Gallium nitride wafer edge collapse control dicing method and system

Abstract

This invention relates to the field of wafer dicing technology, and more particularly to a method and system for controlling edge chipping in gallium nitride (GaN) wafers. The method includes: simultaneously generating a pre-processing structure comprising a vaporized core layer and an annular thermal boundary layer in the wafer dicing area using a coaxial dual-focus laser system; controlling an optically integrated single-blade dicing pre-processing structure integrating a microlens array and a 45° reflector, while simultaneously guiding an auxiliary laser to the dicing front edge for real-time material softening based on the 45° reflector; calculating the crack propagation value based on the dynamic thermal field distribution in the dicing area using the thermal diffusion equation; performing micro-shaping on the dicing front edge when the value is >0.25 mm; and dynamically adjusting the pulse energy spatial ratio of the laser system based on the micro-shaping feedback until the dicing operation is completed. This invention effectively solves the problem of uncontrollable cross-grain boundary edge chipping size in traditional laser dual-blade dicing of gallium nitride wafers due to anisotropic thermodynamic response differences.

Description

Technical Field

[0001] This invention relates to the field of wafer dicing technology, and in particular to a method and system for controlling edge chipping of gallium nitride wafers. Background Technology

[0002] In the field of semiconductor device manufacturing, gallium nitride (GaN) wafers, as the core material of third-generation semiconductors, present a dual challenge to the cutting process due to their high hardness, high brittleness, and multi-domain heterostructure: overcoming the inherent defects of traditional mechanical cutting, such as high tool wear and poor sidewall perpendicularity, and solving the uncontrollable edge chipping problem caused by material anisotropy in laser cutting.

[0003] The current industry standard employs a stepped dual-blade cutting scheme, where a laser is first used to create a pre-treatment groove, followed by a second cutting blade cutting along the bottom of the groove. While this method reduces the risk of macroscopic fragmentation, it suffers from fundamental bottlenecks: First, the dual-blade process accumulates repetitive positioning errors, leading to misalignment between the pre-treatment area and the cutting path, forcing the actual cutting line to deviate from the optimal lattice cleavage plane. Second, the step-by-step operation causes thermal stress to oscillate repeatedly during the two energy loading cycles, especially when cutting across polycrystalline domains, where sudden changes in local thermal strain rate induce microcracks at grain boundaries, which then propagate into edge chipping. Third, each independent process requires two sets of equipment, significantly increasing production costs. The core contradiction faced by advanced packaging production lines both domestically and internationally lies in the fact that existing technologies, in order to ensure cutting yield, are forced to employ complex and energy-intensive processes, directly hindering the economic viability of mass production of gallium nitride devices.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for controlling edge chipping in gallium nitride wafers, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for controlling edge chipping in gallium nitride wafers, the method comprising: Based on a coaxial dual-focus laser system, a pre-processing structure including a vaporized core layer and an annular thermal boundary layer is simultaneously generated in the wafer dicing area, wherein the temperature gradient of the annular thermal boundary layer is ≤250℃ / mm. The optically integrated single-blade cutter, which controls the integrated microlens array and 45° reflector, cuts into the pre-processed structure, while the auxiliary laser is guided to the cutting front edge by the 45° reflector to soften the material in real time. Based on the dynamic thermal field distribution of the cutting area captured by an infrared thermal imager, the crack propagation β value is calculated by inversion using the thermal diffusion equation. When the crack propagation β value is greater than 0.25 mm, a repair laser pulse is emitted towards the cutting front to perform micro-shaping; The pulse energy spatial ratio of the coaxial dual-focus laser system is dynamically adjusted according to the micro-shaping feedback until the cutting operation is completed.

[0007] Furthermore, the conditions that the coaxial dual-focus laser system meets include: The vaporization core layer uses a first laser source with a wavelength range of 355-1064nm and a pulse width of 10-100ns. The annular thermal boundary layer uses a second laser source with a wavelength range of 980-1550nm and adopts a continuous wave mode; The energy ratio between the first laser source and the second laser source is 1.5-3.0.

[0008] Further, defining the gasification core layer and the annular thermal boundary layer includes: Based on the spot position distribution relationship of the coaxial dual-focus laser system, a concentric ring structure radiating from the inside to the outside is formed on the surface of the cutting track as a layer boundary reference. The effective working diameter of the gasification core layer was confirmed by the concave curvature feature of the edge of the gasification core layer in the preprocessed structure captured by optical imaging. The outer edge coordinates of the annular thermal boundary layer are established by combining the detection of temperature transition inflection points on the outer side of the annular thermal boundary layer using the infrared thermal imager. The width of the interlayer melting transition zone is determined based on the difference in spatial overlap between the effective diameter of the gasification core layer and the outer edge coordinates of the annular thermal boundary layer.

[0009] Further, generating the crack propagation β value includes: The real-time location temperature gradient dataset is derived from the dynamic thermal field distribution of the cut area captured by the infrared thermal imager. Input the location temperature gradient dataset into the heat conduction model with the annular thermal boundary layer temperature gradient in the preprocessed structure as the constraint boundary. The heat conduction model is coupled with the displacement rate of change field generated by the optically integrated single-blade cutting motion for spatiotemporal field iteration; Output the spatial distribution of the crack propagation β value, which characterizes the material response state of the cutting region, during the spatiotemporal field iteration.

[0010] Furthermore, a heat conduction model is established, including: The heat flux density control conditions of the heat conduction model are configured using the temperature gradient of the annular thermal boundary layer as a spatial boundary constraint. The dynamic thermal field distribution of the cutting area at the moment when the optically integrated single blade cuts into the pre-processed structure is set as the initial state of heat conduction. Based on the phase transition characteristics of the gallium nitride material in the high-temperature cutting zone, a heat conduction mechanism conversion rule is applied. The update cycle of the dynamic thermal field distribution of the infrared thermal imager is set with a time iteration step size to achieve synchronization with the cutting motion.

[0011] Furthermore, the optical integration of the integrated microlens array and the 45° reflector is precisely cut into the pre-processed structure, including: The cutting laser beam output by the optically integrated single blade is focused by the microlens array and then vertically aligned with the center of the vaporization core layer to initialize the cutting stroke. Adjust the deflection angle of the 45° reflector according to the cutting path so that the reflected light spot always leads and covers the unprocessed area in the cutting direction. Based on the lattice energy distribution characteristics generated by the microlens array, an orientation-adaptive cutting trajectory is generated at the multi-domain boundary. The mechanical pressure gradient of the optically integrated blade is adjusted according to the radial temperature decay characteristics of the annular thermal boundary layer.

[0012] Furthermore, the auxiliary laser is guided to the cutting front edge by the 45° reflector for real-time material softening, including: The auxiliary laser beam that deflects the 45° reflector is focused onto the cutting surface in front of the direction of travel of the cutting laser beam to form a pre-softened spot; The shape and contour of the pre-softened spot are adjusted according to the lattice orientation change of the adaptive cutting trajectory. The energy density range of the pre-softened light spot is adjusted based on the temperature distribution of the annular thermal boundary layer. The distance between the pre-softened spot and the edge of the vaporized core layer is corrected based on the cutting stroke feedback of the cutting laser beam.

[0013] Furthermore, generating crystal orientation-adaptive ingress trajectories includes: Scanning at the polydomain boundaries to obtain micro-region grain boundary distribution maps and identify the orientation of the main cleavage planes; A micro-insulation groove structure is constructed by utilizing the lattice energy distribution characteristics of the microlens array along the orientation extension direction of the main cleavage surface. The projection component of the cutting vector trajectory on the rhombic dodecahedral sliding system is constrained by the real-time numerical constraint of the mechanical pressure gradient. The process feedback based on the cutting stroke dynamically compensates for the trajectory offset caused by the difference in elastic modulus between adjacent grains.

[0014] A gallium nitride wafer edge chipping control dicing system, the system comprising: The preprocessing module, based on a coaxial dual-focus laser system, simultaneously generates a preprocessing structure including a vaporized core layer and an annular thermal boundary layer in the wafer dicing area; The single-blade control module controls the optically integrated single-blade cutting pre-processing structure that integrates a microlens array and a 45° reflector, while simultaneously guiding the auxiliary laser to the cutting front edge for real-time material softening based on the 45° reflector. The value calculation module calculates the crack propagation β value by inverting the thermal diffusion equation based on the dynamic thermal field distribution of the cutting area captured by the infrared thermal imager. The micro-shaping module emits repair laser pulses towards the cutting front to perform micro-shaping when the crack propagation β value is greater than 0.25 mm. The energy adjustment module dynamically adjusts the spatial ratio of the pulse energy of the coaxial dual-focus laser system based on micro-shaping feedback until the cutting operation is completed.

[0015] Furthermore, the β value calculation module includes: Temperature gradient unit: Real-time location temperature gradient dataset is obtained by analyzing the dynamic thermal field distribution of the cut area captured by infrared thermal imager. The model building unit inputs the location temperature gradient dataset into a heat conduction model that uses the temperature gradient of the annular thermal boundary layer in the preprocessed structure as the constraint boundary. The spatiotemporal iteration unit couples the heat conduction model with the displacement rate of change field generated by the optically integrated single-blade cutting motion to perform spatiotemporal field iteration; The value characterization unit outputs the spatial distribution of the crack propagation β value, which represents the material response state of the cutting region during the spatiotemporal field iteration.

[0016] The technical solution of this invention can achieve the following technical effects: By establishing a temperature-controlled gradient annular thermal boundary layer in the coaxial dual-focus laser preprocessing stage, achieving crystal orientation-adaptive dynamic control of the cutting tool based on the collaboration of a 45° reflector and a microlens array, and performing laser repair compensation based on real-time inversion of crack propagation, the ternary closed-loop technology effectively solves the problem of uncontrollable cross-grain boundary chipping size caused by anisotropic thermal response differences in gallium nitride wafers during traditional laser dual-blade cutting.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a gallium nitride wafer edge breakage control dicing method; Figure 2 A schematic diagram of the process for generating the crack propagation β value; Figure 3 A schematic diagram of the process for establishing a heat conduction model; Figure 4 A flowchart illustrating the process of controlling the preprocessing structure. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1; like Figure 1 As shown, this application provides a method for controlling edge chipping in gallium nitride wafers through dicing, the method comprising: S10: Based on a coaxial dual-focus laser system, a pre-processed structure including a vaporized core layer and an annular thermal boundary layer is simultaneously generated in the wafer dicing area, wherein the temperature gradient of the annular thermal boundary layer is ≤250℃ / mm. S20: An optically integrated single-blade pre-processing structure that controls the integrated microlens array and 45° reflector, while simultaneously guiding the auxiliary laser to the cutting front edge for real-time material softening based on the 45° reflector; S30: Based on the dynamic thermal field distribution of the cutting area captured by an infrared thermal imager, the crack propagation degree β value is calculated by inversion using the thermal diffusion equation; S40: When the crack propagation β value is greater than 0.25 mm, a repair laser pulse is emitted towards the cutting front to perform micro-shaping; S50: Dynamically adjust the spatial ratio of pulse energy of the coaxial dual-focus laser system according to micro-shaping feedback until the cutting operation is completed.

[0023] Specifically, firstly, the wafer is fixed on a motion platform with vacuum adsorption and aligned with the dicing area. A coaxial dual-focus laser system is then activated to simultaneously construct a pre-treatment structure on the same dicing track. The pre-treatment structure, from the inside out, sequentially forms a vaporization core layer and an annular thermal boundary layer. Preferably, the vaporization core layer is powered by a first laser source providing pulse energy to achieve rapid material removal and form a pre-treatment trench of controllable depth. The first laser source can be a laser with a wavelength in the range of 355–1064 nm and a pulse width of 10–100 ns. The annular thermal boundary layer is formed by… The second laser source provides continuous wave energy to establish a gradually varying thermal field around the trench and to suppress thermal shock and guide cracks. The second laser source can be selected with a wavelength in the range of 980 to 1550 nm and operates in continuous wave mode. At the same time, through beam shaping and power closed loop, the radial temperature change of the annular thermal boundary layer is kept gradual. Preferably, its temperature gradient is controlled to be no greater than 250 °C / mm to avoid the initiation of microcracks induced by transient thermal strain rate changes at the polycrystalline domain junctions. More preferably, the energy ratio of the first laser source to the second laser source is controlled to be between 1.5 and 3.0, thus balancing the synergistic effect of core vaporization into a groove and circumferential thermal boundary buffering, and confirming the pre-processed structure through optical imaging: using the concentric distribution of coaxial dual-focal spots as the interlayer boundary benchmark, determining its effective diameter by utilizing the concave curvature characteristics of the vaporization core layer edge, and then determining the outer edge coordinates by combining the detection of the temperature transition inflection point on the outer side of the annular thermal boundary layer by infrared thermal imager, thereby verifying the spatial overlap and transition zone width between the two layers, ensuring that the cutting path stably falls within the composite window of the core removal zone and the thermal boundary buffer zone during subsequent single-blade cutting; subsequently, entering the cutting stage The process involves controlling an integrated microlens array and a 45° reflector to perform a single-cut and continuous advance of the pre-processed structure. Preferably, the cutting laser beam output by the single cutter is first processed by the microlens array to form a lattice-like, uniform energy distribution, and then vertically aligned with the center of the vaporized core layer to initialize the cutting stroke. This ensures that the cutting edge or cutting energy is preferentially expanded in the removed or weakened area to reduce the initial risk of edge chipping. Simultaneously, the deflection angle of the 45° reflector is adjusted in real time according to the cutting path, ensuring that the reflected light spot always leads and covers the unprocessed area in the cutting direction. This process is further utilized by the 45° reflector... A 5° reflector guides the auxiliary laser to the cutting front to form a pre-softening spot for real-time material softening. The pre-softening spot preferably falls on the surface of the cutting path ahead of the laser beam's travel direction. Its energy density is adaptively adjusted according to the temperature distribution of the annular thermal boundary layer, and the spot profile can be adjusted based on the lattice orientation change of the adaptive cutting trajectory. For example, at the point crossing a multi-domain boundary, the spot transitions from a near-circular shape to an ellipse stretched along the main cleavage plane to reduce the impact of the lateral thermal gradient on the grain boundary. Simultaneously, the distance between the pre-softening spot and the edge of the vaporization core layer is corrected based on the cutting stroke feedback, thus softening the material. The heat treatment zone can reduce the equivalent hardness and brittleness of the material in advance without causing excessive thermal effects that could lead to sidewall collapse. During the cutting process, an infrared thermal imager is used to capture the dynamic thermal field distribution of the cutting zone in real time. The temperature field data is output with an update cycle synchronized with the cutting motion. This data is then parsed into a real-time position temperature gradient dataset and input into a heat conduction model with the annular thermal boundary layer temperature gradient as the constraint boundary. At the same time, the displacement rate of change field generated by the single-blade cutting motion is coupled for spatiotemporal iteration, thereby inverting the crack propagation β value and its spatial distribution, which characterizes the material response state of the cutting region. When the β value exceeds 0, the crack propagation β value is detected.At 25mm, a repair laser pulse is immediately triggered to the cutting front to perform micro-shaping. Micro-shaping preferably employs short pulses and low heat accumulation to create microscale remelting and resolidification near the crack tip, thus blunting the crack tip, reducing stress concentration, and preventing further crack propagation. After micro-shaping, the system dynamically adjusts the pulse energy spatial ratio of the coaxial dual-focus laser system based on the thermal image feedback and cutting stability feedback. Specifically, the preferred approach is: when crack propagation tends to increase, the relative energy ratio of the annular thermal boundary layer is increased to enhance thermal buffering and reduce radial temperature abrupt changes; when cutting efficiency decreases but crack propagation is controlled, the relative energy of the vaporization core layer is moderately increased. The ratio is used to restore the efficiency of trenching or removal, so as to maintain the closed-loop stability of pre-processing structure, single-blade cutting, thermal field inversion, repair and shaping, and energy ratio until the cutting is completed; for example, for a 6-inch GaN wafer on a cutting track that crosses a multi-domain region, a first laser source (e.g., 355nm or 1064nm, pulse width in the range of 10-100ns) can be preferably used to form a continuous vaporized core layer trench in the center of the cutting track, while a second laser source (e.g., 980nm or 1550nm continuous wave) is used to form an annular thermal boundary layer with a temperature gradient of no more than 250℃ / mm around the trench, and the energy ratio of the two is maintained at about 2:1, during the single-blade advance process. A mirror-guided auxiliary laser forms a pre-softening zone at the leading edge. When the β value rises and exceeds the 0.25mm threshold obtained from thermal image inversion, a repair pulse is immediately used to micro-shape the leading edge. Subsequently, the crack propagation and fallback are achieved by increasing the energy ratio of the annular thermal boundary layer. Compared with the uncontrollable cross-grain boundary chipping caused by repeated positioning errors and two thermal load oscillations in traditional stepped double-blade cutting, this solution can achieve real-time controllable and stable convergence of chipping size in a single-blade single-cutting process, reducing equipment and process complexity and improving yield and mass production economy.

[0024] The technical solution of this invention establishes a temperature-controlled gradient annular thermal boundary layer in the coaxial dual-focus laser preprocessing stage, achieves dynamic control of the cutting tool with crystal orientation adaptation based on the collaboration of a 45° reflector and a microlens array, and performs laser repair compensation based on real-time inversion of crack propagation. This ternary closed-loop technology effectively solves the problem of uncontrollable cross-grain boundary chipping size caused by anisotropic thermal response differences in gallium nitride wafers during traditional laser dual-blade cutting.

[0025] Furthermore, the conditions that a coaxial dual-focus laser system must meet include: The gasification core layer uses a first laser source with a wavelength range of 355-1064nm and a pulse width of 10-100ns. The annular thermal boundary layer uses a second laser source with a wavelength range of 980-1550nm and adopts a continuous wave mode; The energy ratio of the first laser source to the second laser source is 1.5-3.0.

[0026] As a preferred embodiment of the above, the vaporization core layer uses a first laser source with a wavelength selected in the range of 355–1064 nm and a pulse width controlled in the range of 10–100 ns. This allows for the generation of a high instantaneous energy density at the center of the dicing channel, causing the gallium nitride material to melt rapidly and further vaporize, thereby forming a stable pre-treatment trench. Preferably, when a metal layer exists on the surface of the wafer to be diced or when it is necessary to reduce the heat-affected zone width, a 355 nm short-wavelength pulse can be preferentially selected to improve the energy coupling efficiency to the surface layer and reduce lateral heat diffusion. When a certain depth of action and processing efficiency need to be considered, a 1064 nm pulse can be preferentially selected, and the pulse width is preferably set in the range of 20–60 ns to achieve rapid... A balance is achieved between rapid removal and controllable thermal shock. The pulse repetition frequency of the first laser source can be matched with the platform feed speed, so that adjacent pulses form a continuous and uniform vaporization core layer at the center of the cutting path. Simultaneously, a second laser source is used for the annular thermal boundary layer. Its wavelength is selected in the range of 980-1550nm and it adopts a continuous wave mode to continuously provide a controllable heat flux around the vaporization core layer, forming an annular preheating zone distributed around the trench circumferentially. This reduces the abrupt change in thermal strain rate caused by the anisotropic thermodynamic response difference when gallium nitride is cut across grain boundaries and inhibits the initiation of microcracks. Preferably, the second laser source can form an annular spot by beam shaping, for example, shaping the continuous wave beam into a lower energy center and a lower energy outer ring. The high-energy distribution, coaxially superimposed with the pulse spot of the first laser source, ensures stable spatial alignment between the annular thermal boundary layer and the vaporization core layer, guaranteeing that subsequent single-blade cuts always fall within the composite window of the central weakening zone and the circumferential buffer zone. Regarding the energy ratio, the energy ratio of the first laser source to the second laser source is limited to 1.5–3.0, preferably achieved through independent adjustable attenuation after coaxial beam combining. For example, power control is set for each laser path and calibrated at the workpiece focal plane to ensure that the removal energy contributed by the first laser source per unit length of the cutting path is always higher than the preheating energy contributed by the second laser source, but not excessively higher. This ensures both the formation of a clear vaporization core layer groove and the formation of annular... The thermal boundary layer continuously plays a temperature buffering role. More preferably, when infrared thermal imaging feedback shows insufficient temperature rise around the cutting track and a tendency for edge chipping to expand, the energy ratio can be adjusted towards 1.5 to enhance annular preheating and buffering. When the cutting efficiency decreases but the thermal field is stable, the energy ratio can be adjusted towards 3.0 to improve the central vaporization efficiency. For example, in the process chain of film lamination, thinning, laser grooving, and single-blade scribing, a first laser source with a pulse width of about 30ns and a 355nm wavelength can be used to form a continuous vaporization core layer at the center of the cutting track during the laser grooving stage. At the same time, a second laser source with a continuous wave of 1550nm wavelength can be used to form an annular thermal boundary layer around the groove, and the energy ratio of the two can be preferably set at about 2.The temperature is around 0, ensuring the pre-processed structure has sufficient center removal depth while providing a stable thermal buffer on the periphery. This provides low-damage, low-chipping conditions for subsequent single-blade cutting, thus meeting the limitations on the wavelength and pulse width of the first laser source, the wavelength and continuous wave mode of the second laser source, and the energy ratio range.

[0027] Furthermore, defining the vaporization core layer and the annular thermal boundary layer includes: Based on the spot position distribution relationship of the coaxial dual-focus laser system, a concentric ring structure radiating from the inside to the outside is formed on the surface of the cutting track as a reference for interlayer boundary. The effective working diameter of the gasification core layer was confirmed by the concave curvature features of the edge of the gasification core layer in the preprocessed structure captured by optical images. By combining infrared thermal imager to detect the temperature transition inflection point on the outer edge of the annular thermal boundary layer, the coordinates of the outer edge of the annular thermal boundary layer are established. The width of the interlayer melting transition zone is determined based on the difference in spatial overlap between the effective diameter of the gasification core layer and the outer coordinates of the annular thermal boundary layer.

[0028] As a preferred embodiment of the above embodiments, during preprocessing, the first laser source and the second laser source are kept coaxially focused and output in a co-focused or quasi-co-focused manner, so that a concentric ring structure radiating from the inside out is formed on the surface of the cutting channel. The inner ring corresponds to the main working area of ​​the first laser source, and the outer ring corresponds to the continuous heating area of ​​the second laser source. Preferably, the geometric center of the concentric ring is aligned with the center line of the cutting channel through equipment calibration. For example, the center line of the cutting channel is used as the zero point, and the coordinates of the center point of the concentric ring are recorded in the platform coordinate system, so that the concentric ring structure is used as the interlayer boundary benchmark for subsequent measurements. When confirming the effective working diameter of the gasification core layer, it is preferable to use coaxial visible light microscopy or a process camera to perform preprocessing. The processing structure is used for optical image capture, and the concave curvature feature of the edge of the vaporized core layer is used as an identification marker. Specifically, in the vaporized core layer region, due to rapid melting or vaporization of the material, grooves are formed, and their edges typically appear as a concave contour transitioning from the center outwards in reflected or scattered light imaging. Preferably, multiple cross-sections are scanned along the width of the cut path during imaging, and the boundary position where the concave contour changes from descending to rising is selected as the core layer edge point set. Then, this point set is fitted with the center of the concentric ring structure as a reference to obtain the effective working diameter of the vaporized core layer. When establishing the coordinates of the outer edge of the annular thermal boundary layer, it is preferable to synchronously trigger the infrared thermal imager and the platform movement. Under the condition of stable continuous wave output from two laser sources, the temperature distribution image near the pre-processed cutting path is acquired, and the temperature transition inflection point is identified on the radial decay curve of the temperature field as the outer edge criterion. Specifically, preferably, the temperature change trend is read along multiple radial directions starting from the center of the concentric ring. When the temperature changes from a relatively gradual decay to a region close to the base temperature and shows a significant change trend inflection, the spatial position corresponding to this inflection is marked as the outer boundary point set of the annular thermal boundary layer, and then the outer edge coordinates of the annular thermal boundary layer are established in the platform coordinate system. Finally, in order to obtain the width of the interlayer melting transition zone that can reflect the interaction between the two layers and the subsequent cutting stability, it is preferable to vaporize... The effective diameter of the core layer and the coordinates of the outer edge of the annular thermal boundary layer are projected onto the same coordinate reference for spatial overlap difference assessment. That is, taking the center of the concentric ring as a common reference, the radial distance between the edge circle of the core layer and the outer edge circle (or the outer edge closed curve) of the thermal boundary layer is regarded as the width of the melt preheating transition zone between the two working zones. It is preferable to statistically analyze the consistency of this width segmentally along the cutting path length direction: when the transition zone is too narrow, it indicates insufficient thermal buffer and the cross-grain boundary thermal strain rate may be too large; when the transition zone is too wide, it indicates an expansion of the thermal influence range and the possible introduction of unnecessary lateral thermal damage. Based on this, it can provide a basis for adjusting subsequent process parameters, such as the first or second laser energy ratio or spot shaping.For example, after performing coaxial dual-focus preprocessing on the dicing track of a 6-inch GaN wafer, a clear concentric ring image can be obtained on the dicing track surface using a process camera, and its center can be locked. Then, the concave contour of the vaporization trench edge can be identified from the optical image to determine the effective diameter of the core layer. Simultaneously, infrared thermography can be used to capture the temperature transition inflection point at the transition from the outer ring temperature zone to the substrate temperature zone to determine the outer edge coordinates of the thermal boundary layer. Finally, the radial difference between the two under the same central reference can be used to characterize the width of the interlayer melting transition zone. This achieves precise definition of the vaporized core layer and the annular thermal boundary layer, ensuring a stable and consistent reference for subsequent single-blade cutting and crack control strategies.

[0029] Furthermore, the crack propagation β value includes: The real-time location temperature gradient dataset is derived from the dynamic thermal field distribution of the cut area captured by the infrared thermal imager. Input the location temperature gradient dataset into the heat conduction model with the temperature gradient of the annular thermal boundary layer in the preprocessed structure as the constraint boundary; The spatiotemporal field iteration of the coupled heat conduction model and the displacement rate field generated by the optically integrated single-blade cutting motion is performed. The spatial distribution of crack propagation β values, which characterize the material response state of the cutting region, during the spatiotemporal field iteration.

[0030] As a preferred embodiment, the crack propagation β value refers to the equivalent propagation amount, expressed in length, obtained by heat conduction inversion calculation based on the dynamic thermal field distribution of the cutting zone and combined with cutting motion parameters, used to characterize the degree of propagation of the crack at the cutting edge into the material or laterally. The infrared thermal imager is preferably synchronized with the motion platform and the single-blade cutting trigger signal, so that each frame of thermal image corresponds to a specific position and time point on the cutting path. Coordinate calibration and distortion correction are performed on the thermal image to establish a one-to-one correspondence between the thermal image pixel coordinates and the platform coordinates and the cutting path centerline coordinates. Subsequently, a resolution window is set on both sides of the cutting path centerline for each frame of thermal image, preferably covering the area where the vaporization core layer and the outer edge of the annular thermal boundary layer are located. Within this window, the resolution window is simultaneously positioned radially and in the cutting advancement direction. Temperature change trends are extracted by spatially subtracting the temperature differences between adjacent pixels or sampling points to obtain a real-time location temperature gradient dataset. This dataset preferably includes gradient direction and intensity information at each time point and location. Time-series filtering is used to suppress transient noise caused by surface emissivity fluctuations, plasma scintillation, or local reflections to ensure the stability of the gradient data. During the modeling phase, when inputting the aforementioned location temperature gradient dataset into the heat conduction model, the temperature gradient of the annular thermal boundary layer in the preprocessed structure is preferably used as the constraint boundary. Specifically, the temperature transition inflection point on the outer side of the annular thermal boundary layer, determined by infrared thermography, and its corresponding gradient range are used as boundary conditions, ensuring that the model consistently satisfies the outer ring gradient during iteration. The physical constraints of gradient thermal buffering prevent the transient thermal shock of cutting from being misjudged as crack propagation. Simultaneously, within the model, different thermal diffusion response trends are preferably assigned based on the difference between the weakened zone at the center of the cutting path and the circumferential buffer zone, reflecting the objective state that the vaporization core layer has been removed or weakened and the annular thermal boundary layer is continuously preheated. Furthermore, to ensure consistency between the thermal field simulation and the actual cutting process, the thermal conduction model is preferably coupled with the displacement rate of change field generated by the optically integrated single-blade cutting motion. The displacement rate of change field can be directly obtained from the platform encoder and cutting trajectory planning, including information such as cutting advance speed, acceleration and deceleration changes, speed changes at corners, and the trend of single-blade penetration depth with stroke. The preferred coupling method is: within each thermal imaging time step, the thermal field is... The displacement change rate at each time step is mapped to the update of the position of the moving heat source at the cutting front and the update of the duration of local heat input. The thermal field is iteratively updated in the next time step, so that the evolution of the thermal field is not only affected by the thermal diffusion of the material, but also reflects the difference in heat accumulation caused by the rapid movement of the cutting front. In the spatiotemporal field iterative output stage, the material response state of the cutting area is preferably used as an intermediate characterization quantity. By comprehensively evaluating the characteristics such as the temperature gradient concentration area, the duration of the gradient peak, and the positional offset of the gradient concentration area relative to the cutting front obtained by the iteration, the spatial distribution of the crack propagation β value is output. The spatial distribution preferably gives the thermodynamic response map of the β value changing with position on both sides of the center line of the cutting path, so that it can be directly used for threshold trigger repair laser pulse and energy ratio adjustment in the future.For example, when the cutting process advances into the polycrystalline domain boundary region, if infrared thermography reveals a persistent high gradient band near the cutting front, accompanied by this high gradient band shifting outward toward the sidewall of the cutting path, then in the spatiotemporal iteration coupled with the displacement rate field, this outward shift trend will be stably preserved and reflected as an increase in the β value in the lateral direction. This forms a lateral propagation zone on the spatial distribution map. Compared to methods that only use the highest temperature of a single frame for judgment, the spatial distribution of the β value can characterize the possible propagation direction and extent of cracks earlier and more accurately. This provides a reliable basis for subsequent closed-loop control, such as micro-shaping when the β value exceeds the threshold, and adjusting the dual-focus energy ratio based on the shaping feedback, and achieves real-time suppression of edge breakage.

[0031] Furthermore, establishing a heat conduction model includes: The temperature gradient of the annular thermal boundary layer is used as a spatial boundary constraint to configure the heat flux density control condition of the heat conduction model. The dynamic thermal field distribution of the cutting zone at the moment when the optically integrated single blade cuts into the pre-processed structure is set as the initial state of heat conduction. Based on the phase transition characteristics of gallium nitride materials in the high-temperature cutting region, a heat conduction mechanism conversion rule is applied; The update cycle of the dynamic thermal field distribution of the associated infrared thermal imager is set with a time iteration step size to achieve synchronization with the cutting motion.

[0032] As a preferred embodiment of the above, the temperature gradient of the annular thermal boundary layer is first used as a spatial boundary constraint to configure the heat flux density control condition of the heat conduction model. Preferably, the outer edge coordinates are obtained by continuously tracking the temperature transition inflection point on the outer side of the annular thermal boundary layer using an infrared thermal imager, and the allowable range of the temperature gradient at these outer edge coordinates is written into the heat flux density control condition. This allows the model to limit and gradually constrain the heat flux diffusing outward from the cutting area in each time iteration, thereby keeping the outer ring region in a preset gradual gradient thermal buffer state and avoiding lateral thermal stress spikes caused by sudden changes in heat flux density at the outer edge due to local heat input fluctuations. Based on this, an optically integrated single-blade cutting preprocessing structure is then implemented. The moment of entry is taken as the starting moment of the heat conduction model, and the dynamic thermal field distribution of the cutting area captured by the infrared thermal imager at this moment is set as the initial state of heat conduction. Preferably, one or more frames of thermal images are acquired at the moment of single-blade entry and time-series fusion is performed to reduce the accidental noise of the entry transient. Then, the fused temperature distribution is mapped to the local computational domain of the cutting path in the platform coordinate system, so that the initial value of the model can simultaneously reflect the high temperature concentration characteristics of the central weakened region of the vaporization core layer and the circumferential preheating characteristics of the annular thermal boundary layer, thereby ensuring that subsequent iterations start from the actual entry conditions rather than from an idealized uniform temperature field. Subsequently, the heat conduction mechanism conversion rule is applied according to the phase transition characteristics of gallium nitride material in the high-temperature cutting area. Preferably, the changes in the equivalent heat conduction path caused by solid-state heat conduction, local melting heat conduction, vaporization, or removal that the cutting area may experience are considered as the switching scenarios of the heat conduction mechanism, and the temperature range and the gradient concentration pattern shown in the thermal image are used as the triggering criteria: when the thermal field shows that the temperature in the central region is continuously concentrated at a high temperature and accompanied by a strong gradient ring corresponding to the edge recessed groove, the model preferably switches to a conduction mechanism reflecting the existence of local melting, so as to reflect the change in the effective heat diffusion capacity and thermal capacity response near the cutting front; when the thermal image shows that the temperature in the central region drops rapidly and coincides with the location of the groove formation, the model preferably switches to a mechanism reflecting the change in the equivalent conduction path after the material is removed, so that the model cools down in that region. The low-efficiency heat transfer channel distributes heat diffusion mainly to the circumferential buffer zone, which is more consistent with the actual state where the gasification core layer has been formed and the outer ring continues to buffer. Finally, in order to achieve synchronization with the cutting motion and ensure the temporal consistency of β value inversion, the time iteration step size is preferably set by the update cycle of the dynamic thermal field distribution of the associated infrared thermal imager. That is, the time interval between two adjacent frames of thermal images output by the infrared thermal imager is used as the iteration step size of the model, and each iteration step corresponds to the displacement advance of the platform within the time interval. In this way, the synchronous relationship of one frame of thermal image, one iteration step, and one displacement update is achieved in the spatiotemporal iteration, avoiding misjudgment caused by thermal field inference lagging behind or ahead of the cutting front.For example, when a single blade advances along the cutting path at a constant feed rate, and the infrared thermal imager outputs thermal images at a fixed period, the model is iteratively updated at the same period. Each iteration applies a ring-shaped thermal boundary layer gradient constraint based on the true thermal field given by the previous frame's thermal image and updates the local conduction mechanism according to phase transition rules. This yields a continuously evolving thermal field prediction result as the cutting front moves, providing a stable and traceable thermal basis for the subsequent spatial distribution output of crack propagation β values. Furthermore, it supports real-time identification and repair laser-triggered control of edge breakage risk.

[0033] Furthermore, the optically integrated single-blade pre-processing structure for controlling the integrated microlens array and the 45° reflector includes: The cutting laser beam output by the optically integrated single blade is focused by a microlens array and then vertically aligned with the center of the vaporization core layer to initialize the cutting stroke. Adjust the deflection angle of the 45° reflector according to the cutting path so that the reflected light spot always leads and covers the untreated area in the cutting direction. Based on the lattice energy distribution characteristics generated by the microlens array, an orientation-adaptive cutting trajectory is generated at the multi-domain boundary. The mechanical pressure gradient of the optically integrated blade is adjusted according to the radial temperature decay characteristics of the annular thermal boundary layer.

[0034] As a preferred embodiment of the above, the cutting laser beam output by the optically integrated single blade is first focused and homogenized by a microlens array. The microlens array is preferably positioned in the collimation section of the cutting laser beam, resulting in a stable lattice energy distribution and an equivalent processing area with multiple superimposed micro-spots on the focal plane. Subsequently, the focused cutting laser beam is vertically aligned with the center of the vaporized core layer for initial cutting stroke. Preferably, the concentric ring boundary of the pre-processed structure is used as the alignment reference, ensuring the cutting starting point falls near the center line of the weakened groove in the vaporized core layer. This ensures that material removal or fracture in the initial cutting stage preferentially develops along the weakened region, significantly reducing the energy loss during the first contact of the blade edge or energy with the dense GaN. Lateral impact; during the cutting process, the deflection angle of the 45° reflector is adjusted in real time according to the cutting path, so that the reflected light spot always leads and covers the untreated area in the cutting direction. Preferably, the deflection of the 45° reflector is dynamically related to the platform feed direction: when the cutting path is straight, the reflected light spot is kept within a fixed distance in front of the cutting edge; when the cutting path changes direction or crosses the polycrystalline domain boundary and requires trajectory fine-tuning, the reflector quickly corrects at a small angle so that the reflected light spot remains on the side from the cutting edge to the untreated area, thereby forming a continuous energy preload at the cutting edge, reducing the equivalent hardness and brittleness of the cutting edge material and reducing the peak cutting load; at the same time, the lattice energy distribution characteristics generated by the microlens array are utilized in the polycrystalline... At the domain boundary, a crystal orientation-adaptive cutting trajectory is generated. The preferred approach is as follows: when the cutting advances near the polycrystalline domain boundary region, the cutting trajectory is slightly biased by observing the energy superposition morphology of the lattice micro-spots in the processing area and the cutting feedback, such as cutting stability or the gradient concentration direction shown in the thermal image. This causes the equivalent processing area to exhibit a distributed, gradual cutting along the main crystal orientation at the boundary. This avoids stress concentration caused by a single high-energy center point directly crossing the grain boundary. Instead, the fracture path is first laid out through multi-point micro-fractures and micro-removals formed by the lattice micro-spots, and then continuous penetration is completed by the cutting front, thus achieving adaptive passage to different crystal orientations / grain boundaries. Furthermore, to further suppress edge chipping, it is preferable to consider the radial temperature attenuation characteristics of the annular thermal boundary layer. The mechanical pressure gradient of the optically integrated single blade is adjusted accordingly. This means that the lower cutting resistance in the central region of the cutting path, due to the weakening of the vaporized core layer and the higher temperature, is matched with the gradual decrease in temperature and the gradual increase in material resistance in the outer annular thermal boundary layer region. In terms of cutting load control, a pressure gradient strategy is adopted that gradually increases from the inside to the outside and increases with the decrease in radial temperature: at the initial cutting stage, a lower mechanical pressure is applied at the center of the core layer to avoid lateral compression of the groove edge; as the cutting front approaches the edge of the core layer and advances into the outer transition zone, the mechanical pressure is gradually increased with the decrease in radial temperature to ensure cutting continuity but avoid sudden increases. Preferably, the pressure change is a continuous and smooth transition to weaken the transient load peaks at the polycrystalline domain junctions.For example, on the same cutting path, when the platform detects that it is about to enter the multi-domain boundary region, the lattice equivalent processing area output by the microlens array can be slightly offset relative to the cutting center line to conform to the preferred fracture direction at the boundary. Simultaneously, a 45° reflector keeps the reflected light spot ahead of the leading edge to continuously pre-soften the untreated area. Furthermore, as the infrared thermography shows accelerated temperature decay on the outer side, the mechanical pressure of the single blade is simultaneously increased while maintaining a gentle gradient. This allows for stable crossing of GaN grain boundaries during a single cut and continuous advancement, significantly reducing the probability of edge chipping and its size fluctuations.

[0035] Furthermore, the auxiliary laser is guided to the cutting front edge by a 45° reflector for real-time material softening, including: The auxiliary laser beam, which deflects the 45° reflector, is focused onto the cutting surface in front of the direction of travel of the cutting laser beam to form a pre-softened spot. The shape and contour of the pre-softened spot are adjusted according to the lattice orientation change of the adaptive cutting trajectory. Adjusting the energy density range of the pre-softened spot based on the temperature distribution of the annular thermal boundary layer; The distance between the pre-softened spot and the edge of the vaporized core layer is corrected based on the feedback of the cutting stroke of the cutting laser beam.

[0036] As a preferred embodiment of the above, the auxiliary laser beam deflected by a 45° reflector is first focused and falls on the cutting path surface in front of the cutting laser beam's travel direction to form a pre-softened spot. Preferably, this pre-softened spot is placed in the untreated area in front of the cutting edge and aligned with the centerline of the cutting laser beam, thus applying controllable heat input to the material to be cut in advance. This raises the local temperature to a state more easily broken or removed before the cutting laser beam arrives, reducing the initial impact of the cut. Based on this, the shape and contour of the pre-softened spot are adjusted according to the lattice orientation change reflected by the adaptive cutting trajectory. Preferably, when the cutting path is within a single crystal domain with relatively consistent crystal orientation... The pre-softening spot is set to an approximately circular or short-axis ellipse to achieve uniform softening. When the cutting progresses close to the polydomain boundary or the crystal orientation changes, the pre-softening spot is stretched into a long-axis ellipse or strip-shaped profile along the expected main fracture or cleavage tendency direction. This allows the heat input to be distributed along a direction more conducive to controlled crack propagation, thereby weakening the traction effect of the transverse thermal gradient at the grain boundary on the crack and reducing the risk of edge breakage and outward propagation. At the same time, the profile adjustment is preferably linked with the deflection of the 45° reflector, so that the long axis direction of the spot always rotates synchronously with the cutting path and the adaptive trajectory of the crystal orientation, avoiding local overheating caused by the softening spot still being stretched along the original direction when the trajectory has changed. Furthermore, based on the annular thermal boundary layer... The energy density range of the pre-softening spot is adjusted according to the temperature distribution, preferably with reference to the radial temperature decay trend of the annular thermal boundary layer shown by the infrared thermal imager: when the annular thermal boundary layer shows a faster temperature decay and a weaker thermal buffer on the outer side, the energy density of the pre-softening spot is appropriately increased to compensate for insufficient softening of the leading edge material; when the temperature distribution of the annular thermal boundary layer is high and stable, the energy density of the pre-softening spot is reduced to avoid heat accumulation and expansion of the heat-affected zone, and preferably the energy density adjustment is made to be a continuous and smooth change, thereby maintaining the stability of the thermal state of the cutting edge; at the same time, to ensure that the softening effect closely follows the cutting edge and does not cross the boundary to interfere with the structure of the vaporization core layer, it is preferable to correct the pre-softening spot based on the cutting laser beam's entry stroke feedback. The distance between the pre-softening spot and the edge of the gasification core layer is dynamically feedforward compensated for by using the leading edge position obtained from the platform encoder and cutting trajectory planning, as well as travel information such as cutting depth or feed speed, so that the spot is always located outside the edge of the core layer but not far from the cutting leading edge: when the cutting speed increases or there is an acceleration or deceleration section, the leading distance of the pre-softening spot is appropriately increased to ensure the softening time window; when the cutting advance enters the transition zone of the core layer edge and it is necessary to prevent lateral overheating, the leading distance is appropriately reduced and the spot covers more of the narrow strip area that is about to be cut, thereby avoiding the softening energy from crossing to the edge of the formed gasification groove and causing the groove sidewall to collapse or re-solidify burrs;For example, when the cutting path crosses a multi-domain boundary, the lattice orientation change can be determined by the adaptive cutting trajectory. The pre-softening spot is then adjusted from a near-circular shape to an ellipse stretched along the new crystal orientation. Simultaneously, the energy density is appropriately increased to maintain the softening effect, referencing the temperature decay of the annular thermal boundary layer at that location. Based on the cutting stroke feedback, the distance between the pre-softening spot and the edge of the vaporization core layer is maintained close to the cutting edge but not exceeding the boundary. This achieves real-time softening of the material at the cutting edge and pre-suppression of crack propagation, further ensuring controlled and stable edge chipping during continuous single-blade cutting.

[0037] Furthermore, generating crystal orientation-adaptive cutting trajectories includes: Scanning at polydomain boundaries to obtain micro-region grain boundary distribution maps and identify the orientation of major cleavage planes; A micro-insulation groove structure is constructed by utilizing the lattice energy distribution characteristics of a microlens array along the orientation direction of the main cleavage plane. The projection components of the cutting vector trajectory onto the rhombic dodecahedral sliding system are constrained by the real-time numerical constraint of the mechanical pressure gradient. The process feedback based on the cutting stroke dynamically compensates for the trajectory offset caused by the difference in elastic modulus between adjacent grains.

[0038] As a preferred embodiment of the above, a scan is first performed at the multi-domain boundary to obtain a micro-region grain boundary distribution map and identify the orientation of the main cleavage plane. Preferably, an optical imaging method coaxially aligned with the cutting platform is used to perform a narrow-band line scan or surface scan of the vicinity of the cutting path. A grain boundary distribution map is formed by grain texture, reflection difference, or grain boundary contrast. Then, the direction in which the grain boundary orientation changes most significantly with the main grain orientation is extracted from the map as the reference for the orientation of the main cleavage plane of the micro-region. Preferably, the orientation result is written into the trajectory planning in the form of the local coordinate system of the cutting path so that it can be updated continuously as the cutting process progresses. Subsequently, the lattice energy distribution characteristics of the microlens array are used to scan along the main cleavage plane. The preferred approach for constructing a micro-insulating groove structure along the orientation extension direction is to create an elongated multi-point weakening band of lattice energy in the orientation direction of the main cleavage plane as the cutting edge approaches the grain boundary region. This weakening is achieved by progressively stacking these point-to-point grooves on the surface or near the surface of the cutting path, forming a series of continuous or semi-continuous microscale weakening grooves. Thermally, this acts as a barrier against lateral heat diffusion; mechanically, it acts as a guide for crack propagation. This prioritizes the placement of heat input and stress concentration at the cutting edge along directions more aligned with the material's cleavage tendency, without altering the overall cutting path layout. This reduces the probability of cracks being pulled outwards by unfavorable crystal orientations and causing edge breakage when crossing grain boundaries. During the entry and penetration process, the projection component of the cutting vector trajectory on the rhombic dodecahedral slip system is constrained by the real-time numerical constraint of the mechanical pressure gradient. The mechanical pressure gradient is preferably used as the upper limit constraint on the allowable deflection of the cutting vector: when the temperature decay on the outer side of the annular thermal boundary layer leads to increased material resistance and the mechanical pressure gradient is in the rising stage, trajectory planning preferably reduces the projection component of the cutting vector that easily induces lateral slip, keeping the cutting vector more inclined to stably penetrate along the thickness direction of the cutting path, thus avoiding side cracks at grain boundaries caused by high load superposition and unfavorable slip orientation; when the mechanical pressure gradient is low and the thermal buffer is sufficient, the cutting vector is allowed to make small deviations near the orientation of the main cleavage plane. The system guides the trajectory to improve the smoothness and continuity across grain boundaries, thereby achieving a trajectory constraint logic that is more conservative when the load is higher and more adaptive when the load is lower. At the same time, the system dynamically compensates for the trajectory offset caused by the difference in elastic modulus between adjacent grains based on the progress feedback of the cutting stroke. It preferably uses the leading edge position error, the trend of the propulsion resistance change or the cutting stability feedback that can be obtained in the cutting stroke to determine whether the trajectory has shifted to one side. When a continuous offset trend is detected near the grain boundary, the subsequent short-distance trajectory is immediately finely adjusted and compensated in the opposite direction, and linked with the laying direction of the lattice weakening zone, so that the compensated trajectory is still within the range of being guided along the main cleavage plane but not exceeding the cutting path window.For example, when the cutting front approaches the boundary where the elastic responses of two grains differ significantly, the scanned grain boundary map shows that the orientation of the main cleavage planes is tilted relative to the centerline of the cutting path. In this case, a continuous micro-insulating groove weakening zone can be constructed in this tilted direction using a microlens array to form a thermally and mechanically guiding channel. Subsequently, during actual cutting, a mechanical pressure gradient is used as a constraint to suppress the cutting vector component in unfavorable slip directions. Furthermore, when the stroke feedback shows signs of lateral drift at the cutting front, a small reverse compensation is performed on the trajectory, ensuring that the cutting front remains controlled and continuous across the grain boundary. This achieves the generation and online correction of an orientation-adaptive cutting trajectory, further supporting the chipping suppression effect of this invention in multi-domain gallium nitride wafer cutting.

[0039] Example 2; Based on the same inventive concept as the gallium nitride wafer edge-breaking control dicing method in the foregoing embodiments, the present invention also provides a gallium nitride wafer edge-breaking control dicing system, the system comprising: The preprocessing module, based on a coaxial dual-focus laser system, simultaneously generates a preprocessing structure including a vaporized core layer and an annular thermal boundary layer in the wafer dicing area; The single-blade control module controls the optically integrated single-blade cutting pre-processing structure that integrates a microlens array and a 45° reflector, while simultaneously guiding the auxiliary laser to the cutting front edge for real-time material softening based on the 45° reflector. The value calculation module calculates the crack propagation β value by inverting the thermal diffusion equation based on the dynamic thermal field distribution of the cutting area captured by the infrared thermal imager. The micro-shaping module emits repair laser pulses towards the cutting front to perform micro-shaping when the crack propagation β value is greater than 0.25 mm. The energy adjustment module dynamically adjusts the spatial ratio of the pulse energy of the coaxial dual-focus laser system based on micro-shaping feedback until the cutting operation is completed.

[0040] The adjustment system described above in this invention can effectively realize a gallium nitride wafer edge breakage control dicing method, and the technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0041] Furthermore, The value calculation module includes: Temperature gradient unit: Real-time location temperature gradient dataset is obtained by analyzing the dynamic thermal field distribution of the cut area captured by infrared thermal imager. The model building unit inputs the location temperature gradient dataset into a heat conduction model that uses the temperature gradient of the annular thermal boundary layer in the preprocessed structure as the constraint boundary. The spatiotemporal iteration unit couples the heat conduction model with the displacement rate of change field generated by the optically integrated single-blade cutting motion to perform spatiotemporal field iteration; The value characterization unit outputs the spatial distribution of the crack propagation β value, which represents the material response state of the cutting region during the spatiotemporal field iteration.

[0042] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0043] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for controlling edge chipping in gallium nitride wafers, characterized in that, The method includes: Based on a coaxial dual-focus laser system, a pre-processing structure including a vaporized core layer and an annular thermal boundary layer is simultaneously generated in the wafer dicing area, wherein the temperature gradient of the annular thermal boundary layer is ≤250℃ / mm. The optically integrated single-blade cutter, which controls the integrated microlens array and 45° reflector, cuts into the pre-processed structure, while the auxiliary laser is guided to the cutting front edge by the 45° reflector to soften the material in real time. Based on the dynamic thermal field distribution of the cut area captured by an infrared thermal imager, the crack propagation degree is calculated by inversion using the thermal diffusion equation. value; When the crack propagation When the value is >0.25mm, a repair laser pulse is emitted towards the cutting front to perform micro-shaping; The pulse energy spatial ratio of the coaxial dual-focus laser system is dynamically adjusted according to the micro-shaping feedback until the cutting operation is completed.

2. The gallium nitride wafer edge chipping control dicing method according to claim 1, characterized in that, The conditions that the coaxial dual-focus laser system meets include: The vaporization core layer uses a first laser source with a wavelength range of 355-1064nm and a pulse width of 10-100ns. The annular thermal boundary layer uses a second laser source with a wavelength range of 980-1550nm and adopts a continuous wave mode; The energy ratio between the first laser source and the second laser source is 1.5-3.

0.

3. The gallium nitride wafer edge chipping control dicing method according to claim 2, characterized in that, Defining the gasification core layer and the annular thermal boundary layer includes: Based on the spot position distribution relationship of the coaxial dual-focus laser system, a concentric ring structure radiating from the inside to the outside is formed on the surface of the cutting track as a layer boundary reference. The effective working diameter of the gasification core layer was confirmed by the concave curvature feature of the edge of the gasification core layer in the preprocessed structure captured by optical imaging. The outer edge coordinates of the annular thermal boundary layer are established by combining the detection of temperature transition inflection points on the outer side of the annular thermal boundary layer using the infrared thermal imager. The width of the interlayer melting transition zone is determined based on the difference in spatial overlap between the effective diameter of the gasification core layer and the outer edge coordinates of the annular thermal boundary layer.

4. The gallium nitride wafer edge chipping control dicing method according to claim 1, characterized in that, Generate the crack propagation degree Values, including: The real-time location temperature gradient dataset is derived from the dynamic thermal field distribution of the cut area captured by the infrared thermal imager. Input the location temperature gradient dataset into the heat conduction model with the annular thermal boundary layer temperature gradient in the preprocessed structure as the constraint boundary. The heat conduction model is coupled with the displacement rate of change field generated by the optically integrated single-blade cutting motion for spatiotemporal field iteration; Output the crack propagation degree, which characterizes the material response state of the cutting region during the spatiotemporal field iteration. Spatial distribution of values.

5. The gallium nitride wafer edge chipping control dicing method according to claim 4, characterized in that, Establish a heat conduction model, including: The heat flux density control conditions of the heat conduction model are configured using the temperature gradient of the annular thermal boundary layer as a spatial boundary constraint. The dynamic thermal field distribution of the cutting area at the moment when the optically integrated single blade cuts into the pre-processed structure is set as the initial state of heat conduction. Based on the phase transition characteristics of the gallium nitride material in the high-temperature cutting zone, a heat conduction mechanism conversion rule is applied. The update cycle of the dynamic thermal field distribution of the infrared thermal imager is set with a time iteration step size to achieve synchronization with the cutting motion.

6. The gallium nitride wafer edge chipping control dicing method according to claim 1, characterized in that, The optical integration of the integrated microlens array and the 45° reflector is controlled by a single blade cutting into the pre-processed structure, including: The cutting laser beam output by the optically integrated single blade is focused by the microlens array and then vertically aligned with the center of the vaporization core layer to initialize the cutting stroke. Adjust the deflection angle of the 45° reflector according to the cutting path so that the reflected light spot always leads and covers the unprocessed area in the cutting direction. Based on the lattice energy distribution characteristics generated by the microlens array, an orientation-adaptive cutting trajectory is generated at the multi-domain boundary. The mechanical pressure gradient of the optically integrated blade is adjusted according to the radial temperature decay characteristics of the annular thermal boundary layer.

7. The gallium nitride wafer edge chipping control dicing method according to claim 6, characterized in that, The auxiliary laser is guided to the cutting front edge by the 45° reflector for real-time material softening, including: The auxiliary laser beam that deflects the 45° reflector is focused onto the cutting surface in front of the direction of travel of the cutting laser beam to form a pre-softened spot; The shape and contour of the pre-softened spot are adjusted according to the lattice orientation change of the adaptive cutting trajectory. The energy density range of the pre-softened light spot is adjusted based on the temperature distribution of the annular thermal boundary layer. The distance between the pre-softened spot and the edge of the vaporized core layer is corrected based on the cutting stroke feedback of the cutting laser beam.

8. The gallium nitride wafer edge chipping control dicing method according to claim 6, characterized in that, Generate crystal orientation-adaptive infeed trajectories, including: Scanning at the polydomain boundaries to obtain micro-region grain boundary distribution maps and identify the orientation of the main cleavage planes; A micro-insulation groove structure is constructed by utilizing the lattice energy distribution characteristics of the microlens array along the orientation extension direction of the main cleavage surface. The projection component of the cutting vector trajectory on the rhombic dodecahedral sliding system is constrained by the real-time numerical constraint of the mechanical pressure gradient. The process feedback based on the cutting stroke dynamically compensates for the trajectory offset caused by the difference in elastic modulus between adjacent grains.

9. A gallium nitride wafer edge chipping control dicing system, characterized in that, The system includes: The preprocessing module, based on a coaxial dual-focus laser system, simultaneously generates a preprocessing structure including a vaporized core layer and an annular thermal boundary layer in the wafer dicing area; The single-blade control module controls the optically integrated single-blade cutting pre-processing structure that integrates a microlens array and a 45° reflector, while simultaneously guiding the auxiliary laser to the cutting front edge for real-time material softening based on the 45° reflector. The value calculation module calculates the crack propagation β value by inverting the thermal diffusion equation based on the dynamic thermal field distribution of the cutting area captured by the infrared thermal imager. The micro-shaping module emits repair laser pulses towards the cutting front to perform micro-shaping when the crack propagation β value is greater than 0.25 mm. The energy adjustment module dynamically adjusts the spatial ratio of the pulse energy of the coaxial dual-focus laser system based on micro-shaping feedback until the cutting operation is completed.

10. The gallium nitride wafer edge chipping control dicing system according to claim 9, characterized in that, The The value calculation module includes: Temperature gradient unit: Real-time location temperature gradient dataset is obtained by analyzing the dynamic thermal field distribution of the cut area captured by infrared thermal imager. The model building unit inputs the location temperature gradient dataset into a heat conduction model that uses the temperature gradient of the annular thermal boundary layer in the preprocessed structure as the constraint boundary. The spatiotemporal iteration unit couples the heat conduction model with the displacement rate of change field generated by the optically integrated single-blade cutting motion to perform spatiotemporal field iteration; The value characterization unit outputs the crack propagation degree, which characterizes the material response state of the cutting region during spatiotemporal field iteration. Spatial distribution of values.

Images