Wafer heavy scratch three-dimensional shape repairing process based on DFG 8560 single-axis accurate grinding mode

Abstract

The invention particularly discloses a wafer heavy scratch three-dimensional shape repairing process based on a DFG 8560 single-axis accurate grinding mode, and aims to solve the problems that deep subsurface damage is introduced by rough grinding in a traditional double-stage grinding process, the machining efficiency is low, and material waste is serious. The process comprises the following steps: identifying a heavy scratch defect existing on the surface of a wafer through an automatic optical detection system; a ceramic bond diamond grinding wheel with the particle size of 8000 is adopted, uniaxial accurate grinding is conducted through ultrapure water cooling under the optimized parameters that the rotating speed of a main shaft is 1300-2300 rpm, the rotating speed of a workbench is 1785-2100 rpm and the rotating speed ratio is 0.6-1.3, and three-dimensional shape repairing is directly completed by skipping rough grinding; the rotating speed and pressure of grinding are dynamically regulated and controlled in three stages, the total time is less than or equal to 135 seconds, and the removal amount is 6-12 microns. By the adoption of the technical scheme, the surface roughness Ra can be smaller than 0.5 nm, the subsurface damage layer can be smaller than 50 nm, the repair success rate is larger than 98.5%, the machining time is remarkably shortened by 35% or above, and the material removal amount is reduced by 40%.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of semiconductor manufacturing, and particularly relates to a wafer severe scratch three-dimensional morphology repair process based on a DFG 8560 single-axis fine grinding mode. BACKGROUND

[0002] In the back-end process of semiconductor manufacturing, wafer back thinning is a key step to realize device thinning, improve thermal management performance, and support advanced packaging. Wafer regeneration, as an important part of circular economy, requires surface defect repair and thickness reconstruction of used wafers to meet the stringent standards of modern production lines for material flatness, surface roughness, and subsurface integrity. Among them, severe scratch, as a macroscopic surface defect with large depth and irregular profile, accounts for about 15% of incoming wafers. Its existence not only affects the stability of subsequent film pasting, grinding, and polishing processes, but also may induce crack propagation or cause device failure, so efficient and low-damage repair methods are urgently needed.

[0003] However, the existing mainstream grinding equipment such as DISCO DFG 8560 generally adopts a "coarse grinding + fine grinding" two-stage fixed process. This mode performs well when dealing with uniform surfaces, but has significant limitations when facing local severe scratches. To completely remove deep scratches, the traditional process is forced to increase the coarse grinding removal amount overall, not only causing unnecessary material loss, but also introducing a deeper subsurface damage layer due to the large particle size and strong cutting force of the coarse grinding wheel, significantly increasing the burden of subsequent polishing. In addition, even if some equipment has high-precision motion control or online thickness detection capabilities, its process logic is still limited to global uniform thinning, lacking sensing and response mechanisms for scratch three-dimensional morphology (including depth, width, and edge steepness), and cannot achieve adaptive repair of defect areas in a single operation.

[0004] After searching, a wafer thinning and polishing device with publication number CN114473822B was found, published on July 5, 2022. This patent provides an integrated device that integrates the functions of glue sticking, transportation, and thinning and polishing. Its thinning unit includes a coarse grinding disc and a fine grinding disc, which can sequentially perform coarse grinding and fine grinding on the wafer, and monitor the thickness through a laser ranging sensor. However, this scheme still follows the traditional "coarse first and fine second" two-stage fixed process and does not differentiate between local severe scratches on the wafer surface. When facing deep scratches, in order to ensure complete removal of defects, the coarse grinding removal amount must be increased overall, resulting in unnecessary material loss and deepening of the subsurface damage layer, which cannot achieve adaptive repair of the three-dimensional morphology of the severe scratch area, and also cannot balance efficiency and surface quality.

[0005] According to the search, a wafer processing equipment with publication number CN114454024B was disclosed on April 25, 2023. This patent proposes a wafer grinding system with high-precision XYZ linear motion module, multi-turn rotation capability of rotary table, and online 3D displacement detection, supporting grinding force monitoring and real-time error compensation, suitable for high-precision edge grinding and centering control. Although this equipment has advantages in positioning accuracy and process feedback, its core still focuses on conventional grinding or edge finishing scenarios, and does not propose a special repair strategy for wafer backside re-scratch defects. Especially lacking a three-dimensional morphology (such as depth, width, and profile) identification-response-repair closed-loop mechanism based on single-axis fine grinding mode, it cannot achieve efficient repair of re-scratches directly through dynamic adjustment of fine grinding parameters while avoiding the introduction of deep damage by rough grinding.

[0006] Therefore, there is an urgent need for a new process that can skip the rough grinding stage and directly repair re-scratches based on fine grinding mode with three-dimensional morphology adaptation. This process should be able to avoid introducing additional mechanical damage while optimizing fine grinding parameters (such as grinding wheel granularity, spindle and workbench speed matching, cooling conditions, etc.) to accurately remove scratch protrusions and smooth the transition of the surrounding area, thereby shortening the processing cycle while significantly improving the surface quality and yield of regenerated wafers. SUMMARY

[0007] The purpose of the present invention is to provide a wafer re-scratch three-dimensional morphology repair process based on DFG 8560 single-axis fine grinding mode, which effectively solves the technical bottlenecks of deep subsurface damage, low processing efficiency, and poor surface continuity in existing regeneration processes by integrating three unique technical features: "skipping the traditional rough grinding stage", "staged fine grinding strategy", and "cooling system optimization", combined with mathematical modeling and control algorithms of key process parameters.

[0008] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is as follows:

[0009] A wafer re-scratch three-dimensional morphology repair process based on DFG 8560 single-axis fine grinding mode, comprising the following specific steps:

[0010] Step 1: Defect detection and evaluation

[0011] The state of heavy scratch defects existing on the wafer surface is identified by an automated optical detection system, which is configured with a high-resolution camera and a structured light source, and can obtain three-dimensional topographic data of the wafer surface; based on the three-dimensional topographic data, a scratch with a depth of 3-8 μm, a width of 50-200 μm and an asymmetric steep feature of edge profile is automatically identified by a preset algorithm, and is marked as a target defect that needs three-dimensional topographic repair, the algorithm first calculates the depth and width of the scratch based on the three-dimensional topographic data; secondly, the 'asymmetric steep feature' is quantified by calculating the slope gradient difference of the scratch profile on the left and right sides, and when the difference is greater than a preset threshold K, the scratch is determined as a target defect that needs to be repaired.

[0012] Step 2: Process parameter optimization and algorithm modeling

[0013] For the three-dimensional topographic features of heavy scratch defects, an optimized process parameter combination suitable for single-axis fine grinding operation is set, and the following core technical innovations are introduced:

[0014] (1) Skip the traditional coarse grinding stage: abandon the conventional 'coarse grinding + fine grinding' two-stage process, and directly enable the high-precision fine grinding unit to handle heavy scratches, avoiding deep subsurface damage introduced by coarse grinding. This strategy is based on the following material removal amount constraints:

[0015] , wherein 6 μm≤T remove ≤12 μm, wherein D scratch is the maximum measured scratch depth, = 2-4 μm is a safety margin to ensure complete elimination of defects while minimizing material loss.

[0016] (2) Staged fine grinding strategy: The fine grinding process is divided into three dynamically controlled stages, each stage using a different spindle / worktable speed combination to achieve a gradual repair from rapid passivation to smooth transition to super-smooth final grinding. The running time and speed matching of each stage are controlled by the following formula:

[0017] , wherein (R(t) = (N spindle , N table ) represents the spindle speed and worktable speed (unit: rpm) at time t. The total grinding time T total ≤ 135 s.

[0018] (3) Cooling system optimization: High-purity ultrapure water (resistivity > 18 MΩ·cm, pH = 6.8-7.2) is used as the cooling medium, which is precisely covered on the wheel-wafer contact area by nozzles at an incident angle of 35°-45°, forming a stable liquid film. The local temperature rise △T local satisfies:

[0019] wherein is the frictional heat power, c is the specific heat capacity, p is the density, = 4-6L / min is the water flow rate, ensuring that the warping caused by thermal stress is effectively inhibited.

[0020] The grinding tool adopts a ceramic-bonded diamond grinding wheel with a particle size of 8000 (average abrasive particle size 2 pm, binder hardness H level), and the wheel linear speed v s is controlled at 18-25 m / s, satisfying:

[0021] (D: grinding wheel diameter, unit mm),

[0022] The spindle-to-table speed ratio is controlled within the range of 0.6-1.3; when D scratch > 5 pm, it is preferable that η = 1800 / 1950 ≈ 0.923, so as to form a high-density spiral superimposed track and improve the local material removal efficiency.

[0023] Step 3: Perform single-axis fine grinding

[0024] The wafer to be processed is firmly fixed on the worktable of the DISCO DFG 8560 grinding equipment, and the corresponding stage parameters are loaded according to the above algorithm model, and a one-time continuous fine grinding operation is started. The grinding pressure adopts constant force control mode, dynamically maintained at 30-50 kPa, to ensure uniform contact pressure and prevent corrugation or overgrinding.

[0025] Step 4: Quality verification of repair

[0026] (1) Surface topography detection: use a white light interferometer to scan the repair area, requiring:

[0027] Surface roughness R a < 0.5 nm,

[0028] Height difference Δh between scratch area and periphery < 1 nm,

[0029] Edge slope change rate ,

[0030] (2) Subsurface damage evaluation: through cross-sectional transmission electron microscopy (TEM) observation, the subsurface damage layer depth d sub < 50 nm, without dislocation network or microcrack propagation, reduced by more than 60% compared to traditional processes.

[0031] (3) Process stability verification: 100 consecutive wafers containing heavy scratches were subjected to the process, and the results showed:

[0032] Single wafer processing time ≤135s (35.7% shorter than traditional 210s)

[0033] Total material removal reduced by 40%,

[0034] Defect repair success rate >98.5%,

[0035] Subsequent polishing time of regenerated wafers reduced by an average of 35%,

[0036] Overall yield ≥97%

[0037] The process is suitable for 200mm or 300mm silicon-based regenerated wafers, especially for the case of local heavy scratches after multiple uses, without the need to add coarse grinding equipment, and only using the existing fine grinding unit DFG 8560 to complete high-precision repair, realizing deep functional exploration of equipment and process simplification.

[0038] Compared with the prior art, the present application first integrates the three major innovative features of "skipping coarse grinding", "staged fine grinding" and "algorithmic control of cooling system" into a single-axis fine grinding platform, realizes efficient, low-damage and high-continuity repair of three-dimensional topography of heavy scratches by establishing quantitative control models of material removal, rotational speed timing and thermal management. Experiments show that the surface roughness Ra after repair is <0.5 nm, the subsurface damage is <50 nm, the processing time is ≤135s, and the material utilization rate is significantly improved. The process is fully compatible with the existing DFG 8560 equipment, has outstanding industrialization value, and provides a new paradigm with precision, efficiency and economy for semiconductor regenerated wafer manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is the overall technical scheme architecture schematic diagram of the wafer heavy scratch three-dimensional topography repair process based on the DFG 8560 single-axis fine grinding mode of the present application;

[0040] Figure 2 is the core principle framework schematic diagram of skipping the traditional coarse grinding stage and combining the staged fine grinding strategy and the cooling system optimization in the present application;

[0041] Figure 3 is the logic flow framework diagram of the staged fine grinding strategy in the present application;

[0042] Figure 4 is the multi-level interaction relationship and data flow schematic diagram of the cooling system optimization and thermal management control algorithm in the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with specific embodiments.

[0044] Currently, in the back-end process of semiconductor manufacturing, wafer back thinning is a key step to achieve device thinning, improve thermal management performance and support advanced packaging. Wafer regeneration, as an important part of circular economy, requires surface defect repair and thickness reconstruction of used wafers to meet the stringent standards of modern production lines for material flatness, surface roughness and subsurface integrity. Among them, the re-scratch, as a macroscopic surface defect with large depth and irregular profile, accounts for about 15% of incoming wafers. Its existence not only affects the stability of subsequent film pasting, grinding and polishing processes, but also may induce crack propagation or cause device failure, so efficient and low-damage repair methods are urgently needed. In view of the above technical problems, the present application proposes three technical features of "skipping the traditional coarse grinding stage", "staged fine grinding strategy" and "cooling system optimization", combined with mathematical modeling and control algorithm of key process parameters, to effectively solve the technical bottlenecks of subsurface damage depth, low processing efficiency and poor surface continuity in existing regeneration processes, and applied to wafer re-scratch three-dimensional topography repair process based on DFG 8560 single-axis fine grinding mode.

[0045] Reference is made to the accompanying drawings Figure 1 The overall technical scheme architecture of the wafer re-scratch three-dimensional topography repair process based on DFG 8560 single-axis fine grinding mode proposed by the present application includes a defect detection and evaluation module, a process parameter optimization and algorithm modeling module, a single-axis fine grinding execution module, and a repair quality verification module. Each module works together to form a closed-loop control process, ensuring that the re-scratch area is repaired efficiently and accurately without introducing additional mechanical damage.

[0046] In the above wafer re-scratch three-dimensional topography repair process based on DFG 8560 single-axis fine grinding mode, step (1) identifies the re-scratch defect state existing on the wafer surface through an automatic optical detection system, which is configured with a high-resolution camera and a structured light source, and can obtain three-dimensional topography data of the wafer surface; based on the three-dimensional topography data, a preset algorithm automatically identifies scratches with a depth of 3-8 μm, a width of 50-200 μm and an asymmetric steep edge profile, and marks them as target defects that need three-dimensional topography repair, and records their position coordinates, maximum depth D scratch and profile characteristics, the algorithm first calculates the depth and width of the scratch based on the three-dimensional topography data; secondly, it quantifies the 'asymmetric steep feature' by calculating the slope gradient difference of the scratch profile on both sides, and when the difference is greater than a preset threshold K, it is determined that the scratch is a target defect that needs to be repaired.

[0047] In the wafer scratch three-dimensional morphology repair process based on the DFG 8560 single-axis fine grinding mode, the step (2) sets an optimized process parameter combination suitable for single-axis fine grinding operation according to the three-dimensional morphology characteristics of the scratch defect. This step is the core of the present application, which includes three strongly correlated technical sub-features that must be implemented simultaneously to achieve the expected technical effect.

[0048] Specifically, the first sub-feature in step (2) is to skip the traditional coarse grinding stage. This strategy discards the conventional "coarse grinding + fine grinding" two-stage process and directly enables a high-precision fine grinding unit to handle the scratch, thereby avoiding the deep subsurface damage introduced by the coarse grinding wheel due to its large particle size and strong cutting force. The implementation of this strategy strictly follows the material removal amount constraint condition, i.e., the total removal thickness T remove The measured maximum scratch depth D scratch is determined together with the safety margin ΔT, and the mathematical expression is:

[0049] where 6 μm ≤ T remove ≤ 12 μm, where D scratch is measured accurately by a contact profilometer or a white light interferometer at the defect position marked in step (1); ΔT is 2-4 μm, which is used to compensate for measurement errors and ensure that the scratch bottom is completely removed. For example, if the measured maximum depth of a scratch is 7.2 μm, T remove can be set to 9.5 μm, which satisfies the constraint condition and minimizes unnecessary material loss. This strategy allows the entire repair process to be completed within the fine grinding unit, fully utilizing the existing high-precision motion control capabilities of the DFG 8560 device, without the need to start the coarse grinding unit, significantly simplifying the process flow.

[0050] Specifically, the second sub-feature in step (2) is the staged fine grinding strategy. This strategy divides the fine grinding process into three dynamically regulated stages, each using a different spindle / worktable speed combination to achieve gradual repair from fast passivation to smooth transition to super-smooth final grinding.

[0051] Referring to the attached Figure 3 , the logical flow framework diagram clearly shows the timing division and parameter switching logic of the three stages. The running time of each stage and the speed matching are precisely controlled by the following formula:

[0052] , where (R(t) = (N spindle , N table ) represents the spindle speed and worktable speed at time t (units: rpm). The total grinding time T total≤ 135s. The first stage (0-30s) adopts high spindle speed and relatively low table speed (2100 / 1800 rpm), forming high linear speed and high cutting frequency, aiming to quickly remove the scratch protrusions and achieve preliminary blunting. The second stage (30-90s) reduces the spindle speed and increases the table speed (1600 / 2000 rpm), increasing the relative motion trajectory density, smoothing the rough surface formed in the first stage, and eliminating the corrugation. The third stage (90-135s) further reduces the spindle speed to 1300 rpm, while increasing the table speed to 2100 rpm, at which the spindle-to-table speed ratio η drops to about 0.619, forming extremely dense spiral superimposed trajectories, cooperating with extremely fine abrasive grains, achieving ultra-smooth final grinding, and ensuring the high continuity of the repair area and the surrounding substrate. This three-stage strategy solves the contradiction between efficiency and surface quality that traditional single-parameter fine grinding cannot achieve by precisely controlling the timing and dividing the single fine grinding process into a micro-operation sequence with clear functions.

[0053] Specifically, the third sub-feature in step (2) is cooling system optimization. Referring to the accompanying drawings Figure 4 The multi-level interaction relationship and data flow diagram reveal the feedback mechanism between the cooling medium parameters, thermodynamic model, and device control unit. The present application uses high-purity ultrapure water as the cooling medium, with a resistivity greater than 18 MΩ·cm and a pH value controlled within the range of 6.8-7.2 to minimize ion contamination. The cooling water precisely covers the grinding wheel-wafer contact area through a special nozzle at an incident angle of 35°-45°, forming a stable and continuous liquid film that effectively isolates the friction heat and flushes the grinding dust. The local temperature rise ΔT local is strictly controlled and needs to meet the following thermal management control algorithm:

[0054] wherein, is the friction heat power, determined by the grinding pressure, relative speed, and material properties; c is the specific heat capacity of water, with a value of 4186 J / (kg·K); ρ is the density of water, with a value of 1000 kg / m³; Q flow is the water flow rate, set within the range of 4-6 L / min. T ambient is the ambient temperature, T inlet is the cooling water inlet temperature. By monitoring in real time or presetting Q friction and dynamically adjusting Q flow , the local temperature rise in the contact area can be kept below 5℃, effectively inhibiting wafer warping caused by thermal stress and ensuring the geometric fidelity of the repair topography.

[0055] In addition, the grinding tool adopts a ceramic bond diamond grinding wheel with a particle size of 8000, an average abrasive particle size of 2 μm, and a binder hardness of H level. The wheel linear speed v_s is controlled in the safe and efficient interval of 18-25 m / s, and the calculation formula is (D: grinding wheel diameter, unit mm), the spindle and worktable speed ratio is controlled in the range of 0.6-1.3; when D scratch > 5 μm, η=1800 / 1950≈0.923 is preferred, and the specific ratio can form high-density, non-missing spiral superposition tracks in the scratch area, significantly improve the local material removal efficiency, and ensure that the deep scratches are thoroughly repaired.

[0056] In the wafer re-scratch three-dimensional topography repair process based on the DFG 8560 single-spindle fine grinding mode described above, the step (3) performs single-spindle fine grinding. Referring to the attached Figure 2 , the core principle framework integrates three major elements of skipping coarse grinding, staged fine grinding, and cooling optimization. The wafer to be processed evaluated and marked in step (1) is firmly fixed on the worktable of DISCO DFG 8560 grinding equipment through vacuum adsorption or wax sticking, ensuring no displacement or vibration during the entire 135 s fine grinding process. The device control system automatically loads the spindle speed, worktable speed, grinding time, and cooling water flow parameters corresponding to the three stages according to the algorithm model generated in step (2). Start one-time continuous fine grinding operation, and the grinding pressure adopts constant force control mode during the operation, which is dynamically maintained in the set range of 30-50 kPa through high-response pneumatic or electric actuators. This constant force control ensures uniform distribution of pressure on the contact surface between the grinding wheel and the wafer, preventing over-grinding or new ripple defects caused by excessive local pressure, while avoiding incomplete repair caused by insufficient pressure.

[0057] In the wafer re-scratch three-dimensional topography repair process based on the DFG 8560 single-spindle fine grinding mode described above, the step (4) verifies the repair quality. This step includes quantitative evaluation in three dimensions. First, surface topography detection is performed, and a white light interferometer is used to scan the repaired area at high resolution. The detection results need to meet three hard indicators: surface roughness Ra<0.5 nm, height difference Δh between the scratch area and the surrounding untreated area <1 nm, and slope change rate of the edge of the repaired area <0.1%. These three indicators together ensure the super-smoothness, high continuity, and smooth transition of the repaired surface. Second, sub-surface damage evaluation is performed, cross-section samples are prepared by focused ion beam (FIB), and observations are made using transmission electron microscopy (TEM). The evaluation standard is the sub-surface damage layer depth d subless than 50 nm, and there is no dislocation network or micro-crack propagation phenomenon in this depth range. Experimental data show that, compared with the traditional "rough grinding + fine grinding" process, the subsurface damage depth is reduced by more than 60%. Finally, the process stability verification is carried out, and the repair process is carried out on 100 continuous 200 mm or 300 mm silicon-based regeneration wafers containing typical heavy scratches. The statistical results show that: the single piece processing time is not more than 135s, which is 35.7% shorter than the traditional 210s process; the total amount of material removal is reduced by an average of 40%; the defect repair success rate is higher than 98.5%; when the regeneration wafer enters the subsequent polishing process, the polishing time is reduced by an average of 35%; and the final overall yield reaches 97% or higher. These data fully prove the comprehensive advantages of the process in efficiency, economy and reliability.

[0058] To further illustrate the practical application of the present application, a specific application example is constructed. A 300 mm diameter regeneration silicon wafer is selected, and a typical heavy scratch exists on the back surface after use. After step (1) detection, the maximum depth D scratch of the scratch is 6.8 μm, and the width is 120 μm. According to the algorithm model of step (2), T remove = 6.8+3.2 =10.0 μm is calculated. A ceramic bond diamond grinding wheel with a particle size of 8000 (D=300 mm) is selected, the cooling water flow rate Q flow is set to 5 L / min, and the pH is 7.0. In step (3), the equipment performs three-stage fine grinding according to the preset program: running at 2100 / 1800 rpm for the first 30 s, switching to 1600 / 2000 rpm for the next 60 s, and completing the final grinding at 1300 / 2100 rpm for the last 45 s, with a constant pressure of 40 kPa throughout the process. After processing is completed, step (4) verification is performed, and the white light interferometer scanning shows that the repair area Ra=0.42 nm, Δh=0.8 nm, =0.042 μm / μm; TEM observation shows that d sub =42 nm, and there is no micro-crack; the entire processing time is 132 s. This example fully demonstrates the whole process of the present application from defect identification to high-quality repair, and all the technical features work together to achieve the expected technical effect.

[0059] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode, characterized in that: The specific steps include the following: Step 1: Defect detection and assessment. The automated optical inspection system is used to identify the state of severe scratches on the wafer surface. The automated optical inspection system is equipped with a high-resolution camera and a structured light source, which can acquire three-dimensional morphological data of the wafer surface. Based on the three-dimensional topography data, a preset algorithm automatically identifies scratches with a depth of 3–8 μm, a width of 50–200 μm, and an asymmetrical and steep edge profile, and marks them as target defects that need to be repaired in three-dimensional topography. Step 2: Process parameter optimization. Based on the three-dimensional morphological characteristics of the heavy scratch defects, an optimized combination of process parameters suitable for single-axis precision grinding is set. The grinding tool uses a ceramic-bonded diamond grinding wheel with a grit size of 8000. By adjusting the speed ratio between the spindle and the worktable, the spacing between adjacent grinding tracks formed by the grinding wheel in the heavy scratch area is less than 1 / 5 of the scratch width, thereby forming a grinding track that can completely cover and densely intersect, so as to achieve a progressive and smooth removal of the scratch protruding edge. The cooling system uses ultrapure water as the cooling medium, with the water flow rate controlled at 4 to 6 liters per minute and the water temperature maintained in the range of 20 to 23 degrees Celsius, to ensure the thermal stability of the grinding area and effectively flush away and remove grinding debris. Step 3: Perform single-axis fine grinding. Securely fix the wafer to be processed on the worktable of the DISCO DFG 8560 grinding equipment. Skip the traditional rough grinding stage and directly use the fine grinding wheel with the above optimized parameters to perform a one-time continuous grinding operation on the back of the wafer. The grinding process continues until the heavy scratch defects are completely eliminated and the surface roughness reaches the preset standard to obtain a flat surface that meets the requirements of the regeneration process. The grinding process is carried out in stages. In the initial stage, the spindle speed is 2100 rpm and the table speed is 1800 rpm for 30 seconds. In the middle stage, the spindle speed is adjusted to 1600 rpm and the table speed is 2000 rpm for 60 seconds. In the final stage, the spindle speed is 1300 rpm and the table speed is 2100 rpm for 45 seconds. The total grinding time is controlled within 135 seconds.

2. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: The depth of the severe scratches ranges from 3 to 8 micrometers, and the width ranges from 50 to 200 micrometers, with an asymmetrical and steep edge profile.

3. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: The ceramic-bonded diamond grinding wheel with a particle size of 8000 has an average abrasive size of 2 micrometers, a bond hardness of H grade, and a grinding wheel linear speed controlled between 18 and 25 meters per second.

4. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: The ratio of the spindle speed to the table speed is controlled within the range of 0.6 to 1.3; when the depth of the heavy scratch is greater than 5 micrometers, the spindle speed is 1800 rpm and the table speed is 1950 rpm.

5. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: The ultrapure water has a resistivity greater than 18 megohm-cm and a pH value controlled between 6.8 and 7.

2. The coolant nozzle is aimed at the contact area between the grinding wheel and the wafer at an angle of 35 to 45 degrees.

6. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: The total removal amount of the single-axis fine grinding is controlled between 6 and 12 micrometers, and the grinding pressure is dynamically maintained between 30 and 50 kPa, using a constant force control mode.

7. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: It also includes a surface quality inspection step after grinding, which uses a white light interferometer to perform three-dimensional morphological scanning of the repaired area. The requirements are that the surface roughness Ra after repair is less than 0.5 nanometers, the height difference between the scratched area and the surrounding area is less than 1 nanometer, and the edge slope change rate is less than 0.05 micrometers per micrometer.

8. The wafer re-scrubbing three-dimensional morphology repair process based on DFG 8560 uniaxial fine grinding mode according to claim 1, characterized in that: It also includes a subsurface damage layer assessment step, which uses a cross-sectional transmission electron microscope to observe the subsurface structure of the polished wafer. The subsurface damage layer depth is required to be less than 50 nanometers, with no obvious dislocation network or microcrack extension.

Images