Multi-station collaborative operation method and system for wafer femtosecond laser cutting

Abstract

This invention discloses a multi-station collaborative operation method and system for femtosecond laser dicing of wafers, relating to the field of wafer dicing technology. The method includes: acquiring the dicing path to be diced on the wafer, determining the total modified layer depth mapped to points on the dicing path; interactively acquiring wafer material data, configuring the crack initiation control parameters for the first station using the wafer material data and the total modified layer depth, and performing laser dicing using the configuration results; performing birefringence stress imaging of the crack initiation layer to establish a stress vector diagram of the crack initiation layer; synchronizing the stress vector diagram of the crack initiation layer and the total modified layer depth to the second station, and performing extended control optimization for the second station; and performing collaborative dicing operations for the second station using the extended control optimization results. This solves the technical problem in the prior art where the lack of collaborative control between crack initiation and subsequent extended processing during femtosecond laser dicing of wafers leads to unstable dicing quality, achieving the technical effect of improving wafer dicing quality and dicing stability.

Description

Technical Field

[0001] This invention relates to the field of wafer dicing technology, and more specifically to a multi-station collaborative operation method and system for femtosecond laser dicing of wafers. Background Technology

[0002] As semiconductor manufacturing processes continue to advance towards higher integration, smaller size, and ultra-thinness, wafer dicing, as a critical step in chip manufacturing, directly impacts chip edge integrity, mechanical strength, and subsequent packaging yield. Femtosecond lasers, with their ultra-short pulse width, small heat-affected zone, and high processing precision, are widely used in the precision dicing of wafers made of hard and brittle materials such as silicon, silicon carbide, and sapphire. Current femtosecond laser wafer dicing typically employs a combination of crack initiation and propagation processes to achieve crack formation and propagation. Crack initiation is responsible for forming the initial modified layer or crack source, while propagation is responsible for propagating the crack along a predetermined dicing path to achieve wafer separation. However, in existing technologies, crack initiation and propagation are mostly performed independently according to preset process parameters. Subsequent propagation processing is usually executed based on preset parameters, failing to dynamically adjust the propagation parameters based on the actual stress distribution formed after crack initiation. Furthermore, there is a lack of effective information exchange and collaborative control between different workstations. When the properties of the wafer material, the depth of the modified layer, or the local stress state change, it can easily cause the crack propagation direction to shift, the propagation consistency to decrease, and the quality of the cut surface to fluctuate, thereby affecting the wafer cutting quality and processing stability. Summary of the Invention

[0003] This application provides a multi-station collaborative operation method and system for femtosecond laser cutting of wafers, which solves the technical problem of unstable cutting quality caused by the lack of coordinated control between crack initiation and subsequent expansion processing in the existing femtosecond laser cutting process.

[0004] A first aspect of this application provides a multi-station collaborative operation method for femtosecond laser dicing of wafers, the method comprising:

[0005] The process involves: acquiring the cutting path on the wafer and determining the total modified layer depth mapped to points along the cutting path; interactively acquiring wafer material data; configuring the crack initiation control parameters for the first station using the wafer material data and the total modified layer depth; and performing laser cutting using the configuration results, wherein the first station is the crack initiation layer processing station; performing birefringence stress imaging of the crack initiation layer to establish a crack initiation layer stress vector diagram; synchronizing the crack initiation layer stress vector diagram and the total modified layer depth to the second station and performing extended control optimization for the second station; and performing collaborative cutting operations for the second station using the extended control optimization results, wherein the second station is the extended layer processing station.

[0006] A second aspect of this application provides a multi-station collaborative operation system for femtosecond laser dicing of wafers, the system comprising:

[0007] Cutting Determination Module: Acquires the cutting path to be cut on the wafer and determines the total modified layer depth mapped to the points on the cutting path; First Processing Module: After interactively acquiring wafer material data, it uses the wafer material data and the total modified layer depth to execute the crack initiation control parameter configuration of the first station, and uses the configuration result to execute laser cutting, wherein the first station is the crack initiation layer processing station; Imaging Module: Performs birefringence stress imaging of the crack initiation layer and establishes a crack initiation layer stress vector diagram; Optimization Module: Synchronizes the crack initiation layer stress vector diagram and the total modified layer depth to the second station and executes the extended control optimization of the second station; Second Processing Module: Executes the collaborative cutting operation of the second station using the extended control optimization result, wherein the second station is the extended layer processing station.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, the dicing path on the wafer to be diced is acquired, and the total modified layer depth mapped to the points on the dicing path is determined. After interactively acquiring the wafer material data, the crack initiation control parameters for the first station are configured using the wafer material data and the total modified layer depth. Laser dicing is then performed using the configuration results. Here, the first station is the crack initiation layer processing station. Next, birefringence stress imaging of the crack initiation layer is performed to establish a stress vector diagram of the crack initiation layer. Then, the stress vector diagram of the crack initiation layer and the total modified layer depth are synchronized to the second station, and expansion control optimization is performed at the second station. Finally, the collaborative dicing operation at the second station is performed using the expansion control optimization results. Here, the second station is the expansion layer processing station. This solves the technical problem in the prior art where the lack of coordinated control between crack initiation processing and subsequent expansion processing during wafer femtosecond laser dicing leads to unstable dicing quality, achieving the technical effect of improving wafer dicing quality and dicing stability. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a multi-station collaborative operation method for femtosecond laser cutting of wafers provided in an embodiment of this application;

[0012] Figure 2 This is a schematic diagram of the stress vector distribution of the crack initiation layer provided in an embodiment of this application;

[0013] Figure 3 This is a schematic diagram of a multi-station collaborative operation system for femtosecond laser cutting of wafers, provided in an embodiment of this application.

[0014] Explanation of reference numerals in the attached drawings: Cutting determination module 11, first processing module 12, imaging module 13, optimization module 14, second processing module 15. Detailed Implementation

[0015] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0016] Example 1, as Figure 1 As shown, this application provides a multi-station collaborative operation method for femtosecond laser cutting of wafers, wherein the method includes:

[0017] Obtain the cutting path to be cut on the wafer, and determine the total modified layer depth of the point mapping on the cutting path.

[0018] Specifically, the system uses the center line of the dicing track in the wafer layout file as the source of the dicing path to be cut. It reads the two-dimensional coordinate sequence of the dicing track center line in the layout coordinate system, and uses the wafer center positioning point as the origin and the wafer notch direction as the angular reference to transform the two-dimensional coordinate sequence into the wafer processing coordinate system, obtaining the dicing path to be cut on the wafer. Then, it performs equidistant sampling of the dicing path according to a preset point spacing, generating a path point sequence continuously arranged according to the processing direction. Each path point has a point number, planar coordinates, and adjacent point numbers. After determining the path points, the wafer thickness measurement module measures the local wafer thickness at each corresponding position along the path point sequence, obtaining the local wafer thickness value for each path point. It also retrieves preset values ​​for the surface protective layer thickness, bottom retention layer thickness, and crack propagation safety factor. Full margin; for any path point, subtract the surface protective layer thickness and the bottom retention layer thickness from the local wafer thickness value of the path point to obtain the target effective segmentation depth of the path point. Then, add the crack propagation safety margin to the target effective segmentation depth to obtain the total modified layer depth mapped to the path point. The total modified layer depth represents the cumulative internal modified depth that needs to be formed by the first station crack initiation layer processing and the second station expansion layer processing at the path point. The system traverses all path points in the above manner and establishes a point depth mapping table consisting of path point number, path point coordinates, local wafer thickness value, surface protective layer thickness, bottom retention layer thickness, crack propagation safety margin, and total modified layer depth. The point depth mapping table is used as the data input for subsequent first station crack initiation control parameter configuration and second station expansion control optimization.

[0019] After interactively acquiring wafer material data, the first station is configured with the wafer material data and the total modified layer depth to perform crack initiation control parameter configuration. The configuration result is then used to perform laser cutting. The first station is the crack initiation layer processing station.

[0020] The system establishes an interactive connection between the first-station control terminal and the process database. After the wafer to be processed is loaded and identified, the system reads the corresponding wafer material data according to the wafer batch number. This wafer material data includes at least the wafer material's bandgap, thermal conductivity, crystal cleavage plane orientation, and known crystal orientation distribution. Simultaneously, the system reads the previously established dicing path point depth mapping table. Then, using each path point on the dicing path as a configuration unit, and combining it with the total modified layer depth of the corresponding point, the system configures the crack initiation control parameters point-by-point. Specifically, the energy range for nonlinear absorption of the femtosecond laser within the wafer is determined based on the wafer material's bandgap; the allowable heat accumulation degree and repetition frequency of adjacent pulses are determined based on the wafer material's thermal conductivity; the angle between the laser scanning direction and the crystal orientation is determined based on the crystal cleavage plane orientation and known crystal orientation distribution; and the focusing position is determined based on the total modified layer depth of the corresponding path point. The target depth of the initiation layer is determined, and based on this, the pulse energy, pulse width, repetition frequency, and focusing depth corresponding to each path point are jointly generated to establish a set of initiation control parameters corresponding to the cutting path. Then, the femtosecond laser processing equipment is controlled to sequentially perform initiation layer processing according to the cutting path point sequence. During the processing, the motion control platform drives the laser focus to move continuously along the cutting path according to the preset scanning speed, and automatically calls the corresponding initiation control parameters according to the current processing point, adjusting the pulse energy, pulse width, repetition frequency, and focusing depth of the femtosecond laser in real time, so that the laser focus is always located at the target initiation layer position of the corresponding path point, continuously forming a modified region distributed along the cutting path inside the wafer. After the initiation processing of all path points is completed, the initiation layer processing result of the first station is formed, and the corresponding initiation control parameters and processing records are saved synchronously for subsequent initiation layer stress imaging and second station expansion control optimization.

[0021] Furthermore, the wafer material data includes at least the wafer material's bandgap, thermal conductivity, crystal cleavage plane orientation, and known crystal orientation distribution.

[0022] The bandgap width is used to characterize the energy threshold characteristics of nonlinear multiphoton absorption and electronic excitation of wafer materials under femtosecond laser irradiation, and serves as a basic constraint parameter for determining the pulse energy range in crack initiation control; the thermal conductivity is used to characterize the thermal diffusion capability of the wafer during local laser heating, and is used to limit the pulse repetition frequency and upper limit of heat accumulation per unit path point to avoid excessive expansion of the local heat-affected zone; the crystal cleavage plane orientation is used to characterize the preferential propagation direction of natural fractures inside the wafer material, and is used to determine the angle relationship between the laser scanning direction and the crystal structure to guide the initiation crack to propagate stably along the preset cutting path; the known crystal orientation distribution is used to characterize the spatial distribution characteristics of the lattice orientation inside the wafer, and is used to differentiate the focusing depth, energy deposition density, and uniformity of the modified layer formation at different path points.

[0023] Furthermore, the crack initiation control parameters include pulse energy, pulse width, repetition frequency, and depth of focus.

[0024] The pulse energy is used to control the energy input intensity of the femtosecond laser to generate a nonlinear absorption effect inside the wafer, and is set in stages according to the bandgap of the wafer material and the total depth of the modified layer to ensure the formation of a stable local modified region within the target depth range. The pulse width is used to control the time-spreading characteristics of a single laser pulse. By adjusting the pulse width, the instantaneous peak power is affected, thereby controlling the free electron excitation rate and the initial nucleation density of microcracks inside the wafer. The repetition frequency is used to control the number of laser pulses emitted per unit time, and, combined with the wafer's thermal conductivity, to constrain the thermal accumulation effect to avoid excessive thermal superposition between adjacent pulses, which could lead to uncontrolled expansion of the modified region. The focusing depth is used to control the spatial positioning of the laser focus in the wafer thickness direction, and, based on the total depth of the modified layer, the focus is aligned point by point with the target depth range of the initiation layer to ensure that the modified region is continuously formed within the preset depth range.

[0025] Furthermore, performing laser cutting using the configuration results also includes:

[0026] The cutting results of the initiation layer are monitored, and a monitoring dataset is established; the monitoring dataset is evaluated for smoothing of adjacent path points, and smoothing anomaly feedback is established; the control parameter configuration compensation process of the first station is performed according to the smoothing anomaly feedback.

[0027] After the crack initiation process is completed at the first workstation, synchronous data acquisition is performed using a coaxial microscopic imaging unit positioned above the wafer processing area and a beam-splitting interference detection unit positioned in the processing optical path. The coaxial microscopic imaging unit acquires visible light microscopic image sequences of the cutting area at a preset frame rate, while the beam-splitting interference detection unit acquires phase difference interference signals at corresponding positions. Both data streams are synchronously triggered at the nanosecond level by a clock trigger module synchronized with the femtosecond laser drive signal, ensuring a one-to-one time-space correspondence between the acquired data and the cutting path points. A crack initiation layer cutting result monitoring dataset is constructed based on the path point numbers. This monitoring dataset stores data including crack boundary coordinates, grayscale distribution of the modified region, interference phase difference, and local deformation displacement field data, indexed by the point location. Based on this, structured numerical processing is performed on the monitoring dataset. Specifically, the Sobel gradient operator is used to extract the crack edge coordinate sequence from the microscopic images, and the crack propagation continuity is calculated using least-squares linear fitting. The local variance statistical method is used to calculate the uniformity index of the modified region for the grayscale distribution data. The digital image correlation method (DIP) is used for the displacement field data. IC) calculates subpixel-level displacement vectors and obtains edge fracture amplitude; performs phase expansion and inversion calculations on interferometric phase difference data to obtain equivalent quality improvement depth deviation values, and constructs the above four types of indicators into path point mass vectors according to a unified normalization method; further, constructs a sliding calculation window of length N=3 for the path point mass vectors according to the cutting path order, performs Euclidean distance difference calculation on adjacent point mass vectors in each window, and simultaneously calculates the second-order difference to characterize local change acceleration, introduces path curvature κ as a weighting coefficient, and performs Euclidean distance difference calculation on straight segments and corner segments. Different weighting functions are used for weighted fusion to obtain a local smoothing evaluation value. When the evaluation value exceeds the threshold (μ+λσ, where μ is the mean, σ is the standard deviation, and λ is the empirical coefficient) obtained based on the historical processing statistical distribution, the center point in the corresponding window is marked as an anomaly point, and an anomaly index set is generated. Based on the anomaly index set, the crack initiation control parameter vector corresponding to the first station is updated in reverse. The quality deviation vector of the anomaly point is mapped to the control parameter gradient matrix. The control parameter correction amount is obtained by calculating the Jacobian inversion increment Δu = −α·J⁻¹·e (where e is the quality deviation vector and α is the step size coefficient), and is applied to the pulse energy and repetition frequency parameters respectively. At the same time, a neighborhood constraint smoothing strategy is adopted for the pulse width and focusing depth. That is, the parameters of one path point before and after the anomaly point are reconstructed by weighted mean, so that the updated parameter sequence satisfies the first-order continuity constraint, thereby realizing the closed-loop compensation update of the crack initiation control parameters of the first station.

[0028] Perform birefringence stress imaging of the initiation layer to create a stress vector diagram of the initiation layer.

[0029] Specifically, a polarized light stress detection device is used to perform point-by-point scanning inspection on the wafer after the first-stage crack initiation process. The device incident linearly polarized light on the wafer surface along a preset cutting path and collects transmitted light intensity signals in an analysis direction fixed orthogonal to the incident polarization direction. During the scanning process, optical response data at each path point is acquired sequentially at a fixed spatial sampling interval along the cutting path, and the corresponding phase delay is calculated based on the light intensity change at each path point. Furthermore, the phase delay is bound one-to-one with the path point coordinates to obtain the local stress response value corresponding to each path point. The local stress response value is used to characterize the stress concentration caused by laser modification inside the initiation layer. Then, according to the spatial order of the cutting path, the local stress response values ​​of adjacent path points are continuously interpolated to obtain stress response curves continuously distributed along the cutting path. On this basis, the local stress response values ​​of each path point are used as vector amplitudes, and the gradient of the stress response value in the tangent direction of the path is used as the vector direction component to construct a stress vector field distribution that strictly corresponds to the cutting path. This stress vector field is then organized into an initiation layer stress vector diagram according to the path point order to characterize the stress distribution and crack propagation direction characteristics inside the initiation layer.

[0030] like Figure 2 The diagram shows the stress vector distribution of the initiation layer provided in an embodiment of this application. During the femtosecond laser cutting of a wafer, spatial sampling is performed on the wafer surface along a preset cutting path to obtain stress response data at discrete points along the path. The stress vector distribution at each point is then reconstructed based on a birefringence imaging method. The horizontal axis represents the spatial position on the wafer cutting path, and the vertical axis represents the spatial geometric offset distribution of the cutting path; the curve characterizes the spatial trajectory of the cutting path. At each sampling point, the direction and magnitude of the principal stress corresponding to that point are represented by a vector arrow, thus forming a stress vector field distributed along the cutting path.

[0031] The stress vector diagram of the initiation layer and the total modified layer depth are synchronized to the second station, and the extended control optimization of the second station is performed.

[0032] Furthermore, the stress vector diagram of the initiation layer and the total modified layer depth are synchronized to the second station, and extended control optimization is performed at the second station, including:

[0033] The actual crack initiation depth data of the initiation layer is obtained. Based on the actual crack initiation depth data and the total modified layer depth, a cutting depth target mapped to the cutting path points is established. The second station traverses the cutting path point by point, reads the stress amplitude and stress azimuth angle of the current point and adjacent points in the stress vector diagram of the initiation layer, and obtains the cleavage plane orientation of the wafer. The stress azimuth angle and cleavage plane orientation are compared. If the angle comparison result is less than a preset angle threshold, the current point is determined to be a cleavage plane sensitive point. The cutting depth target, stress amplitude, and cleavage plane sensitivity determination result are sent to the pulse energy prediction channel to establish the initial pulse energy parameters of the point. The stress amplitude gradient between the current point and adjacent points is extracted. When the stress amplitude gradient exceeds a preset gradient threshold, an energy compensation amount opposite to the gradient direction is applied to the initial pulse energy parameters of the current point, and the initial pulse energy parameters of the adjacent points are simultaneously corrected in the opposite direction to establish the extended control optimization result.

[0034] First, the actual crack initiation depth data after crack initiation processing at the first station is obtained. The actual crack initiation depth data is then compared point-by-point with the total modified layer depth at the corresponding path points to calculate the cutting depth deviation for each path point. Based on this cutting depth deviation, a cutting depth target sequence corresponding one-to-one with each cutting path point is constructed. Second, the second station traverses the cutting path points sequentially. For the current path point, the stress amplitude and stress azimuth angle of that point and its adjacent points are extracted from the crack initiation layer stress vector diagram. Simultaneously, the cleavage plane orientation parameters of the wafer material are acquired. Furthermore, the angle between the stress azimuth of the current path point and the azimuth of the wafer cleavage surface is calculated. When the angle between the stress azimuth and the cleavage surface is less than a preset angle threshold, the current path point is identified as a sensitive point on the cleavage surface, and a corresponding sensitivity weight coefficient is assigned to the path point. The sensitivity weight coefficient is determined based on the degree of angle deviation. Specifically, using the angle threshold as a normalization benchmark, the angle value of the current path point is compared with the preset angle threshold to obtain an angle normalization factor. The angle normalization factor is then multiplied with the stress amplitude to generate the sensitivity weight coefficient of the current path point. The sensitivity weight coefficient is increased when the angle is closer to zero and gradually decreased when the angle is closer to the threshold, thereby achieving a continuous quantitative expression of the sensitive area of ​​the cleavage surface. Subsequently, the target cutting depth, stress amplitude, and weight coefficients of sensitive points on the cleavage surface are input into the pulse energy prediction channel to generate the initial pulse energy parameters of the current path point and establish the initial energy distribution sequence corresponding to the cutting path point. Based on this, the stress amplitude gradient between the current path point and adjacent path points is calculated. When the stress amplitude gradient exceeds a preset gradient threshold, an energy compensation amount opposite to the gradient change direction is applied to the initial pulse energy parameters of the current path point, and the initial pulse energy parameters corresponding to adjacent path points are synchronously corrected to maintain the continuity constraint of energy distribution in space, thereby forming an extended control optimization result that satisfies the consistency of stress continuity and cutting depth.

[0035] Furthermore, the sensitive points on the cleavage surface are subject to an upper limit constraint on pulse energy, which is lower than that of ordinary points.

[0036] For the cleavage-sensitive points, during the second-station expansion control optimization process, the corresponding pulse energy parameters are subject to an energy constraint upper limit lower than that of conventional path points. This energy constraint upper limit is dynamically determined based on the stress concentration at the corresponding point in the wafer material's cleavage orientation and stress vector diagram, and is adjusted downwards from the conventional point pulse energy upper limit. Specifically, when a path point is identified as a cleavage-sensitive point, the system sets the allowed upper limit of pulse energy for that point to a preset reduction factor corresponding to the conventional point pulse energy upper limit. This reduction factor is then further adjusted based on the stress amplitude at that point, ensuring a lower energy input level during the expansion cutting process. This reduces the probability of uncontrolled crack propagation along the crystal plane and ensures stable crack extension along the preset cutting path.

[0037] Furthermore, the cutting depth target, stress amplitude, and cleavage surface sensitivity determination results are sent to the pulse energy prediction channel to establish the initial pulse energy parameters for the point, including:

[0038] Based on the configuration results, a predetermined processing target for the first workstation is established; a deviation analysis is performed between the predetermined processing target and the actual crack initiation depth data to establish a deviation dataset; a source tracing analysis is performed on the deviation dataset to trace the influence of wafer material data to establish a source tracing influence feedback; the source tracing influence feedback is used as an additional item to perform an influence feedback update of the initial pulse energy parameters.

[0039] Specifically, a predetermined processing target sequence is established based on the crack initiation control parameter configuration results of the first workstation. This predetermined processing target sequence is then matched point-by-point with the actual crack initiation depth data collected at the first workstation. By calculating the difference between the target depth and the actual crack initiation depth at each path point, a corresponding deviation data sequence is formed. This deviation data sequence is then organized according to the cutting path order to construct a deviation dataset. Further, a wafer material data correlation analysis is performed on the deviation dataset, mapping the deviation data to the bandgap width, thermal conductivity, and crystal orientation distribution in the wafer material data to identify the material sensitivity caused by the deviation. The source is identified, and the propagation trend of the deviation is extracted based on the continuous distribution characteristics of the deviation along the spatial path, forming a source-tracing influence feedback. The source-tracing influence feedback is used to characterize the degree of influence of material properties on the local crack initiation depth offset. The source-tracing influence feedback information is introduced as an additional correction term into the initial pulse energy calculation process. Under the combined effect of the cutting depth target, stress amplitude, and cleavage surface sensitivity judgment results, the initial pulse energy parameters are dynamically updated so that the pulse energy parameters of each path point simultaneously satisfy the depth constraint, stress constraint, and material source-tracing correction constraint, thereby generating an initial pulse energy parameter set that corresponds one-to-one with the cutting path points.

[0040] Furthermore, a source analysis of the impact of wafer material data on the aforementioned deviation dataset is performed, and a source impact feedback mechanism is established, including:

[0041] Obtain the depth deviation data sequence distributed along the cutting path in the deviation dataset; perform path direction correlation analysis on the depth deviation data sequence to identify continuous deviation segments; extract the deviation propagation direction and deviation accumulation trend through the continuous deviation segments; perform source tracing analysis using the deviation propagation direction and deviation accumulation trend to establish source impact feedback.

[0042] First, a depth deviation data sequence arranged sequentially along the cutting path is obtained from the deviation dataset, where each data point corresponds to the initiation depth deviation value of a path point. This sequence is then reconstructed temporally according to the path point numbers to form a one-dimensional ordered depth deviation sequence. Second, the path direction consistency of the depth deviation data sequence is detected. By calculating the same-direction index and amplitude change rate of deviation changes between adjacent path points, the continuity of the deviation in the path direction is determined. When the deviation change direction of multiple consecutive path points remains consistent and the amplitude change rate is lower than a preset fluctuation threshold, the region is divided into a continuous deviation segment. Further, within the continuous deviation segment, spatial evolution analysis is performed on the deviation data, calculating the local derivative of the deviation with path position, and extracting the deviation's... The propagation direction characteristics along the path are analyzed, and a deviation accumulation trend curve is obtained by fitting a function of deviation variation with path length. The deviation accumulation trend is used to characterize the enhancement or attenuation characteristics of deviation during spatial propagation. Finally, the deviation propagation direction is compared with the crystal orientation distribution and cleavage plane orientation in the wafer material data. The deviation accumulation trend is correlated with the thermal conductivity and bandgap parameters of the wafer material. When the deviation propagation direction and crystal orientation distribution show consistent expansion and the deviation accumulation trend shows monotonically changing, the deviation is determined to originate from the influence of lattice anisotropy. When the deviation propagation direction and crystal orientation distribution do not show significant consistency and the deviation shows fluctuating diffusion characteristics, the deviation is determined to originate from the influence of material doping or structural inhomogeneity, and source-tracing feedback information is generated accordingly.

[0043] Furthermore, source tracing analysis is performed using the direction of deviation propagation and the trend of deviation accumulation to establish a source tracing impact feedback mechanism, including:

[0044] The deviation propagation direction of the continuous deviation segment is compared with the known crystal orientation distribution of the wafer material data to determine whether the deviation propagation direction extends linearly along a specific crystal orientation. When the angle between the deviation propagation direction and the specific crystal orientation is less than a preset crystal orientation deviation threshold, and the deviation accumulation trend is monotonically increasing or decreasing within the continuous deviation segment, it is determined that there is a material lattice distortion anomaly in the corresponding region of the continuous deviation segment, and the slope of the deviation accumulation trend is used as the lattice distortion gradient quantification index. When the deviation propagation direction is unrelated to the specific crystal orientation and is distributed in a scattering manner, and the deviation accumulation trend is non-monotonic fluctuation within the continuous deviation segment, it is determined that there is a material doping concentration anomaly in the corresponding region of the continuous deviation segment, and the fluctuation amplitude of the deviation accumulation trend is used as the doping uniformity control index. The lattice distortion gradient quantification index or the doping uniformity control index is used as the source influence feedback output.

[0045] First, the deviation propagation direction of the continuous deviation segment is matched and compared one by one with the known crystal orientation distribution in the wafer material data to calculate the angle between the deviation propagation direction and each crystal orientation. Based on the minimum angle principle, the target crystal orientation is determined to determine whether the deviation propagation direction extends linearly along a specific crystal orientation. When the angle between the deviation propagation direction and the target crystal orientation is less than a preset crystal orientation deviation threshold, and the deviation accumulation trend shows a monotonically increasing or decreasing change within the continuous deviation segment, it is determined that there is an abnormal lattice structure distortion in the corresponding region of the continuous deviation segment. Furthermore, a linear fitting is performed on the deviation accumulation trend to obtain the fitted straight line. The slope value of the line is used as a lattice distortion gradient quantization index. When the deviation propagation direction has no consistent correspondence with the wafer crystal orientation distribution, and the deviation accumulation trend exhibits non-monotonic fluctuation within the continuous deviation segment, the standard deviation or peak-valley difference value of the deviation accumulation trend is calculated, and the standard deviation or peak-valley difference value is used as a doping uniformity control index to characterize the uniformity of the material doping distribution. Finally, the lattice distortion gradient quantization index or doping uniformity control index is used as a source influence feedback output and bound to the corresponding path point for storage, so as to be called for subsequent pulse energy parameter updates and second station extended control optimization.

[0046] The collaborative cutting operation of the second station is executed using the optimization results of the extended control. The second station is the extended layer processing station.

[0047] Specifically, the optimization result of the extended control is parsed into a set of processing control parameters for the second station, which corresponds one-to-one with the cutting path points. The set of processing control parameters includes at least pulse energy parameters, scanning speed parameters, and focus position adjustment parameters. During execution, the second station processes the extended layer point by point along the preset cutting path of the wafer, and calls the corresponding processing control parameters according to the path point sequence, so that the femtosecond laser continuously acts inside the wafer extended layer according to the preset depth trajectory, thereby achieving directional propagation and penetration of cracks based on the crack initiation layer already formed in the first station. At the same time, the second station receives path point constraint information from the stress vector diagram in real time during the processing, and dynamically fine-tunes the local processing parameters to ensure that the crack propagation direction is consistent with the cutting path and to suppress unexpected deflection propagation, thereby completing the collaborative cutting operation of the extended layer and achieving stable separation of the wafer along the cutting path.

[0048] For example, using a silicon-based wafer with a thickness of 725 μm as the processing object, a preset cutting path extends along the wafer diameter direction, with a total path length of 200 mm. Discrete sampling is performed at a point spacing of 2 μm to obtain a total of 100,000 path points. Based on the wafer edge positioning results and thickness distribution measurement results, the target cutting depth is set to 710 μm, and the safety margin is set to 5 μm, thus obtaining a dynamic distribution of the total modified layer depth corresponding to each path point between 705 μm and 710 μm. During the crack initiation process at the first station, the femtosecond laser pulse energy is dynamically adjusted within the range of 200 nJ to 450 nJ, the pulse width is set to 220 fs to 340 fs, the repetition frequency is set to 120 kHz to 480 kHz, and the focusing depth is corrected point by point within the range of ±3 μm of the target crack initiation layer depth. This process forms a continuous modified initiation layer inside the wafer. During birefringence stress imaging, a 1μm scanning interval is used to acquire phase delay data of the initiation layer, with a phase delay range of 0 to 75nm. The corresponding converted stress amplitude range is 0MPa to 120MPa, the stress vector direction resolution is 1°, and the stress gradient threshold is set to 15MPa / μm to construct a continuous stress vector field. During the second station expansion control optimization process, a pulse energy reduction coefficient of 0.65 to 0.85 is applied to the sensitive points on the cleavage surface, with a corresponding included angle threshold set to 12°. The continuous deviation judgment window is 3 to 5 path points. When the standard deviation of the stress amplitude fluctuation is greater than 8MPa, energy compensation correction is triggered, with a compensation ratio of 10% to 25%. Ultimately, the stable connection of the expansion layer cutting path and the controllable separation of the wafer along the preset path are achieved.

[0049] In summary, the embodiments of this application have at least the following technical effects:

[0050] First, the dicing path on the wafer to be diced is acquired, and the total modified layer depth mapped to the points on the dicing path is determined. After interactively acquiring the wafer material data, the crack initiation control parameters for the first station are configured using the wafer material data and the total modified layer depth. Laser dicing is then performed using the configuration results. Here, the first station is the crack initiation layer processing station. Next, birefringence stress imaging of the crack initiation layer is performed to establish a stress vector diagram of the crack initiation layer. Then, the stress vector diagram of the crack initiation layer and the total modified layer depth are synchronized to the second station, and expansion control optimization is performed at the second station. Finally, the collaborative dicing operation at the second station is performed using the expansion control optimization results. Here, the second station is the expansion layer processing station. This solves the technical problem in the prior art where the lack of coordinated control between crack initiation processing and subsequent expansion processing during wafer femtosecond laser dicing leads to unstable dicing quality, achieving the technical effect of improving wafer dicing quality and dicing stability.

[0051] Example 2, based on the same inventive concept as the multi-station collaborative operation method for femtosecond laser cutting of wafers in the foregoing examples, such as... Figure 3 As shown, this application provides a multi-station collaborative operation system for femtosecond laser cutting of wafers, wherein the system includes:

[0052] Cutting determination module 11: Obtains the cutting path to be cut on the wafer and determines the total modified layer depth mapped on the cutting path; First processing module 12: After interactively obtaining wafer material data, it uses the wafer material data and the total modified layer depth to execute the crack initiation control parameter configuration of the first station, and uses the configuration result to execute laser cutting, wherein the first station is the crack initiation layer processing station; Imaging module 13: Performs birefringence stress imaging of the crack initiation layer and establishes a crack initiation layer stress vector diagram; Optimization module 14: Synchronizes the crack initiation layer stress vector diagram and the total modified layer depth to the second station and executes the extended control optimization of the second station; Second processing module 15: Executes the collaborative cutting operation of the second station using the extended control optimization result, wherein the second station is the extended layer processing station.

[0053] Furthermore, the optimization module 14 is used to perform the following method:

[0054] The actual crack initiation depth data of the initiation layer is obtained. Based on the actual crack initiation depth data and the total modified layer depth, a cutting depth target mapped to the cutting path points is established. The second station traverses the cutting path point by point, reads the stress amplitude and stress azimuth angle of the current point and adjacent points in the stress vector diagram of the initiation layer, and obtains the cleavage plane orientation of the wafer. The stress azimuth angle and cleavage plane orientation are compared. If the angle comparison result is less than a preset angle threshold, the current point is determined to be a cleavage plane sensitive point. The cutting depth target, stress amplitude, and cleavage plane sensitivity determination result are sent to the pulse energy prediction channel to establish the initial pulse energy parameters of the point. The stress amplitude gradient between the current point and adjacent points is extracted. When the stress amplitude gradient exceeds a preset gradient threshold, an energy compensation amount opposite to the gradient direction is applied to the initial pulse energy parameters of the current point, and the initial pulse energy parameters of the adjacent points are simultaneously corrected in the opposite direction to establish the extended control optimization result.

[0055] Furthermore, the optimization module 14 is used to perform the following method:

[0056] The sensitive points on the cleavage surface are subject to an upper limit constraint on pulse energy, which is lower than that of normal points.

[0057] Furthermore, the optimization module 14 is used to perform the following method:

[0058] Based on the configuration results, a predetermined processing target for the first workstation is established; a deviation analysis is performed between the predetermined processing target and the actual crack initiation depth data to establish a deviation dataset; a source tracing analysis is performed on the deviation dataset to trace the influence of wafer material data to establish a source tracing influence feedback; the source tracing influence feedback is used as an additional item to perform an influence feedback update of the initial pulse energy parameters.

[0059] Furthermore, the optimization module 14 is used to perform the following method:

[0060] Obtain the depth deviation data sequence distributed along the cutting path in the deviation dataset; perform path direction correlation analysis on the depth deviation data sequence to identify continuous deviation segments; extract the deviation propagation direction and deviation accumulation trend through the continuous deviation segments; perform source tracing analysis using the deviation propagation direction and deviation accumulation trend to establish source impact feedback.

[0061] Furthermore, the optimization module 14 is used to perform the following method:

[0062] The deviation propagation direction of the continuous deviation segment is compared with the known crystal orientation distribution of the wafer material data to determine whether the deviation propagation direction extends linearly along a specific crystal orientation. When the angle between the deviation propagation direction and the specific crystal orientation is less than a preset crystal orientation deviation threshold, and the deviation accumulation trend is monotonically increasing or decreasing within the continuous deviation segment, it is determined that there is a material lattice distortion anomaly in the corresponding region of the continuous deviation segment, and the slope of the deviation accumulation trend is used as the lattice distortion gradient quantification index. When the deviation propagation direction is unrelated to the specific crystal orientation and is distributed in a scattering manner, and the deviation accumulation trend is non-monotonic fluctuation within the continuous deviation segment, it is determined that there is a material doping concentration anomaly in the corresponding region of the continuous deviation segment, and the fluctuation amplitude of the deviation accumulation trend is used as the doping uniformity control index. The lattice distortion gradient quantification index or the doping uniformity control index is used as the source influence feedback output.

[0063] Furthermore, the first processing module 12 is used to perform the following methods:

[0064] The wafer material data includes at least the wafer material's bandgap, thermal conductivity, crystal cleavage plane orientation, and known crystal orientation distribution.

[0065] Furthermore, the first processing module 12 is used to perform the following methods:

[0066] The crack initiation control parameters include pulse energy, pulse width, repetition frequency, and focusing depth.

[0067] Furthermore, the first processing module 12 is used to perform the following methods:

[0068] The cutting results of the initiation layer are monitored, and a monitoring dataset is established; the monitoring dataset is evaluated for smoothing of adjacent path points, and smoothing anomaly feedback is established; the control parameter configuration compensation process of the first station is performed according to the smoothing anomaly feedback.

[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-station collaborative operation method for femtosecond laser cutting of wafers, characterized in that, The method includes: Obtain the cutting path to be cut on the wafer, and determine the total modified layer depth of the point mapping on the cutting path; After interactively acquiring wafer material data, the first station is configured with the wafer material data and the total modified layer depth to perform crack initiation control parameter configuration. The configuration result is then used to perform laser cutting. The first station is the crack initiation layer processing station. Perform birefringence stress imaging of the initiation layer and establish a stress vector diagram of the initiation layer; The stress vector diagram of the crack initiation layer and the total modified layer depth are synchronized to the second station, and the extended control optimization of the second station is performed. The collaborative cutting operation of the second station is executed using the optimization results of the extended control. The second station is the extended layer processing station.

2. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 1, characterized in that, The stress vector diagram of the initiation layer and the total modified layer depth are synchronized to the second station, and the extended control optimization of the second station is performed, including: Obtain the actual initiation depth data of the initiation layer, and establish a cutting depth target that maps to the cutting path points based on the actual initiation depth data and the total modified layer depth; The second station traverses the cutting path point by point, reads the stress amplitude and stress azimuth angle of the current point and adjacent points in the stress vector diagram of the initiation layer, and obtains the orientation of the cleavage plane of the wafer. The stress azimuth angle and the cleavage plane azimuth are compared. If the angle comparison result is less than the preset angle threshold, the current point is determined to be a cleavage plane sensitive point. The cutting depth target, stress amplitude, and cleavage surface sensitivity determination results are sent to the pulse energy prediction channel to establish the initial pulse energy parameters of the point. Extract the stress amplitude gradient between the current point and adjacent points. When the stress amplitude gradient exceeds a preset gradient threshold, apply an energy compensation amount opposite to the gradient direction to the initial pulse energy parameter of the current point, and simultaneously correct the initial pulse energy parameters of the adjacent points to establish an extended control optimization result.

3. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 2, characterized in that, The sensitive points on the cleavage surface are subject to an upper limit constraint on pulse energy, which is lower than that of normal points.

4. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 2, characterized in that, The cutting depth target, stress amplitude, and cleavage surface sensitivity determination results are sent to the pulse energy prediction channel to establish the initial pulse energy parameters for the point, including: Based on the configuration results, establish the predetermined processing target for the first workstation; A deviation analysis was performed on the predetermined processing target and the actual crack initiation depth data to establish a deviation dataset; A source analysis of the impact of wafer material data on the aforementioned deviation dataset is conducted, and a source impact feedback mechanism is established. The source tracing impact feedback is added as an additional item to perform the impact feedback update of the initial pulse energy parameters.

5. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 4, characterized in that, A source analysis of the impact of wafer material data on the aforementioned deviation dataset is performed, and a source impact feedback mechanism is established, including: Obtain the depth deviation data sequence distributed along the cutting path in the deviation dataset; Perform path direction correlation analysis on the depth deviation data sequence to identify continuous deviation segments; The deviation propagation direction and deviation accumulation trend are extracted from the continuous deviation segments. Utilize the direction of deviation propagation and the trend of deviation accumulation to perform source tracing analysis and establish a source tracing impact feedback mechanism.

6. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 5, characterized in that, Utilizing the direction of deviation propagation and the trend of deviation accumulation, conduct source tracing analysis and establish feedback mechanisms to trace the impact of the source tracing, including: The deviation propagation direction of the continuous deviation section is compared with the known crystal orientation distribution of the wafer material data to determine whether the deviation propagation direction extends linearly along a specific crystal orientation. When the angle between the deviation propagation direction and a specific crystal orientation is less than the preset crystal orientation deviation threshold, and the deviation accumulation trend is monotonically increasing or decreasing in the continuous deviation section, it is determined that there is a material lattice distortion anomaly in the region corresponding to the continuous deviation section, and the slope of the deviation accumulation trend is used as the lattice distortion gradient quantification index. When the deviation propagation direction is independent of a specific crystal orientation and is distributed in a scattering manner, and the deviation accumulation trend is non-monotonic fluctuation within the continuous deviation section, it is determined that there is an abnormal material doping concentration in the region corresponding to the continuous deviation section, and the fluctuation amplitude of the deviation accumulation trend is used as the doping uniformity control index. The lattice distortion gradient quantization index or the doping uniformity quality control index is used as the source influence feedback output.

7. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 1, characterized in that, The wafer material data includes at least the wafer material's bandgap, thermal conductivity, crystal cleavage plane orientation, and known crystal orientation distribution.

8. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 1, characterized in that, The crack initiation control parameters include pulse energy, pulse width, repetition frequency, and focusing depth.

9. The multi-station collaborative operation method for femtosecond laser cutting of wafers as described in claim 1, characterized in that, Performing laser cutting using the configuration results also includes: Monitor the cutting results of the initiation layer and establish a monitoring dataset; A smooth evaluation of adjacent path points is performed on the monitoring dataset to establish smooth anomaly feedback; Based on the smooth anomaly feedback, the control parameter configuration compensation process for the first station is executed.

10. A multi-station collaborative operation system for femtosecond laser cutting of wafers, characterized in that, For implementing the multi-station collaborative operation method for femtosecond laser cutting of wafers according to any one of claims 1-9, the system comprises: Cutting determination module: Obtain the cutting path to be cut on the wafer, and determine the total modified layer depth of the points mapped on the cutting path; First processing module: After interactively acquiring wafer material data, the module uses the wafer material data and total modified layer depth to execute the crack initiation control parameter configuration of the first station, and uses the configuration result to execute laser cutting. The first station is the crack initiation layer processing station. Imaging module: Performs birefringence stress imaging of the initiation layer and establishes a stress vector diagram of the initiation layer; Optimization module: Synchronizes the stress vector diagram of the crack initiation layer and the total modified layer depth to the second station, and performs extended control optimization in the second station; The second processing module: Utilizes the optimization results of extended control to execute the collaborative cutting operation of the second station, where the second station is the extended layer processing station.

Images