the ultrathin wafer to be tested, obtain the topography mapping matrix of the surface of the ultrathin wafer to be tested, and extract irregular contour trajectory data based on the topography mapping matrix;

[0057] S102, calculate the local curvature value of each sampling point in the irregular contour trajectory data, and generate a trajectory curvature distribution sequence; extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold, and obtain a set of high curvature region coordinates;

[0058] S103, Based on the trajectory curvature distribution sequence and the high curvature region coordinate set, generate an aperiodic time jitter sequence; superimpose the aperiodic time jitter sequence onto the basic trigger frequency sequence to obtain a time modulation trigger sequence;

[0059] S104, extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set, calculate and generate a multi-focus spatial coordinate matrix and corresponding phase mask data based on the trajectory tangent vector, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix;

[0060] S105, the topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data, and the excitation delay sequence are associated to generate the wafer parameter test results of the ultrathin wafer under test.

[0061] Regarding the above S101:

[0062] In practice, the first step is to perform wafer parameter testing on the ultrathin wafer under test. The thickness of the ultrathin wafer under test can range from 10μm to 50μm, or it can be a silicon wafer, silicon carbide wafer, gallium arsenide wafer, gallium nitride wafer, glass-based wafer, or other thin substrates suitable for semiconductor manufacturing and advanced packaging with a thickness of less than or equal to 50μm, depending on the actual process requirements. Wafer parameter testing is the process of measuring the actual morphology of the wafer surface to obtain height or morphology information at different planar positions on the wafer surface. This wafer parameter testing can be completed through non-contact surface height measurement methods, such as confocal displacement measurement, white light interferometry, laser triangulation, or other measurement methods that can obtain the height distribution of the wafer surface; it can also be completed using equivalent measurement methods that can output three-dimensional morphology data of the wafer surface.

[0063] During the test, using the planar coordinates of the wafer carrier platform or processing platform as a reference, multiple surface sampling points are collected within the test area of ​​the ultrathin wafer under test. Each surface sampling point includes at least its planar coordinates and the corresponding surface height value, which can be recorded as follows: ,in, , These represent the positions of the surface sampling points along the wafer plane. This indicates the height value of the wafer surface at that location.

[0064] When performing wafer parameter testing, the sampling area should at least cover the area where the subsequent irregular contour trajectory is located. To ensure that the topographic data at the trajectory boundary can be accurately interpolated, preferably, the sampling area can also extend outward by a certain width on both sides of the irregular contour trajectory to form a buffer area. This buffer area can be determined according to the wafer thickness, contour width, and path curvature, for example, it can be set to 20μm to 500μm. For irregular regions such as arcs, S-shapes, L-shapes, inner R-angles, outer R-angles, and the edges of internal voids, since these regions usually have significant curvature changes, complex edge contours, or large local topographic changes, a smaller sampling interval can be used during wafer parameter testing; for straight segments or regions with small curvature changes, a larger sampling interval can be used. For example, for regions with large curvature changes, the sampling interval can be set to 1μm to 10μm, and for regions with small curvature changes, the sampling interval can be set to 10μm to 50μm. Through the above sampling method, the topographic resolution of key areas of irregular contours can be improved without significantly increasing the testing time.

[0065] After completing wafer parameter testing, the obtained surface sampling points are organized to obtain the wafer surface topography mapping matrix. The topography mapping matrix is ​​a matrix-based data structure formed by arranging wafer surface topography data according to preset coordinate intervals, used to characterize the height distribution of the wafer surface at different planar locations. Specifically, the wafer test area can be divided into a grid, with each grid node corresponding to a matrix element. The topography mapping matrix can be represented as M, where M[m1,n1] represents the wafer surface height value at the grid node in row m1 and column n1. If an actual measured surface sampling point exists at this grid node, its height value is directly used as M[m1,n1]. If no actual measured surface sampling point exists at this grid node, the height value corresponding to this grid node is obtained by interpolation based on the height values ​​of its surrounding adjacent surface sampling points. Interpolation methods can include bilinear interpolation, bicubic interpolation, or spline interpolation.

[0066] For example, when using bilinear interpolation to generate the topography mapping matrix, if the location of the mesh node to be determined is... The four known sampling points or known grid points around it are respectively , , , ,in , Then you can first in Linear interpolation is performed in the direction, and then... Linear interpolation is performed along the direction to obtain the height value at that location. Therefore, discretely measured wafer surface height data can be converted into continuously queryable matrix data. To reduce the impact of outlier measurements on the topography mapping matrix, sampling points that significantly deviate from the neighborhood height variation trend can be removed or corrected before generating the topography mapping matrix. For example, if the difference between the height value of a surface sampling point and the average height of its neighboring sampling points exceeds a preset height anomaly threshold, then the surface sampling point is considered an anomaly, and its height value is replaced by the interpolation result of the valid neighboring sampling points. The preset height anomaly threshold can be set according to the wafer thickness and measurement accuracy, for example, it can be set to 0.1 μm to 5 μm.

[0067] After obtaining the topography mapping matrix of the wafer surface, the irregular contour trajectory data of the ultrathin wafer under test is extracted. Irregular contour trajectory data refers to the trajectory data located on the surface of the ultrathin wafer under test, used to characterize complex edge contours, internal opening boundaries, circular arc transition sections, S-shaped transition sections, L-shaped concave corners, inner R-angles, or outer R-angles, etc. It can be obtained based on the coordinate correspondence between the preset irregular contour data and the topography mapping matrix, or extracted from height abrupt boundaries, edge contour points, or local topography change areas in the topography mapping matrix. The preset irregular contour data can be provided by process files, layout files, or processing path files, and typically contains multiple two-dimensional path points or multiple geometric path segments, such as straight line segments, circular arc segments, or spline curve segments.

[0068] In practical implementation, the preset irregular contour data can first be converted to a wafer coordinate system, and the height value at the corresponding position of the data can be queried in the topography mapping matrix, thus obtaining the irregular contour trajectory data containing planar coordinates and height coordinates. Any sampling point on the irregular contour trajectory can be denoted as... ,in, , These represent the position coordinates of the trajectory sampling point in the wafer plane. This represents the wafer surface height value at that location, determined by the topography mapping matrix. By processing all trajectory sampling points sequentially in the manner described above, the irregular contour trajectory data can be obtained. ,in This represents the total number of trajectory sampling points.

[0069] To ensure the accuracy of subsequent local curvature calculations, sampling can be performed along the arc length of the irregular contour trajectory. For straight segments, the sampling interval can be relatively large; for circular arc segments, S-shaped transition segments, L-shaped concave angles, and other path segments with significant curvature changes, the sampling interval can be relatively small to avoid distortion in subsequent local curvature calculations due to insufficient sampling points. For example, the sampling interval for straight segments can be set to 5μm to 20μm, and the sampling interval for circular arc segments or areas with significant curvature changes can be set to 1μm to 5μm.

[0070] If the pre-defined irregular contour data is represented by a continuous function or spline curve, then equidistant sampling can be performed according to the path arc length. If the pre-defined irregular contour data is represented by discrete design points, then curve fitting can be performed on the discrete design points first, and then sampling can be performed along the fitted curve according to the arc length. The irregular contour trajectory data obtained in this way can fully express the geometric changes and surface morphology changes of the irregular path, providing a data foundation for subsequent local curvature calculation, high curvature region identification, and wafer parameter test result generation.

[0071] Regarding S102 above:

[0072] As an optional implementation, the local curvature values ​​of each sampling point in the irregular contour trajectory data are calculated to generate a trajectory curvature distribution sequence, including steps S201~S203, wherein:

[0073] S201, extract the coordinates of continuous sampling points in the irregular contour trajectory data and perform polynomial curve fitting to construct a smooth contour trajectory curve;

[0074] S202, calculate the first and second derivatives of the coordinates of each sampling point at the smooth contour trajectory curve, and use the first and second derivatives to calculate the local curvature value;

[0075] S203, arrange the local curvature values ​​of all sampling points in the sampling order to generate a trajectory curvature distribution sequence.

[0076] In this implementation, the irregular contour trajectory data generated in step S101 is used as the input for this step. The coordinates of continuous sampling points in the irregular contour trajectory data are extracted and subjected to polynomial curve fitting to construct a smooth contour trajectory curve. Here, the coordinates of continuous sampling points refer to a set of trajectory sampling points arranged adjacently in the sampling order in the irregular contour trajectory data, used to describe the geometric direction of a certain local trajectory segment. Polynomial curve fitting refers to using the coordinate data of several continuous sampling points to fit a continuously differentiable curve function for subsequent calculation of the first and second derivatives.

[0077] Since directly calculating curvature from discrete sampling points is easily affected by sampling interval, topographic measurement errors, and local noise points, step S102 first performs curve fitting on continuous sampling points to obtain a continuous and differentiable smooth contour trajectory curve, thereby improving the stability of curvature calculation. Polynomial curve fitting can be performed using quadratic polynomial fitting, cubic polynomial fitting, spline curve fitting, or other curve fitting methods that can obtain continuous first and second derivatives.

[0078] Specifically, for any sampling point in the irregular contour trajectory data , can be A local sampling window is formed by selecting several consecutive sampling points before and after the center. , For example, select to common 1 consecutive sampling point, of which For the window radius, This represents the sampling point where the curvature is currently to be calculated. Window radius. The window radius can be determined based on the sampling interval, the degree of change in contour curvature, and the accuracy of shape detection; for example, it can be between 2 and 10. For contour regions with large curvature changes, a smaller window radius can be used; for contour regions with gentler curvature changes, a larger window radius can be used to improve the smoothness of curve fitting.

[0079] After determining the local sampling window, to facilitate fitting, it is preferable to use the cumulative path length or local parameters of the trajectory. As the independent variable, the cumulative path length of the trajectory refers to the path distance accumulated along the trajectory direction from the starting point of the irregular contour trajectory to the current sampling point, used to represent the sequential position of the sampling point on the entire irregular contour trajectory. Specifically, it can be set as follows: For the first sampling points Calculate the cumulative path length of the trajectory ,in adjacent sampling points and Spatial distance between them; The range is 1 to When using three-dimensional coordinates, When using two-dimensional planar coordinates, .

[0080] For each sampling point within a local sampling window, the cumulative path length of the trajectory can be converted to a value centered on the sampling point. Local parameters as reference Specifically, let the sampling points contained within the local sampling window be... , This represents the index of the sampling point within the local sampling window in the global sampling sequence, and Located in Within a local sampling window centered on [the point]. Corresponding local parameters =0, any sampling point within the window The corresponding local parameters are .in, Sampling points The corresponding cumulative path length of the trajectory, Central sampling point The corresponding cumulative path length of the trajectory.

[0081] In local sampling window Within this framework, polynomial fitting is performed on the coordinate components to obtain a smooth contour trajectory curve. If three-dimensional trajectory coordinates are used, the curve can be fitted as follows: , , If two-dimensional planar trajectory coordinates are used, then it can be fitted as follows: and .in, Indicates the first sampling within the local sampling window Each sampling point relative to the center sampling point Local parameters; These are the polynomial fitting coefficients corresponding to the x-coordinate components. These are the polynomial fitting coefficients corresponding to the y-coordinate components. These are the polynomial fitting coefficients corresponding to the z-coordinate components. These fitting coefficients correspond to the constant term, first-order term, second-order term, and so on up to the m-th-order term of the respective coordinate components, and can be adjusted based on the local sampling window. Coordinates and local parameters of each sampling point within the area The order m is determined through least squares fitting. It can be set according to the complexity of the irregular contour trajectory, the sampling interval, and the noise level of the shape detection; typically, it can be between 2nd and 5th order. For local arc segments, inner R-angle segments, outer R-angle segments, or S-shaped transition segments, a 3rd order polynomial is preferred to balance fitting smoothness and curvature representation ability. For shape data with high noise, the order can be appropriately reduced or the window length increased to avoid overfitting. The polynomial coefficients can be solved using the least squares method, minimizing the sum of squared coordinate residuals between the fitted curve and each sampling point within the window.

[0082] After the above fitting process, the sampling points are constructed. The nearby smooth contour trajectory curve. The smooth trajectory curve refers to a trajectory curve expressed by a polynomial function and at least second-differentiable, used to provide a continuous function basis for subsequent calculations of the first derivative, second derivative, and local curvature values.

[0083] Next, the first and second derivatives of the coordinates of each sampling point at the smooth contour trajectory curve are calculated. The first derivative represents the direction and rate of change of the tangential aspect of the trajectory curve at the current sampling point; the second derivative represents the change in the rate of change of the tangential aspect of the trajectory curve at the current sampling point, i.e., the curvature trend of the trajectory. Specifically, for any sampling point... It is possible to calculate its corresponding local parameters. The first derivative vector at point and second derivative vector If three-dimensional trajectory coordinates are used, then , If two-dimensional planar trajectory coordinates are used, then the calculation is... and .

[0084] Then, the local curvature value is calculated using the first and second derivatives. When calculating the local curvature value using three-dimensional coordinates, the local curvature value... The following formula can be used for calculation:

[0085]

[0086] in, Indicates the first The local curvature value corresponding to each sampling point; "×" represents the vector cross product; ||·|| represents the vector magnitude. This formula characterizes the curvature of the space curve at the current sampling point and is applicable to cases where irregular paths are superimposed on the height undulations of the wafer surface.

[0087] When calculating the local curvature value using two-dimensional planar coordinates, the local curvature value The following formula can be used for calculation:

[0088]

[0089] in, and These represent the smooth trajectory curves at the current sampling point with respect to the parameters. The first derivative, and These represent the smooth trajectory curves at the current sampling point with respect to the parameters. The second derivative of .

[0090] Then, the local curvature values ​​of all sampling points are arranged in the sampling order to generate a trajectory curvature distribution sequence. The trajectory curvature distribution sequence is used to reflect the curvature change of the irregular contour trajectory data along the sampling direction, and can characterize the geometric change features of complex contour areas such as arc segments, S-shaped transition segments, L-shaped concave angles, inner R angles, outer R angles, and the edges of internal cavities.

[0091] Specifically, for trajectory sampling points to By sequentially performing the local fitting and curvature calculation described above, we obtain... , … … And form a trajectory curvature distribution sequence. Where N represents the total number of sampling points in the irregular contour trajectory data. Indicates the first The local curvature value of each sampling point.

[0092] After obtaining the trajectory curvature distribution sequence, the coordinates of sampling points whose absolute values ​​of local curvature exceed the preset curvature threshold are extracted to obtain the set of coordinates of high curvature regions.

[0093] As an optional implementation, the coordinates of sampling points whose absolute values ​​of the local curvature exceed a preset curvature threshold are extracted to obtain a set of coordinates for high curvature regions, including steps S301~S302, wherein:

[0094] S301, Calculate the absolute value of each local curvature value in the trajectory curvature distribution sequence;

[0095] S302, calculate the difference between the absolute value and the preset curvature threshold, and extract the coordinates of the sampling points with a difference greater than zero as candidate coordinate points; calculate the adjacent path distance between each candidate coordinate point, and cluster and merge the candidate coordinate points whose adjacent path distance is less than the preset merging spacing threshold to generate a set of coordinates for high curvature regions.

[0096] The preset curvature threshold is a threshold parameter used to determine whether a certain trajectory sampling point belongs to a high-risk curvature location. It can be determined based on the thickness, material type, contour design size, topography detection resolution, and parameter testing accuracy requirements of the ultrathin wafer under test. For example, for irregular contour regions with drastic curvature changes, a higher curvature threshold can be set to avoid misclassifying ordinary curved areas as high-curvature areas; for thinner ultrathin wafers with more sensitive edge contour changes, a lower curvature threshold can be set to improve the sensitivity of identifying local curvature abrupt changes.

[0097] Since curvature and radius of curvature are reciprocals of each other, in one feasible implementation, if the process considers the radius of curvature to be smaller than... If the location belongs to a high curvature region, the preset curvature threshold can be set to... .in, A preset minimum safe radius of curvature is used to represent the curvature safety boundary of the region corresponding to the irregular contour trajectory under the current wafer material, thickness, and process constraints; when the radius of curvature corresponding to a certain sampling point is smaller than this minimum safe radius of curvature, the safety boundary of curvature is determined. If the curve is too sharp, it indicates that the trajectory at that location is prone to energy superposition or thermal stress concentration, and it can be identified as a high curvature location.

[0098] In practice, the absolute values ​​of the local curvature values ​​in the trajectory curvature distribution sequence are first calculated. Specifically, for the trajectory curvature distribution sequence... For each local curvature value, calculate its absolute value to obtain a sequence of absolute curvature values. .

[0099] Next, the absolute value of each local curvature value is calculated to differ from a preset curvature threshold, and the coordinates of sampling points with a difference greater than zero are extracted as candidate coordinate points. The distance between adjacent paths of each candidate coordinate point is calculated, and candidate coordinate points with adjacent path distances less than a preset merging interval threshold are clustered and merged to generate a set of high curvature region coordinates. Specifically, the difference between the absolute value of the local curvature value and the preset curvature threshold is calculated for each sampling point. ,in This is a preset curvature threshold. When... When it is greater than 0, it indicates that the absolute value of the curvature at the sampling point exceeds the preset curvature threshold, and the coordinates of this sampling point are extracted as candidate coordinate points. The candidate coordinate points are trajectory sampling points whose single-point curvature exceeds the threshold, which may belong to the high-curvature region but have not undergone continuous region merging. All candidate coordinate points can form a candidate coordinate point set , where is the number of candidate coordinate points, denotes the th candidate coordinate point, has a value range from 1 to , corresponds to a trajectory sampling point that satisfies is greater than .

[0100] Since the high-curvature regions in the actual special-shaped contour trajectory are usually not single isolated points, but are composed of a plurality of continuous or approximately continuous sampling points with curvature exceeding the threshold, for example, an inner rounded corner region may include a plurality of adjacent candidate coordinate points, it is necessary to further calculate the adjacent path distance between each candidate coordinate point, and perform clustering and merging on candidate coordinate points with relatively close distances. The adjacent path distance is the path distance between two candidate coordinate points along the direction of the special-shaped contour trajectory; adopting the adjacent path distance can more accurately reflect the continuity of the candidate coordinate points on the actual contour trajectory.

[0101] Specifically, if two candidate coordinate points respectively correspond to the trajectory sampling points and , and a<b, the adjacent path distance between the two can be calculated as , where and are respectively the cumulative path lengths corresponding to and . If the candidate coordinate points are adjacent in the candidate sequence, the path distance between the adjacent candidate coordinate points can be directly calculated.

[0102] The preset merging distance threshold is a distance threshold for judging whether adjacent candidate coordinate points belong to the same high-curvature region. If the adjacent path distance between two adjacent candidate coordinate points is less than the preset merging distance threshold, the two candidate coordinate points are considered to belong to the same high-curvature region, and clustering and merging are performed; if the adjacent path distance is greater than or equal to the preset merging distance threshold, the two are considered to belong to different high-curvature regions. The preset merging distance threshold can be set according to the trajectory sampling interval and the laser heat-affected length, for example, it can be set to 2 to 10 times the sampling interval, or can be set to 5 μm to 50 μm.

[0103] During clustering and merging, the candidate coordinate point set C can be scanned sequentially according to the trajectory sampling order. First, the first candidate coordinate point is used as the starting point of the current high-curvature region. Then, it is determined whether the adjacent path distance between the next candidate coordinate point and the current candidate coordinate point is less than a preset merging distance threshold. If it is less, the next candidate coordinate point is merged into the current high-curvature region; if it is not less, the clustering of the current high-curvature region ends, and the next candidate coordinate point is used as the starting point of a new high-curvature region to continue scanning. After traversing all candidate coordinate points in this way, one or more high-curvature regions can be obtained. Each high-curvature region can be represented by the set of all candidate coordinate points within the region, or by parameters such as the region's starting coordinates, ending coordinates, center coordinates, maximum curvature point coordinates, and region length. Preferably, the coordinates of each candidate coordinate point within the region and its corresponding local curvature value are retained in the high-curvature region coordinate set.

[0104] The high-curvature region coordinate set is a data set consisting of one or more curvature-exceeding regions, used to represent areas in irregular contour trajectories prone to fracturing, microcracks, or thermal stress concentration. This set can be represented as... ,in The number of high curvature regions. Indicates the first A high curvature region The value range is 1 to For example, It can be represented as ,in, These are the initial candidate coordinate points for this high curvature region. To terminate candidate coordinates, to This represents all candidate coordinate points arranged in order within the high curvature region; , , They represent , , The corresponding local curvature value is used to characterize the degree of curvature at each candidate coordinate point in the high curvature region; Indicates the initial candidate coordinate point The corresponding cumulative path length, Indicates the termination candidate coordinate point The corresponding cumulative path length is used to determine the start and end positions and the length of the high curvature region in the entire irregular contour trajectory. The coordinates of the sampling point with the largest absolute value of curvature within this region are given. Using this data structure, subsequent steps can directly read the local curvature values ​​corresponding to each candidate coordinate point in the high-curvature region and generate an aperiodic time jitter sequence accordingly.

[0105] Regarding the above S103:

[0106] See Figure 2 The diagram shown is a flowchart of generating a non-periodic time jitter sequence according to an embodiment of this application, including steps S401 to S403, wherein:

[0107] S401, extract the local curvature value of each candidate coordinate point in the high curvature region coordinate set in the trajectory curvature distribution sequence as the target local curvature value;

[0108] S402, The target local curvature value is input into a preset pulse delay mapping function to calculate the pulse delay offset corresponding to each candidate coordinate point;

[0109] S403, the pulse delay offset is converted into a time interval value according to the sampling order of each candidate coordinate point to generate an aperiodic time jitter sequence.

[0110] In practical implementation, the local curvature values ​​corresponding to each candidate coordinate point in the high curvature region coordinate set are first extracted in the trajectory curvature distribution sequence and used as the target local curvature values. Specifically, for any candidate coordinate point... The first one in the corresponding irregular contour trajectory data Trajectory sampling points Then in the trajectory curvature distribution sequence Read the first Local curvature values and will Target local curvature value as candidate coordinate point , can be written as .in, Indicates the first Candidate coordinate points The corresponding target local curvature value is used to characterize the degree of trajectory curvature at the candidate coordinate point.

[0111] Next, the target local curvature value is input into a preset pulse delay mapping function to calculate the pulse delay offset corresponding to each candidate coordinate point. The pulse delay mapping function is a functional relationship used to convert the target local curvature value or curvature exceedance of candidate coordinate points into a laser pulse trigger time offset. This ensures that candidate coordinate points with larger trajectory curvature correspond to more significant trigger time disturbances, thereby reducing heat accumulation and stress superposition at that location. The pulse delay offset is the amount of time added or adjusted based on a fixed pulse time interval, and can be denoted as... . The unit can be nanosecond, microsecond, or millisecond, depending on the laser repetition rate, modulator response capability, and the range of pulse intervals allowed by the process.

[0112] As an optional implementation, the preset pulse delay mapping function can be a linear mapping function. Specifically, the curvature excess of the candidate coordinate points can be calculated first. ,in Indicates the first The degree to which the absolute value of the target local curvature of each candidate coordinate point exceeds a preset curvature threshold. Because the candidate coordinate points themselves satisfy... Greater than ,therefore Greater than zero.

[0113] The pulse delay offset is then calculated based on a linear function. ,in, This is a delay scaling factor used to convert curvature overshoot into a time offset. It can be obtained through process calibration. For example, under the same wafer material, thickness, laser power, and basic repetition frequency conditions, test tracks with different curvature radii are cut, and the delay required without edge chipping or microcracks is recorded. This delay is then fitted based on the correspondence between curvature overshoot and the delay. To avoid excessively large delay offsets leading to excessively wide local pulse spacing on the track, a minimum delay offset can also be set. and maximum delay offset and pulse delay offset Amplitude limiting is applied, that is:

[0114]

[0115] in, and It can be determined based on the laser's allowable trigger time resolution, modulator response time, material thermal diffusion time, and the minimum and maximum pulse intervals allowed by the process. For example, It can be set to the minimum time resolution at which the modulator can respond stably. It can be set to 5% to 30% of the fixed pulse time interval T_0 to ensure that the continuity of contour forming is not significantly disrupted while generating effective time disturbance.

[0116] As another possible implementation, the preset pulse delay mapping function can be a normalized mapping function to adapt to situations where the curvature range varies greatly under different irregular trajectories and different wafer batches. Specifically, for the first... High curvature region Within this region, calculate the maximum absolute value of the target local curvature value corresponding to all candidate coordinate points. and preset curvature threshold As the lower bound for normalization, calculate the... Normalized curvature risk factor corresponding to each candidate coordinate point :

[0117]

[0118] in, Used to characterize the The relative curvature risk level of each candidate coordinate point within its respective high curvature region. The larger the value, the closer the local curvature of the candidate coordinate point is to the maximum absolute value of curvature in the high curvature region; The smaller the value, the better, indicating that the curvature of the candidate coordinate point exceeds the preset curvature threshold. However, the relative risk is lower in this high curvature region. When equal If the difference between the two is too small, it can be Set it to 1 or use a preset constant to avoid the denominator being too small.

[0119] Then the pulse delay offset is calculated. In this way, the candidate coordinate point with the highest curvature in the same high curvature region corresponds to... Its pulse delay offset is close to 1. near The curvature just exceeds the preset curvature threshold. The candidate coordinates corresponding to Its pulse delay offset is close to 0. near This allows the intensity of temporal perturbations to be matched with the distribution of curvature risk within high-curvature regions.

[0120] Then, the pulse delay offset is converted into time interval values ​​according to the sampling order of each candidate coordinate point, generating an aperiodic time jitter sequence. The sampling order refers to the order in which the candidate coordinate points are located within the irregular contour trajectory data. The order of selection is as follows: Specifically, if the candidate coordinate points in the high curvature region coordinate set H are stored in the sampling order of the irregular contour trajectory data, then the order of the candidate coordinate points is directly used; if the candidate coordinate points in the high curvature region coordinate set H are not arranged in the sampling order, then they are first sorted in ascending order according to the trajectory sampling point index or cumulative path length corresponding to each candidate coordinate point to obtain a set of candidate coordinate points arranged in order.

[0121] Then each candidate coordinate point Corresponding pulse delay offset Converted to a time interval value. The pulse delay offset itself represents the additional delay relative to a fixed pulse time interval; therefore, in a preferred embodiment, it can be directly used as the time jitter corresponding to the candidate coordinate point. ,Right now The time jitter corresponding to all candidate coordinate points After arranging according to the sampling order, an aperiodic time jitter sequence is generated. .

[0122] It should be noted that the aperiodic time jitter sequence is a sequence composed of multiple time jitter values ​​that are not completely equal and do not repeat according to a fixed period, used for local time-domain modulation of the basic trigger frequency. Because the target local curvature values ​​of each candidate coordinate point are different, and the spatial position, region length, and curvature change trend of each high-curvature region in the irregular contour trajectory are different, the corresponding values ​​of different candidate coordinate points will vary. They are not usually completely identical, and their arrangement does not exhibit a fixed periodic repetition characteristic, thus forming a non-periodic time jitter sequence.

[0123] Furthermore, for coordinate sets not included in the high curvature region... For ordinary trajectory sampling points, since the absolute value of their local curvature does not exceed the preset curvature threshold, no corresponding time jitter needs to be generated for them; when the non-periodic time jitter sequence is subsequently superimposed on the base trigger frequency sequence, the time jitter corresponding to non-high curvature regions can be regarded as 0. In other words, only candidate coordinate points within high curvature regions correspond to non-zero values. The time jitter corresponding to straight sections, gentle curves, or low curvature regions is 0 or does not participate in modulation.

[0124] Thus, on the time axis of the entire irregular contour trajectory, the laser pulse triggering rhythm only undergoes a local non-periodic shift near the high curvature region, while maintaining a basic fixed pulse time interval in the non-high curvature region.

[0125] As an optional implementation, the aperiodic time jitter sequence is superimposed on the base trigger frequency sequence to obtain a time-modulated trigger sequence, including steps S501~S503, wherein:

[0126] S501, Obtain the preset fixed pulse time interval and generate the basic trigger frequency sequence;

[0127] S502, on the same time axis, the pulse delay offset in the aperiodic time jitter sequence is superimposed with the fixed pulse time interval at the corresponding moment in the basic trigger frequency sequence;

[0128] S503, the pulse timestamp sequence containing non-uniform time intervals generated after superposition is used as the time modulation trigger sequence.

[0129] In practical implementation, a preset fixed pulse time interval is first obtained to generate a basic trigger frequency sequence. The fixed pulse time interval refers to the fixed time difference between the trigger times of adjacent laser pulses when no aperiodic time jitter modulation is performed, and can be denoted as... This serves as the base pulse trigger reference for the entire irregular contour trajectory. It can be obtained by reading preset process parameters or based on the laser's fundamental repetition frequency. Specifically, during the process parameter configuration stage, the laser's basic repetition frequency can be pre-set. Basic repetition frequency This refers to the number of basic pulses output per unit time without time jitter modulation. It can be pre-calibrated based on the material type, wafer thickness, laser power, single pulse energy, trajectory speed, target width, and allowable heat-affected zone width of the ultrathin wafer under test. Fixed pulse time interval With the basic repetition frequency satisfy For example, when the basic repetition frequency At 100kHz, fixed pulse time interval 10 μs; when the fundamental repetition frequency is... At 200kHz, fixed pulse time interval The duration is 5 μs.

[0130] Obtain a fixed pulse time interval Next, the total runtime of the trajectory was further determined. The total trajectory runtime refers to the total time required for the process action point to travel from the starting point to the ending point along the irregular contour trajectory. If the total path length of the irregular contour trajectory is... Furthermore, the process application point moves at a constant speed along the entire irregular contour trajectory. The total runtime of the trajectory is calculated during the run. .in, It can be obtained by accumulating the path lengths between adjacent sampling points in the irregular contour trajectory data.

[0131] If different running speeds are used at different process application points on different path segments, the irregular contour trajectory can be divided into multiple path segments. Let the first... The length of each path segment is The corresponding running speed is The total runtime of the trajectory .

[0132] In determining and Then, a basic trigger frequency sequence is generated. This basic trigger frequency sequence is generated according to a fixed pulse time interval when no aperiodic time jitter modulation is applied. The generated time-interval triggering reference sequence can be represented as follows: ,in Based on the number of basic pulse triggers, Indicates the first A basic trigger timestamp The value range is 1 to If we take the start time of the trajectory as time 0, then we can let , .

[0133] Basic pulse trigger count Based on the total running time of the trajectory Specifically, the number of basic pulse triggers. Desirable satisfaction The largest integer plus one, that is .in, This indicates rounding down to the nearest integer.

[0134] Simultaneously with generating the basic trigger frequency sequence, a basic pulse interval sequence can also be formed. The basic pulse interval sequence is a sequence composed of the time differences between adjacent basic trigger timestamps, and can be represented as follows: , among which, the Each basic pulse interval corresponds to a basic trigger timestamp. and The time interval between Without time jitter modulation, each basic pulse interval is equal to a fixed pulse time interval. .

[0135] Next, on the same time axis, the pulse delay offset in the aperiodic time jitter sequence is superimposed with the fixed pulse time interval at the corresponding moment in the basic trigger frequency sequence. Here, "same time axis" unifies the spatial position of the candidate coordinate point on the trajectory, the elapsed time during the process action point's operation, and the basic trigger timestamp under the same time reference. The time reference is preferably set to the moment when the process action point begins to move along the irregular contour trajectory as time 0, to ensure that the time jitter corresponding to a candidate coordinate point can be applied to the pulse trigger moment when the process action point passes near that candidate coordinate point.

[0136] Specifically, for the first Candidate coordinate points First, based on its cumulative path length Cumulative path length It proceeds from the starting point of the irregular contour trajectory along the contour path to the candidate coordinate point. The length of the path traversed. If the process point operates at a constant speed... Run, then the candidate coordinate point The corresponding theory takes time It can be represented as If different speeds are used at the process application points on different path segments, then the candidate coordinate points can be determined based on the distance from the contour start point to the candidate coordinate point. The lengths and corresponding speeds of each path segment are accumulated and calculated. Let the distance from the starting point of the contour to the candidate coordinate point be... The path segments between them are several, the first... The length of each path segment is The corresponding speed is The theory then takes time. It can be represented as . This indicates that the process application point theoretically passes through the candidate coordinate point. The moment. By calculation Can candidate coordinate points Corresponding time jitter Locate the basic trigger frequency sequence The corresponding pulse interval in the sequence.

[0137] Specifically, the basic trigger frequency sequence In the middle, two adjacent basic trigger timestamps and The formation between The basic pulse intervals correspond to the time intervals as follows: For candidate coordinate points If its theory has passed through time satisfy Then the candidate coordinate point Determined to be the corresponding number The basic pulse interval, and the time jitter corresponding to the candidate coordinate point. Mapped to the One basic pulse interval.

[0138] To convert time jitter at the candidate coordinate point level to time jitter at the pulse interval level, this step defines... Indicates the first The amount of time jitter corresponding to each basic pulse interval. Used to characterize the The basic pulse interval needs to be within a fixed pulse time interval. The additional time added on top of the basic pulse interval. For each basic pulse interval, it can be determined as follows: .

[0139] If the first Each basic pulse interval does not correspond to any candidate coordinate point, meaning there are no candidate coordinate points. The theory has been passed over time fall into Then let At this point, the pulse interval is not subject to time jitter modulation. If the first... Each basic pulse interval corresponds to only one candidate coordinate point. Then let At this point, the time jitter of the basic pulse interval is directly taken from the time jitter corresponding to the candidate coordinate point. If the first... Each basic pulse interval corresponds to multiple candidate coordinate points, for example, candidate coordinate points , , ..., The theory always falls into Then, the time jitter corresponding to these multiple candidate coordinate points can be used to determine Preferred, The maximum value can be taken, that is .

[0140] After obtaining the number Time jitter corresponding to each basic pulse interval Then, it is compared with a fixed pulse time interval. Superimpose the results to obtain the first... The modulated pulse time interval .in, Indicates the first The modulated pulse time interval. If If it equals 0, then equal This indicates that the pulse interval maintains a basic fixed pulse triggering rhythm; if If greater than 0, then Greater than This indicates that the pulse interval is subjected to non-periodic delay modulation driven by the curvature of the high curvature region.

[0141] To ensure that the modulated pulse time interval meets the requirements for stable laser triggering and trajectory continuity, it is also possible to... Amplitude limiting can be performed. Specifically, a lower limit can be set for the modulated pulse time interval. and upper limit And correct it as follows :

[0142]

[0143] in, This is used to ensure that the time interval between adjacent laser pulses is not less than the minimum stable trigger interval allowed by the laser and modulator. This is used to ensure that the time interval between adjacent laser pulses does not exceed the maximum interval allowed for trajectory continuity. and It can be determined based on the laser's trigger response capability, modulator response time, trajectory running speed, material thermal diffusion characteristics, and the pulse spatial spacing allowed by the process.

[0144] Finally, the pulse timestamp sequence containing non-uniform time intervals generated after superposition is used as the time modulation trigger sequence. In a specific implementation, after processing in step S502, the basic pulse interval sequence is transformed into an equally spaced sequence. Modulated into a non-uniform pulse time interval sequence Due to the different basic pulse intervals corresponding to They may differ, and the corresponding regions are not high curvature regions. A value of 0 corresponds to the high curvature region. The pulse time interval sequence described above varies depending on the curvature risk and therefore typically contains non-uniform time intervals.

[0145] Subsequently, a pulse timestamp sequence is generated based on the modulated pulse time interval sequence. The pulse timestamp sequence is a sequence consisting of the actual trigger times of each laser pulse. Taking the trajectory start time as time 0, we can let... Then, subsequent trigger timestamps are generated sequentially using a recursive method. Thus, the time-modulated trigger sequence is obtained. .in, Indicates the time-modulated trigger sequence. Indicates the first The modulated pulse trigger timestamp The value range is 1 to .

[0146] Regarding S104 above:

[0147] As an optional implementation, a multi-focal spatial coordinate matrix is ​​calculated and generated based on the trajectory tangent vector, including steps S601~S603, wherein:

[0148] S601, extract the trajectory tangent vector of each candidate coordinate point in the coordinate set of the high curvature region, and calculate the normal offset vector that is perpendicular to the trajectory tangent vector;

[0149] S602, obtain the preset number of focal points and the basic focal space spacing;

[0150] S603, using each of the candidate coordinate points as a reference, along the direction of the normal offset vector, calculate the three-dimensional coordinate data of the remaining associated focus according to the basic focus spatial spacing, and combine them to generate a multi-focus spatial coordinate matrix.

[0151] In specific implementation, the trajectory tangent vectors of each candidate coordinate point in the coordinate set of the high curvature region are extracted, and the normal offset vector perpendicular to the trajectory tangent vector is calculated. The trajectory tangent vector is the local direction vector of the irregular contour trajectory along the trajectory's forward direction at the candidate coordinate point, used to characterize the instantaneous motion direction of the process action point when it passes through the candidate coordinate point. In step S102, a smooth trajectory curve is constructed through polynomial curve fitting, and the trajectory tangent vector can be directly calculated using the first derivative of this smooth trajectory curve at the candidate coordinate point.

[0152] Specifically, let the smoothing parametric curve of the irregular contour trajectory be represented as... ,in The parameter representing the arc length along the trajectory indicates the candidate coordinate points. The tangent vector of the trajectory at point can be represented as ,in, Indicates candidate coordinate points The corresponding trajectory parameter positions. If the irregular contour trajectory data consists of discrete sampling points, for ease of subsequent calculations, [the following is omitted as the text is incomplete and requires further context]. Normalization is performed to obtain the unit trajectory tangent vector. ,in, express The modulus can be calculated; the trajectory tangent vector can also be approximated by the difference between adjacent sampling points. For example, for non-endpoint candidate coordinate points... Its corresponding trajectory sampling points hour, And then Unitize to obtain .

[0153] The normal offset vector is a direction vector perpendicular to the trajectory tangent vector at the candidate coordinate point, used to determine the lateral unfolding direction of multiple foci relative to the centerline of the irregular contour trajectory. When the local surface of the wafer is approximately planar, the wafer surface normal or the perpendicular direction in the device coordinate system can be denoted as n_z=(0,0,1). Then, the normal offset vector perpendicular to the trajectory tangent vector and located in the tangent plane of the wafer surface is calculated using the cross product. ,in, Indicates the first The unit normal offset vector corresponding to each candidate coordinate point This indicates unitization.

[0154] If there is significant warping on the wafer surface, the local surface normal can be calculated near the candidate coordinate points based on the topography mapping matrix. For example, based on the height function Local gradient calculation Then adopt Calculate the normal offset vector that more closely resembles the actual wafer surface. This is achieved through the method described above. With the direction of trajectory Vertically, this allows subsequent multifocal arrays to be distributed laterally or nearly laterally along the slit, rather than being repeatedly superimposed along the trajectory direction.

[0155] Next, obtain the preset number of focal points and the basic focal spatial spacing; the number of focal points is the number of focal points generated in space for each candidate coordinate point, denoted as . This is used to control the lateral coverage of a multifocal beam array, ensuring that one focal point is located near the candidate coordinate point itself or the center line of the irregular contour trajectory, while the remaining focal points are symmetrically distributed along the normal offset direction. The basic focal spatial spacing is the preset spatial distance between two adjacent focal points, denoted as . It is used to control the unfolding scale of the focal array. The dimensions can be determined based on the wafer thickness, target kerf width, laser spot diameter, material thermal diffusion length, and allowable edge thermal influence range. For example, for ultrathin wafers with a thickness of 10 μm to 50 μm, It can be set to a value in the range of 1μm to 10μm; when the kerf is narrow or the material is highly thermally sensitive, a smaller value can be selected. When it is necessary to expand the energy distribution range to suppress local cracks, a larger energy distribution range can be selected. .

[0156] Then, using each candidate coordinate point as a reference, along the direction of the normal offset vector, and according to the spatial spacing of the basic focus, the three-dimensional coordinate data of the remaining associated focuses are calculated, and combined to generate a multi-focus spatial coordinate matrix. Associated focuses are multiple spatial focuses generated around the same candidate coordinate point; these focuses together constitute a local multi-focus array at that candidate coordinate point. For the first... Candidate coordinate points If the number of focal distributions If it is an odd number, we can let ,in, This represents the number of foci located on one side of the central focus. If this candidate coordinate point is taken as the central focus, then the... Spatial coordinates of each focal point It can be calculated as follows:

[0157]

[0158] in, =1, 2, ..., ; Indicates the first The candidate coordinate point corresponding to the th... The three-dimensional coordinates of each focus. This represents the unit normal offset vector at the candidate coordinate point. This represents the spatial spacing of the basic focal points. According to this formula, when... hour, That is, the central focus is located at the candidate coordinate point itself; when Less than or greater than At that time, the remaining foci are located on both sides of the candidate coordinate point and are symmetrically distributed along the normal offset direction.

[0159] If the number of focal distributions If the number is even, multiple foci can be symmetrically distributed around the candidate coordinate point without requiring any one of the foci to completely coincide with the candidate coordinate point. For example, we can let the th... The offset coefficient of each focus is ,but In this way, the even number of foci are symmetrically distributed with the candidate coordinate points as the geometric center.

[0160] To match the multifocal coordinates to the actual wafer surface, the z-values ​​in the focal coordinates can be corrected based on the topography mapping matrix. Specifically, when based on... and Calculate the plane coordinates of a certain focus Then, bilinear interpolation or spline interpolation can be performed on the plane position in the topography mapping matrix to obtain the surface height at that position. The z-coordinate of the focus is set to the surface height plus a preset focus depth offset. ,Right now:

[0161]

[0162] in, This indicates the depth offset of the laser focus relative to the wafer surface. This is useful when using surface cutting or shallow scribing. It can be 0 or close to 0; when the focus needs to penetrate slightly into the wafer interior to form a stable modified layer or cut trenches. It can be set to a negative value, and its absolute value is determined based on the wafer thickness and laser depth of focus. The multi-focus spatial coordinates generated in this way not only unfold along the trajectory normal, but also follow the warping and local height changes of the ultrathin wafer, avoiding the focus deviating from the actual processing surface due to the non-flatness of the wafer surface.

[0163] By combining the focal coordinates corresponding to all candidate coordinate points according to the candidate coordinate point number and the focal number, a multifocal spatial coordinate matrix is ​​obtained. The multifocal spatial coordinate matrix can be represented as:

[0164]

[0165] in, Let G represent the multifocal spatial coordinate matrix, where G represents the number of candidate coordinate points, and U represents the number of foci corresponding to each candidate coordinate point. Indicates the first The candidate coordinate point corresponding to the th... The three-dimensional spatial coordinates of each focal point.

[0166] As an optional implementation, the corresponding phase mask data is calculated and generated, including steps S701~S703, wherein:

[0167] S701, set the amplitude of each focus in the multi-focus spatial coordinate matrix to a constant value, and construct the target light field distribution function;

[0168] S702, Using the target light field distribution function as input parameters, perform forward and inverse Fourier transform operations using an iterative phase retrieval algorithm;

[0169] S703, extract the frequency domain two-dimensional phase distribution data obtained after the convergence of the iterative phase retrieval algorithm, quantize the two-dimensional phase distribution data into a gray value matrix, and generate phase mask data.

[0170] In practical implementation, the amplitude of each focal point in the multifocal spatial coordinate matrix is ​​set to a constant value to construct the target optical field distribution function. The target optical field distribution function is the desired optical field distribution formed on the wafer processing surface or target focal plane after phase modulation by the spatial light shaping element. To ensure a uniform energy distribution across multiple focal points, the multifocal spatial coordinate matrix can be... The target amplitude at each focus is set to the same constant A_c. For the first... The multifocal array corresponding to each candidate coordinate point can be represented by its target light field distribution function as:

[0171]

[0172] in, Indicates the first The target light field distribution function corresponding to each candidate coordinate point, g(·) represents the spatial distribution function of a single target focus, which can be a two-dimensional Gaussian function, an Airy disk approximation function or a discrete pixel point function, and A_c represents the target amplitude constant of each focus.

[0173] If a phase mask is generated simultaneously for multiple candidate coordinate points or multiple candidate coordinate points within a high curvature region, the target focal points corresponding to multiple candidate coordinate points can be superimposed to form a regional target light field distribution function:

[0174]

[0175] Where Ω represents the set of candidate coordinate points corresponding to the current phase mask, which can be a single candidate coordinate point, several candidate coordinate points within a high curvature region, or a set of candidate coordinate points corresponding to the current processing time window. In actual control, the range of candidate coordinate points corresponding to each frame of the phase mask can be determined based on the refresh rate of the spatial light shaping element and the movement speed of the process action point. If the refresh rate of the spatial light shaping element is high, the phase mask can be updated point by point; if the refresh rate is limited, phase masks can be generated in batches according to high curvature regions or time windows.

[0176] Next, the target light field distribution function is used as the input parameter, and an iterative phase retrieval algorithm is executed to perform forward and inverse Fourier transform operations. The iterative phase retrieval algorithm, given the amplitude distribution of the target focal plane, solves for the phase distribution of the input plane by repeatedly performing Fourier transforms and amplitude constraint substitutions between the input and target planes. This algorithm can employ the Gerchberg-Saxton algorithm or its equivalent variations.

[0177] Specifically, let the amplitude distribution of the incident light on the plane where the spatial light shaping element is located be... ,in This represents the planar coordinates of the spatial light shaping element; the initial phase can be set to a random phase. Alternatively, it can be set to a phase of all zeros. In the nth iteration, the complex amplitude of the input surface can be expressed as:

[0178]

[0179] in, This represents the phase value at this position during the nth iteration, where i is the imaginary unit and satisfies i 2 =-1, The complex exponential phase term can be equivalently represented as .

[0180] right Perform a Fourier transform to obtain the complex optical field of the target focal plane. , and then keep The phase is replaced with the target amplitude distribution constructed in step S701. A new complex light field of the target surface is obtained. , where angle(·) represents taking the complex phase.

[0181] Subsequently Perform an inverse Fourier transform to obtain the updated complex optical field at the input surface. , then keep The phase of the light is determined and its amplitude is replaced with the incident light amplitude of the spatial light shaping element. This yields the complex amplitude of the input surface for the next iteration. .

[0182] Repeat the above process of forward Fourier transform, target surface amplitude replacement, inverse Fourier transform, and input surface amplitude replacement until a preset convergence condition is met. The convergence condition can be either reaching a preset maximum number of iterations or the error between the reconstructed target surface amplitude and the target amplitude being less than a preset error threshold. The preset maximum number of iterations can be set to 20 to 200, and the preset error threshold can be set according to the focal uniformity requirements, for example, to 0.01 to 0.05. Through the above iterative calculations, the frequency domain or two-dimensional phase distribution of the input surface that can form a multi-focal beam array on the target processing surface can be obtained.

[0183] Then, the frequency domain two-dimensional phase distribution data obtained after the convergence of the iterative phase retrieval algorithm is extracted, and the two-dimensional phase distribution data is quantized into a grayscale matrix to generate phase mask data. The two-dimensional phase distribution data is the phase value corresponding to each pixel position of the spatial light shaping element obtained after the convergence of the iterative phase retrieval algorithm, which can be represented as: This phase value is typically within the range of 0 to 2π; if the calculated phase value exceeds this range, it can be reversed using a modulo operation, i.e. ,in, Indicates the pixel position After the modulus is taken and folded back, the actual phase value is finally limited to the range of 0 to 2π.

[0184] The grayscale matrix is ​​a digitized pixel matrix used to drive spatial light shaping elements. If the phase modulation depth of the spatial light shaping element corresponds to E-bit grayscale, for example, 8-bit grayscale, then the grayscale value range is 0 to 255. Phase values ​​can be quantized into grayscale values ​​as follows:

[0185]

[0186] in, Indicates the spatial light shaping element at the pixel position The grayscale value at the specified location, round[·] indicates rounding to the nearest integer, and E indicates the number of grayscale bits.

[0187] Corresponding to all pixel positions The resulting matrix constitutes the phase mask data. This phase mask data can be stored and loaded according to the pixel resolution, refresh format, and driving interface requirements of the spatial light shaping element. For example, when the spatial light shaping element is a phase-type spatial light modulator, the phase mask data can be loaded as a frame of two-dimensional grayscale image into the spatial light modulator; when the spatial light shaping element is a diffractive optical element or a programmable light shaper, the phase mask data can be converted into the corresponding device driving format.

[0188] As an optional implementation, generating the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix includes steps S801~S803, wherein:

[0189] S801, Obtain the material sound velocity constant of the ultrathin wafer to be tested;

[0190] S802, Calculate the actual spatial straight-line distance between adjacent foci in the multi-foci spatial coordinate matrix;

[0191] S803, divide the actual spatial straight-line distance by the material sound velocity constant to calculate the basic sound wave propagation time delay, multiply the basic sound wave propagation time delay by a preset avoidance interference coefficient, and generate an excitation time delay sequence.

[0192] In practice, the material velocity constant of the ultrathin wafer under test is first obtained. The material velocity constant, denoted as c_s, is the speed at which sound waves or mechanical disturbances propagate through the material of the wafer under test. It is used to estimate the time required for acoustic disturbances generated after excitation at adjacent focal points to propagate to the nearest focal point. c_s can be read from a material parameter database based on the wafer material type, or it can be obtained through pre-calibration. For example, for silicon wafers, the corresponding longitudinal wave velocity or equivalent velocity can be selected based on the crystal orientation and acoustic wave mode; for compound semiconductor wafers, glass-based wafers, or composite thin-film wafers, the equivalent velocity of the corresponding material system can be read. For ultrathin wafers, since they may exhibit thin-film or acoustic waveguide characteristics, c_s can also be an experimentally calibrated equivalent velocity constant. This equivalent velocity constant can be obtained by applying pulse excitation to a wafer of the same material and thickness and measuring the sound wave propagation delay.

[0193] Next, the actual spatial straight-line distance between adjacent foci in the multifocal spatial coordinate matrix is ​​calculated. The actual spatial straight-line distance is the Euclidean distance between two adjacent foci in three-dimensional space within the multifocal spatial coordinate matrix. For the ... The multifocal array corresponding to the candidate coordinate points, if the... The focus is on , No. The focus is on The actual straight-line distance between the two is... It can be represented as:

[0194]

[0195] If each focus is strictly offset along the normal vector at a fixed interval If the distribution is such that the z-direction changes little, then... Approximately equal to However, when there is warping on the wafer surface or changes in focal depth, using the three-dimensional actual spatial straight-line distance can more accurately reflect the acoustic propagation path between adjacent focal points.

[0196] Then, the actual linear distance in space is divided by the material's sound velocity constant to calculate the basic sound wave propagation time delay. This basic sound wave propagation time delay is then multiplied by a preset avoidance interference coefficient to generate an excitation time delay sequence. The basic sound wave propagation time delay is the theoretical time required for an acoustic disturbance to propagate from one focal position to an adjacent focal position, denoted as […]. , can be represented as ,in, Indicates the first The candidate coordinate point corresponding to the th... The focus and the first The actual linear distance in space between the focal points, where c_s represents the material's sound velocity constant.

[0197] The interference avoidance coefficient is a process coefficient used to adjust the degree of temporal misalignment between adjacent focal points, denoted as . This coefficient is used to ensure that the excitation time difference between adjacent focal points avoids the time window for synchronous superposition of acoustic impacts. It can be determined based on the material's acoustic attenuation characteristics, wafer thickness, laser single-pulse energy, focal spacing, and trial cutting results. Generally, It can be set to a value greater than 0, for example, 0.5 to 3. When When the value is 1, it means that the delay between adjacent focal points is approximately equal to the time it takes for the sound wave to travel between the two focal points; when... When the value is greater than 1, it indicates a further increase in the excitation time difference between focal points to enhance the effect of avoiding acoustic interference; when... When the value is less than 1, it indicates that a weaker peak shifting is performed while ensuring processing efficiency.

[0198] Excitation delay between adjacent focal points It can be calculated as follows:

[0199]

[0200] in, Indicates the first The candidate coordinate point corresponding to the th... The focus and the first The excitation delay between each focus. By combining the excitation delays of adjacent focuses corresponding to the same candidate coordinate point according to the focus arrangement order, the local excitation delay subsequence corresponding to that candidate coordinate point can be obtained. , .

[0201] Furthermore, by combining the local excitation delay subsequences corresponding to all candidate coordinate points, the excitation delay sequence can be obtained. , where Λ represents the excitation delay sequence between each focus in the multifocal spatial coordinate matrix.

[0202] If subsequent time delay lines require the absolute delay of each focus relative to the first focus as control input, the excitation delays between adjacent focuses can be accumulated to obtain the relative excitation time of each focus. For example, for the first... A multi-focus array corresponding to candidate coordinate points allows the relative excitation time of the first focus to be... and in accordance with The method of calculation yields the relative excitation time sequence. .

[0203] Regarding the above S105:

[0204] In practical implementation, the wafer coordinate system corresponding to the topography mapping matrix is ​​first determined. This wafer coordinate system can be the wafer plane coordinate system used when generating the topography mapping matrix, or it can be a coordinate system determined by the wafer carrier platform, wafer alignment marks, or a preset process reference. Each matrix element in the topography mapping matrix corresponds to a planar position on the wafer surface and stores the surface height value at that position. Therefore, in the subsequent parameter association process, the wafer coordinate system corresponding to this topography mapping matrix is ​​used as a unified coordinate reference, and the trajectory curvature distribution sequence, high curvature region coordinate set, time modulation trigger sequence, multi-focus spatial coordinate matrix, phase mask data, and excitation delay sequence are all converted or matched to this wafer coordinate system.

[0205] When unifying coordinates, for irregular contour trajectory data already generated based on the topography mapping matrix, their corresponding wafer coordinates can be directly used. For parameters generated in the local trajectory coordinate system, device motion coordinate system, focal array coordinate system, or spatial light shaping coordinate system, conversion is performed according to a pre-established coordinate correspondence. This coordinate correspondence can be determined by the correspondence between the wafer alignment mark, the origin of the device motion platform, the starting point of the irregular contour trajectory, the calibration position of the spatial light shaping element, and the focal imaging position. During conversion, the data in the local coordinate system can be converted to the data in the wafer coordinate system based on the translation, rotation, and scale correction relationships between the coordinate systems. For the three-dimensional focal coordinates in the multi-focal spatial coordinate matrix, the coordinate relationship in the height direction or the focal depth direction also needs to be considered to ensure that the focal coordinates are consistent with the height position of the wafer surface.

[0206] After coordinate unification, various parameters are indexed and matched based on the irregular contour trajectory data. The irregular contour trajectory data consists of multiple sampling points arranged sequentially along the trajectory direction. Each sampling point has corresponding wafer plane coordinates and surface height values. Since the trajectory curvature distribution sequence was generated in the previous steps according to the sampling point order, each local curvature value can be directly associated with its corresponding sampling point according to the arrangement order of the sampling points. That is, the first sampling point corresponds to the first local curvature value, the second sampling point corresponds to the second local curvature value, and so on, thus obtaining the local curvature parameter corresponding to each sampling point.

[0207] For a set of coordinates representing high-curvature regions, it needs to be converted into high-curvature labels for each sampling point. Specifically, the coordinates or regions within the high-curvature region coordinate set are matched with the sampling points of the irregular contour trajectory. If a sampling point belongs to the high-curvature region coordinate set, or is located within the clustered high-curvature region range, it is marked as a high-curvature sampling point; if a sampling point does not belong to any high-curvature region, it is marked as a normal sampling point. Furthermore, different levels of labeling can be assigned to high-curvature sampling points based on the degree to which their local curvature value exceeds a preset curvature threshold, to distinguish regions with mild, severe, or significantly abnormal curvature abrupt changes. In this way, the wafer parameter test results can clearly display which locations belong to high-curvature regions and the corresponding degree of curvature anomaly.

[0208] For time-modulated trigger sequences, it is necessary to establish a correspondence between trigger time parameters and sampling points of the irregular contour trajectory. In practice, firstly, the cumulative path length along the trajectory direction for each sampling point is calculated based on the irregular contour trajectory data. Then, the trajectory position corresponding to each trigger pulse is determined based on the trajectory running speed and trigger timestamp. When the trajectory running speed is constant, the theoretical position of the trigger pulse on the trajectory can be determined based on the trigger time and running speed. Then, the sampling point closest to this theoretical position is found, and the trigger time parameter is written into the parameter record corresponding to that sampling point. When the trajectory running speed varies with trajectory segments, a correspondence table between the cumulative path length of the trajectory and the running time can be established first. Then, the corresponding trajectory position is found based on the trigger time, and the trigger time parameter is matched to the corresponding sampling point. If multiple trigger pulses correspond to a sampling point, the multiple trigger times can be written into the trigger time parameter of that sampling point in chronological order. If multiple sampling points correspond to the same trigger pulse, their correspondence can be determined based on the nearest distance principle or the trajectory interval assignment principle.

[0209] For a multifocal spatial coordinate matrix, it needs to be associated with a high-curvature region or high-curvature sampling points. Since the multifocal spatial coordinate matrix is ​​typically generated based on the coordinate set of high-curvature regions and the corresponding trajectory tangent vectors, for sampling points belonging to high-curvature regions, their corresponding multifocal spatial coordinates are written into the parameter record of that sampling point; for sampling points not belonging to high-curvature regions, the focal spatial coordinate parameters can be set to null, invalid, or a default flag. If a high-curvature region contains multiple consecutive sampling points, and this high-curvature region corresponds to a set of multifocal spatial coordinates, then this set of focal spatial coordinates can be associated with all sampling points within that high-curvature region. To improve the association accuracy, the focal spatial coordinates can also be locally corrected based on the local trajectory direction and normal direction of each sampling point within the high-curvature region, ensuring that the focal distribution always remains consistent with the contour direction near the sampling point.

[0210] For phase mask data, it is necessary to establish a correlation between it and the corresponding multifocal spatial coordinate matrix. Since the phase mask data is generated based on the multifocal spatial coordinate matrix, each set of multifocal spatial coordinates should correspond to a set of phase mask data or a phase mask number. When writing to the wafer parameter mapping table, it is not necessary to repeatedly write the complete phase mask grayscale matrix to every sampling point. Instead, a unique number is assigned to the phase mask data, and this phase mask number is written in the parameter record of the corresponding sampling point or high curvature region. This phase mask number can further correspond to the grayscale value matrix, quantization bits, matrix size, applicable focus group, and generation parameters of the phase mask. If multiple high curvature regions use the same focus distribution method, they can also share the same set of phase mask data. In this case, the same phase mask number can be written at different sampling points or different high curvature regions. If the focus distribution direction or position varies for different high curvature regions, the corresponding spatial transformation parameters can be recorded simultaneously so that the corresponding focus modulation method can be restored later based on the phase mask number and spatial transformation relationship.

[0211] For the excitation delay sequence, it needs to be matched with the focal relationships in the multifocal spatial coordinate matrix. The excitation delay sequence is typically generated based on the actual spatial distance between adjacent focal points, the material acoustic velocity constant of the ultrathin wafer under test, and the avoidance interference coefficient. Therefore, it should be written together with the corresponding focal group into the parameter record of the high-curvature sampling point or high-curvature region. In practice, for sampling points belonging to high-curvature regions, the excitation delay sequence of the corresponding focal group is written into its parameter record; for sampling points not belonging to high-curvature regions, the excitation delay parameter can be set to a null value or an invalid flag. If multiple sampling points within a high-curvature region share the same set of focal points and the same set of excitation delay sequences, the same excitation delay sequence number can be written into the records of each sampling point within that region. This method ensures a consistent correspondence between the focal spatial coordinates, phase mask data, and excitation delay sequence.

[0212] After completing the index matching above, the local curvature value, high curvature marker value, trigger time parameter, focal space coordinate parameter, phase mask parameter, and excitation delay parameter corresponding to each sampling point are written into the corresponding wafer coordinate position in the topography mapping matrix to generate a wafer parameter mapping table. In specific implementation, firstly, the corresponding matrix position of the sampling point in the topography mapping matrix is ​​determined based on the wafer plane coordinates of the sampling point. Then, the surface height value at that matrix position is read, and the various parameters corresponding to that sampling point are written into the extended data unit corresponding to that matrix position. This extended data unit includes at least the surface height value, local curvature value, high curvature marker value, trigger time parameter, focal space coordinate parameter, phase mask parameter, and excitation delay parameter. For matrix positions that do not fall on the irregular contour trajectory, only the surface height value can be retained, and other parameters can be set to null values ​​or background markers; for matrix positions that fall on the irregular contour trajectory, the complete parameters corresponding to the sampling point of that trajectory are written.

[0213] When multiple trajectory sampling points correspond to the same topography mapping matrix position, they can be processed according to a preset merging rule. Specifically, the sampling point with the largest absolute value of local curvature can be selected as the representative sampling point of the matrix position, and the parameters of the sampling point can be written into the matrix position; alternatively, a parameter list of multiple sampling points can be stored under the matrix position, so that the same matrix position can correspond to multiple trajectory parameter records. For high curvature regions, it is preferable to retain the parameters of the sampling point with the largest absolute value of curvature to highlight the maximum curvature change at that position; for trigger time parameters, trigger times falling within the neighborhood range corresponding to the matrix position can be listed in order; for focal space coordinate parameters, phase mask parameters, and excitation delay parameters, parameter number references can be used for storage to avoid repeatedly writing a large amount of the same data.

[0214] The wafer parameter mapping table can adopt a matrix structure consistent with the topography mapping matrix, or a table structure with the sampling points of the irregular contour trajectory as the primary key. When using a matrix structure, each matrix position corresponds to a coordinate position on the wafer surface, and the original surface height value is supplemented with corresponding curvature parameters, high curvature marker parameters, trigger timing parameters, focal space parameters, phase mask parameters, and excitation delay parameters. When using a sampling point table structure, each sampling point of the irregular contour trajectory corresponds to a record, which includes the sampling point number, wafer coordinates, surface height value, cumulative path length, local curvature value, high curvature marker value, trigger timing parameter, focal space coordinate parameter, phase mask parameter, and excitation delay parameter. The matrix structure facilitates querying and displaying parameter distribution within the wafer plane, while the sampling point table structure facilitates tracing parameter changes of each sampling point sequentially along the irregular contour trajectory. Both can be generated simultaneously to meet different result output requirements.

[0215] When generating wafer parameter test results for the ultrathin wafer under test based on the wafer parameter mapping table, the output can include at least the following parameters: topography parameters, trajectory curvature parameters, high curvature region identification results, trigger timing parameters, multifocal spatial parameters, phase mask parameters, and excitation delay parameters. Specifically, the topography parameters characterize the height distribution on the wafer surface; the trajectory curvature parameters characterize the local curvature changes at each sampling point on the irregular contour trajectory; the high curvature region identification results provide the location, range, and degree of curvature anomaly of the high curvature region; the trigger timing parameters provide the modulation trigger time corresponding to each trajectory position; the multifocal spatial parameters provide the focal spatial distribution corresponding to the high curvature region; the phase mask parameters provide the phase modulation data required to generate the corresponding multifocal optical field; and the excitation delay parameters provide the excitation timing relationship between the multifoci. These results comprehensively reflect the topography, curvature characteristics, high curvature locations, and related temporal control parameters of the irregular contour region of the ultrathin wafer under test, all within the same wafer coordinate reference.

[0216] Before generating the final wafer parameter test results, consistency verification of the parameter association results can be performed. Specifically, this involves checking whether the number of curvature values ​​in the trajectory curvature distribution sequence matches the number of sampling points on the irregular contour trajectory; checking whether the coordinates in the high curvature region coordinate set can all be mapped to the wafer coordinate range corresponding to the topography mapping matrix; checking whether the trigger time in the time modulation trigger sequence corresponds to the effective path range of the irregular contour trajectory; checking whether the multi-focal spatial coordinate matrices have all undergone wafer coordinate transformation; checking whether the phase mask data has a clear association with the corresponding focal spatial coordinate parameters; and checking whether the excitation delay sequence is consistent with the number of focal points and the relationship between focal points in the corresponding focal group. If a sampling point or a high curvature region has missing parameters, out-of-bounds coordinates, or mismatched indexes, an anomaly flag is written in the wafer parameter mapping table, and the corresponding pending confirmation information is output in the wafer parameter test results. This verification process can improve the completeness and reliability of the wafer parameter test results.

[0217] Therefore, by uniformly transforming the trajectory curvature distribution sequence, high curvature region coordinate set, time-modulated trigger sequence, multi-focal spatial coordinate matrix, phase mask data, and excitation delay sequence to the wafer coordinate system corresponding to the topography mapping matrix, and performing sample-point index matching based on the irregular contour trajectory data, the local curvature value, high curvature marker value, trigger time parameter, focal spatial coordinate parameter, phase mask parameter, and excitation delay parameter corresponding to each sampling point can be obtained. These parameters are then written into the corresponding wafer coordinate positions in the topography mapping matrix to generate a wafer parameter mapping table, and the wafer parameter test results of the ultrathin wafer under test are generated based on this table. This wafer parameter test result enables unified correlation output between topography data, curvature data, high curvature region data, trigger timing data, focal spatial data, phase mask data, and excitation delay data. Based on this result, the parameter distribution and local anomaly locations of the irregular contour region of the ultrathin wafer under test can be directly determined.

[0218] Based on the same inventive concept, this application also provides an ultrathin wafer parameter testing system based on topography detection, corresponding to a method for testing ultrathin wafer parameters based on topography detection. Since the principle of the system in this application is similar to the above-mentioned method for testing ultrathin wafer parameters based on topography detection in this application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0219] See Figure 3 The diagram shown is a schematic of an ultrathin wafer parameter testing system based on topography detection provided in an embodiment of this application, including:

[0220] The parameter testing and analysis module 10 is used to perform wafer parameter testing on the ultrathin wafer under test, obtain the topography mapping matrix of the surface of the ultrathin wafer under test, extract irregular contour trajectory data based on the topography mapping matrix, calculate the local curvature value of each sampling point in the irregular contour trajectory data, generate a trajectory curvature distribution sequence, and extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold to obtain a set of high curvature region coordinates.

[0221] The sequence generation module 20 is used to generate an aperiodic time jitter sequence based on the trajectory curvature distribution sequence and the high curvature region coordinate set; and to superimpose the aperiodic time jitter sequence onto the basic trigger frequency sequence to obtain a time modulation trigger sequence.

[0222] The parameter calculation module 30 is used to extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set, calculate and generate a multi-focus spatial coordinate matrix and corresponding phase mask data based on the trajectory tangent vector, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix.

[0223] The test result generation module 40 is used to associate the topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data and the excitation delay sequence to generate the wafer parameter test results of the ultrathin wafer under test.

[0224] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for testing parameters of ultrathin wafers based on morphology detection, characterized in that, include: Perform wafer parameter testing on the ultrathin wafer under test, obtain the topography mapping matrix of the surface of the ultrathin wafer under test, and extract the irregular contour trajectory data based on the topography mapping matrix; Calculate the local curvature value of each sampling point in the irregular contour trajectory data to generate a trajectory curvature distribution sequence; extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold to obtain a set of high curvature region coordinates. Based on the trajectory curvature distribution sequence and the set of coordinates of the high curvature region, an aperiodic time jitter sequence is generated; The aperiodic time jitter sequence is superimposed on the base trigger frequency sequence to obtain the time-modulated trigger sequence; Extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set. Based on the trajectory tangent vector, calculate and generate a multi-focus spatial coordinate matrix and the corresponding phase mask data, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix. The topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data, and the excitation delay sequence are correlated to generate the wafer parameter test results of the ultrathin wafer under test.

2. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 1, characterized in that, The step of calculating the local curvature values ​​of each sampling point in the irregular contour trajectory data and generating a trajectory curvature distribution sequence includes: Extract the coordinates of continuous sampling points from the irregular contour trajectory data and perform polynomial curve fitting to construct a smooth contour trajectory curve. Calculate the first and second derivatives of the coordinates of each sampling point at the smooth contour trajectory curve, and use the first and second derivatives to calculate the local curvature value; Arrange the local curvature values ​​of all sampling points in the sampling order to generate a trajectory curvature distribution sequence.

3. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 1, characterized in that, The step of extracting the coordinates of sampling points whose absolute values ​​of the local curvature exceed a preset curvature threshold yields a set of coordinates for high curvature regions, including: Calculate the absolute value of each local curvature value in the trajectory curvature distribution sequence; The absolute value is calculated by comparing it with a preset curvature threshold, and the coordinates of the sampling points with a difference greater than zero are extracted as candidate coordinate points. The distance between adjacent paths between each candidate coordinate point is calculated, and the candidate coordinate points with adjacent path distances less than a preset merging spacing threshold are clustered and merged to generate a set of coordinates for high curvature regions.

4. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 3, characterized in that, The generation of aperiodic time jitter sequences based on the trajectory curvature distribution sequence and the set of coordinates of the high curvature region includes: Extract the local curvature value of each candidate coordinate point in the high curvature region coordinate set in the trajectory curvature distribution sequence, and use it as the target local curvature value; The target local curvature value is input into a preset pulse delay mapping function to calculate the pulse delay offset corresponding to each candidate coordinate point; The pulse delay offset is converted into a time interval value according to the sampling order of each candidate coordinate point to generate an aperiodic time jitter sequence.

5. The method for testing parameters of an ultrathin wafer based on morphology detection according to claim 4, characterized in that, The step of superimposing the aperiodic time jitter sequence onto the base trigger frequency sequence to obtain the time-modulated trigger sequence includes: Obtain a preset fixed pulse time interval and generate a basic trigger frequency sequence; On the same time axis, the pulse delay offset in the aperiodic time jitter sequence is superimposed with the fixed pulse time interval at the corresponding moment in the basic trigger frequency sequence; The pulse timestamp sequence containing non-uniform time intervals generated after superposition is used as the time modulation trigger sequence.

6. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 3, characterized in that, The step of calculating and generating a multi-focal spatial coordinate matrix based on the trajectory tangent vector includes: Extract the trajectory tangent vector of each candidate coordinate point in the coordinate set of the high curvature region, and calculate the normal offset vector that is perpendicular to the trajectory tangent vector; Obtain the preset number of focal points and the basic focal space spacing; Using each candidate coordinate point as a reference, along the direction of the normal offset vector, calculate the three-dimensional coordinate data of the remaining related foci according to the basic focal spatial spacing, and combine them to generate a multi-foci spatial coordinate matrix.

7. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 1, characterized in that, Calculate and generate the corresponding phase mask data, including: The amplitude of each focus in the multifocal spatial coordinate matrix is ​​set to a constant value to construct the target light field distribution function; Using the target light field distribution function as input parameters, an iterative phase retrieval algorithm is executed to perform forward and inverse Fourier transform operations; Extract the frequency domain two-dimensional phase distribution data obtained after the convergence of the iterative phase retrieval algorithm, quantize the two-dimensional phase distribution data into a gray value matrix, and generate phase mask data.

8. The method for testing parameters of ultrathin wafers based on morphology detection according to claim 1, characterized in that, The generation of the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix includes: Obtain the material sound velocity constant of the ultrathin wafer to be tested; Calculate the actual spatial straight-line distance between adjacent foci in the multifoci spatial coordinate matrix; Divide the actual spatial straight-line distance by the material's sound velocity constant to calculate the basic sound wave propagation time delay. Multiply the basic sound wave propagation time delay by a preset avoidance interference coefficient to generate an excitation time delay sequence.

9. The method for testing parameters of an ultrathin wafer based on morphology detection according to claim 1, characterized in that, The wafer parameter test results generated for the ultrathin wafer under test include: The trajectory curvature distribution sequence, high curvature region coordinate set, time modulation trigger sequence, multi-focus spatial coordinate matrix, phase mask data and excitation delay sequence are transformed to the wafer coordinate system corresponding to the topography mapping matrix; Based on the irregular contour trajectory data, the trajectory curvature distribution sequence, high curvature region coordinate set, time modulation trigger sequence, multi-focus spatial coordinate matrix, phase mask data and excitation delay sequence are indexed and matched to obtain the local curvature value, high curvature mark value, trigger time parameter, focus spatial coordinate parameter, phase mask parameter and excitation delay parameter corresponding to each sampling point; The local curvature value, high curvature mark value, trigger time parameter, focal space coordinate parameter, phase mask parameter and excitation delay parameter are written into the corresponding wafer coordinate positions in the topography mapping matrix to generate a wafer parameter mapping table. The wafer parameter test results of the ultrathin wafer to be tested are generated based on the wafer parameter mapping table.

10. A morphology-based ultrathin wafer parameter testing system, used to implement the morphology-based ultrathin wafer parameter testing method according to any one of claims 1 to 9, characterized in that, include: The parameter testing and analysis module is used to perform wafer parameter testing on the ultrathin wafer under test, obtain the topography mapping matrix of the surface of the ultrathin wafer under test, and extract irregular contour trajectory data based on the topography mapping matrix. Calculate the local curvature value of each sampling point in the irregular contour trajectory data to generate a trajectory curvature distribution sequence; extract the coordinates of sampling points whose absolute value of the local curvature value exceeds a preset curvature threshold to obtain a set of high curvature region coordinates. The sequence generation module is used to generate an aperiodic time jitter sequence based on the trajectory curvature distribution sequence and the set of coordinates of the high curvature region; The aperiodic time jitter sequence is superimposed on the base trigger frequency sequence to obtain the time-modulated trigger sequence; The parameter calculation module is used to extract the trajectory tangent vector of each coordinate point in the high curvature region coordinate set, calculate and generate a multi-focus spatial coordinate matrix and corresponding phase mask data based on the trajectory tangent vector, and generate the excitation delay sequence between each focus in the multi-focus spatial coordinate matrix. The test result generation module is used to associate the topography mapping matrix, the trajectory curvature distribution sequence, the high curvature region coordinate set, the time modulation trigger sequence, the multi-focus spatial coordinate matrix, the phase mask data, and the excitation delay sequence to generate the wafer parameter test results of the ultrathin wafer under test.

Images