Method for optimizing 12-inch wafer fine grinding process based on sd6000-vbc14 grinding wheel

Abstract

The application discloses a 12-inch wafer fine grinding process optimization method based on an SD6000-vbc14 grinding wheel, and the method comprises the following steps: selecting an SD6000-vbc14 diamond grinding wheel with self-sharpening characteristics and long service life as a fine grinding wheel; configuring an adaptive device and setting an air cutting amount, a spindle speed and a table speed; adopting a multi-stage feeding strategy to perform fine grinding processing, removing macro fluctuations at a high speed in the first stage, accurately adjusting thickness distribution at a medium speed in the second stage, and completing finishing at a low speed in the third stage; performing light grinding processing after the fine grinding to release surface stress; after completion, the grinding wheel is separated from the wafer surface at a controlled escape speed to prevent edge cracking; and a closed-loop monitoring system is established to collect processing data in real time and dynamically adjust process parameters, so that parameter adaptive control is realized. Through the technical scheme, the technical effects that the service life of the grinding wheel is greatly prolonged, the processing cost is significantly reduced, the thickness tolerance after fine grinding is controlled within a preset range, and the yield of continuous batch processing is maintained above a preset yield threshold are achieved, and the comprehensive requirements of high yield, low cost and long service life for regenerative silicon wafer fine grinding are met.

Description

Technical Field

[0001] This application belongs to the field of semiconductor manufacturing technology, specifically relating to an optimization method for the fine grinding process of 12-inch wafers based on SD6000-vbc14 grinding wheels. Background Technology

[0002] As a core pillar of the modern information technology industry, the semiconductor manufacturing industry is constantly driving the evolution of integrated circuits towards smaller processes and higher integration levels through technological advancements and process innovations. As the fundamental material for integrated circuit manufacturing, the increase in wafer size significantly improves the chip yield per unit silicon area, thereby effectively reducing the manufacturing cost per chip. Under this trend, 12-inch (300mm) wafers have become the mainstream specification in the current semiconductor manufacturing field due to their significant advantages in cost-effectiveness and production efficiency. In the entire wafer manufacturing process, the precision grinding process, as a key step in achieving high surface quality and precise dimensional control of silicon wafers, directly determines the performance quality and production economy of the final silicon wafer product through the optimization of its process parameters and the appropriate selection of grinding wheels.

[0003] Among them, 6000-mesh high-precision diamond grinding wheels, with their ultra-fine abrasive grain size and uniform wear characteristics, are increasingly widely used in the fine grinding of 12-inch recycled silicon wafers. Recycled silicon wafers, through processes such as recycling, inspection, stripping, cleaning, fine grinding, and polishing of used wafers, achieve resource recycling and have become an effective way for the semiconductor industry to control costs and achieve green development. Fine grinding, as a core process in the recycled silicon wafer manufacturing process, must ensure the removal of scratches, contamination layers, and residual damage layers from previous processes on the wafer surface, while controlling the silicon wafer thickness tolerance within ±2μm, the total thickness variation (TTV) ≤5μm, and ensuring a yield of over 90% in mass production. This places extremely high demands on the grinding performance of the grinding wheel and the precise matching of process parameters.

[0004] In existing technologies, research on optimizing silicon wafer grinding processes mainly focuses on morphology prediction models, equipment structure improvements, and high-temperature heat treatment processes. However, analysis reveals the following systemic shortcomings in these technologies: First, the neural network-based silicon wafer morphology prediction model relies heavily on training and accumulating historical data. Its adaptability to specific process parameters of new high-grit precision grinding wheels like the 6000-grit wheel is insufficient. It cannot perform targeted process optimization for the abrasive characteristics of the SD6000-vbc14 wheel (e.g., approximately 2μm particle size, wear primarily consisting of microscopic disintegration), nor does it fully consider the impact of the chaotic characteristics of recycled silicon wafers (various films, damage) on grinding stability. Furthermore, it lacks an optimization strategy that balances wheel lifespan and processing costs. Second, for 300mm silicon wafer polishing... High-temperature heat treatment processes in wafer manufacturing primarily focus on polishing wafer quality and process simplification. There is a lack of systematic research on the synergistic optimization of grinding wheel parameters with silicon wafer surface quality and dimensional accuracy in the fine grinding process. This fails to meet the specific requirements of reclaimed silicon wafer fine grinding for grinding wheel lifespan (continuous processing of over 10,000 Pcs) and processing costs (reduction of over 20%). Furthermore, adjustable grinding devices based on displacement sensors and hydraulic control circuits mainly focus on improving equipment structure, correcting surface errors through post-processing adjustments rather than actively optimizing process parameters during grinding. Moreover, they do not involve research on the grinding characteristics of specific grinding wheel types, thus failing to fully leverage the advantages of high-mesh grinding wheels in the fine grinding of reclaimed silicon wafers. In summary, existing silicon wafer grinding process optimization methods lack systematic research on process parameters for 6000-grit high-precision grinding wheels, fail to fully consider the impact of recycled silicon wafer characteristics on grinding stability, and fail to comprehensively balance the relationship between surface quality, dimensional accuracy, grinding wheel life, and processing cost. These methods are insufficient to meet the comprehensive requirements of high yield, long life, and low cost for 12-inch recycled silicon wafer precision grinding, necessitating targeted research and optimization at the process method level. The purpose of this invention is to provide a wafer precision grinding process optimization method based on the SD6000-vbc14 grinding wheel, which can effectively solve the technical problems of existing silicon wafer grinding process optimization methods lacking systematic research on process parameters for high-precision grinding wheels, failing to fully consider the impact of recycled silicon wafer characteristics on grinding stability, and failing to comprehensively balance the relationship between surface quality, dimensional accuracy, grinding wheel life, and processing cost. Summary of the Invention

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] Step 1: Grinding Wheel Selection: An SD6000-vbc14 diamond grinding wheel was selected as the finishing grinding wheel. The key parameters of this wheel are: abrasive grain size of 2μm, wheel matrix material of aluminum alloy (grade 7075), resin-based bond type, wheel concentration of 150%, wheel dimensions of outer diameter φ400mm, inner diameter φ127mm, abrasive layer thickness of 5mm, and total thickness of wheel matrix and abrasive layer of 15mm. Its wear mechanism is mainly microscopic disintegration, exhibiting self-sharpening properties.

[0007] Step 2, Equipment Configuration and Parameter Setting: Use a DFG-8560 type or equivalent / higher precision grinding machine with spindle radial runout ≤0.5μm and table end face runout ≤0.3μm. Set the cut depth to 15μm using the equipment's CNC system. Specifically, based on the wafer's upper surface detection signal, control the grinding wheel feed axis to switch to the grinding feed speed 15μm before contacting the wafer. Simultaneously set the spindle speed to 2330rpm and the table speed to 234rpm.

[0008] Step 3, Multi-stage feed strategy: A three-stage decreasing feed strategy is adopted, and the total removal amount is controlled within the range of 4μm to 6μm.

[0009] First stage (high-speed removal): feed rate 0.35μm / s, removal amount 2.8μm to 3.8μm, used to remove macroscopic surface undulations and damage layers.

[0010] Second stage (medium speed adjustment): feed speed 0.30μm / s, removal amount 0.8μm to 1.8μm, used to precisely adjust thickness distribution.

[0011] The third stage (low-speed finishing): the feed rate is 0.2 μm / s, and the removal amount is 0.4 μm to 0.8 μm, which is used to complete the surface finishing and release stress.

[0012] Step 4, Polishing: After fine polishing, polishing is performed at 5 rev. The selection of this parameter is based on the following: Experiments show that when the polishing speed is less than 3 rev, the residual stress on the surface is not fully released, leading to a 15% increase in the defect rate of subsequent CMP processes; when the polishing speed exceeds 8 rev, the stress release effect tends to saturate, and each additional 1 rev will add an extra 8-10 seconds of processing time without significantly improving quality. 5 rev is the balance point between stress release efficiency and processing time.

[0013] Step 5, Controlled Escape: After processing, an escape velocity of 0.2 μm / s, the same as the feed rate in the third stage, is used to detach the grinding wheel from the wafer surface. Setting both values ​​identical ensures that the kinetic characteristics of the detachment process remain consistent with the final finishing stage, avoiding sudden speed changes that could introduce impact loads. Comparative experiments show that when the escape velocity is 0.5 μm / s, the wafer edge chipping rate increases to 1.2%; when the escape velocity is 0.1 μm / s, the chipping rate is not significantly different from that at 0.2 μm / s (both below 0.03%), but the single-wafer processing time increases by 15 seconds. Therefore, setting the escape velocity to 0.2 μm / s is the optimal choice.

[0014] Step 6: Closed-loop monitoring and adaptive control:

[0015] Hardware configuration: Two Keyence CL-3000 series spectral confocal displacement sensors are installed above the grinding equipment platform, with a measurement range of ±1mm and an accuracy of ±0.1μm, located at the wafer center and edge R / 2, respectively. The online surface roughness measurement device is a Zygo.

[0016] The Nexview white light interferometer is installed at the unloading arm of the equipment and measures a light spot diameter of 50μm.

[0017] Control algorithm: An incremental PID control strategy is adopted, with initial parameters of proportional coefficient P=0.8, integral coefficient I=0.1, and derivative coefficient D=0.15. The control cycle is adjusted once every 5 wafers processed.

[0018] Parameter mapping and triggering:

[0019] The triggering condition is evaluated every 5 wafers processed. When the thickness tolerance center value of 3 consecutive wafers deviates by more than ±0.5μm, the second stage feed endpoint position correction is triggered, and the correction amount ΔP2 = -0.8 * thickness deviation.

[0020] When the total thickness variation (TTV) exceeds 1.5μm for three consecutive pieces, the table speed is fine-tuned, with an adjustment step of ±3rpm.

[0021] When the surface roughness Ra exceeds 0.15 μm for three consecutive pieces, the third stage of feed speed adjustment is triggered, with an adjustment step of ±0.01 μm / s, and the polishing speed is adjusted simultaneously by ±1 rev.

[0022] Preferably, in step 1, the continuous processing capacity of the SD6000-vbc14 type diamond grinding wheel can reach more than a preset number of pieces, significantly extending the service life compared to traditional grinding wheels, reducing the cost of a single grinding wheel by a preset percentage, and greatly reducing the cost loss caused by replacing the grinding wheel when its service life expires.

[0023] Preferably, the setting of the empty cutting amount in step 2 is used to ensure that the wafer smoothly enters the grinding area from the free state, avoiding edge cracking caused by direct collision; the spindle speed and the table speed form a preset proportional relationship to form a stable relative motion force field, so that the grinding wheel abrasive grains contact the wafer surface layer by layer with micro-angle differences, avoiding surface texture defects caused by fixed trajectory grinding.

[0024] Preferably, in step 3, the multi-stage feed strategy achieves efficient material removal and finishing through progressively decreasing feed rates. The higher feed rate in the first stage is used to quickly remove macroscopic undulations on the wafer surface and the damage layer remaining from the previous process. The moderate feed rate in the second stage is used to accurately adjust the thickness distribution. The low feed rate in the third stage is used to complete the final finishing, ensuring uniform release of surface stress. The total removal amount in the three stages is controlled within a preset range.

[0025] Preferably, in step 4, the polishing process establishes a stable dynamic equilibrium between the abrasive grains of the grinding wheel and the wafer surface through continuous grinding contact at a preset number of revolutions. This fully releases the surface stress accumulated during the grinding process, reduces the density of microcracks in the surface layer, and provides a good starting point for the subsequent chemical mechanical planarization process.

[0026] Preferably, in step 5, the escape velocity and the polishing parameters work together to ensure that the grinding wheel smoothly leaves the wafer surface at a controlled rate after the fine grinding process is completed. This eliminates the adverse effects of the impact load generated by high-speed retraction on the integrity of the processed surface and prevents edge chipping due to stress concentration in the wafer edge area.

[0027] Preferably, the present invention also includes process stability control of the batch processing process. After each wafer is finished in the fine grinding process, the closed-loop monitoring system automatically collects thickness tolerance data, total thickness change value and surface roughness data, compares and analyzes the collected data with preset standards, and dynamically corrects the process parameters of the next wafer based on the comparison results, so as to realize the adaptive adjustment of parameters and ensure that the yield is maintained above the preset yield threshold when a preset number of wafers are continuously processed.

[0028] Preferably, the present invention supports setting processing parameters independently for each wafer to adapt to the differences in the characteristics of recycled silicon wafers and the special processing needs of customers, thereby improving the flexibility of the process and market adaptability, and meeting the requirements of the wafer manufacturing industry for high precision, high efficiency and low cost of the grinding process.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. Significantly extended grinding wheel lifespan: Replacing traditional grinding wheels with SD6000-vbc14 type diamond grinding wheels utilizes the self-sharpening mechanism of micro-abrasive grain disintegration to achieve continuous processing of more than 10,000 pieces. This represents a lifespan increase of over 40% compared to the traditional PW06 type grinding wheel (7,600 pieces), and the cost per grinding wheel is reduced by approximately 74%, effectively minimizing cost losses due to grinding wheel replacement at the end of its lifespan.

[0031] 2. Significantly reduced processing costs: Due to fewer processes and longer lifespan, high-mesh grinding wheels reduce the grinding cost of a single silicon wafer by about 22% compared to traditional solutions. The overall process solution significantly reduces the cost per wafer, meeting the urgent need in the recycled silicon wafer industry for low-cost, high-performance fine grinding processes.

[0032] 3. Stable and reliable processing quality: The average thickness tolerance of the silicon wafers after fine grinding can be controlled within ±1.2μm (standard products), and the average total thickness variation can be controlled within 1.1μm. There is no obvious mechanical damage layer on the surface. After chemical mechanical planarization repair, the particles can meet customer requirements. It is guaranteed that there are no grinding wheel residue scratches on the surface. The yield rate of continuous batch processing of 10,000 wafers is maintained at over 99.3%, which meets the comprehensive requirements of high yield and long life of regenerated silicon wafer fine grinding. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the overall technical solution architecture of the wafer finishing process optimization method based on SD6000-vbc14 grinding wheel according to an embodiment of this application;

[0034] Figure 2 This is a schematic diagram of the core principle framework of closed-loop monitoring and adaptive parameter control in the wafer finishing process optimization method based on SD6000-vbc14 grinding wheel according to an embodiment of this application.

[0035] Figure 3 This is a flowchart illustrating the logic of the multi-stage feed strategy in the wafer finishing process optimization method based on the SD6000-vbc14 grinding wheel according to an embodiment of this application. Detailed Implementation

[0036] Example 1

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0038] This invention discloses an optimization method for wafer grinding based on the SD6000-vbc14 grinding wheel, aiming to achieve high efficiency, high precision, and low cost synergistic optimization in the grinding process of 12-inch reclaimed silicon wafers. The technical architecture of this method includes core modules such as grinding wheel selection and characteristic verification, grinding equipment configuration and parameter setting, multi-stage feed strategy implementation, polishing treatment and stress control, and closed-loop monitoring and adaptive parameter adjustment. These modules form a complete technical closed loop, jointly supporting the stability, reliability, and economic requirements of reclaimed silicon wafer grinding.

[0039] In terms of grinding wheel selection and characteristic verification, this invention selects the SD6000-vbc14 diamond grinding wheel as the core grinding tool for the fine grinding process. This grinding wheel belongs to the 6000-mesh high-precision grade, with an abrasive grain size of approximately 2μm. High-quality diamond particles are bonded to the grinding wheel matrix through a specific process. From the perspective of wear mechanism analysis, the abrasive wear of the SD6000-vbc14 grinding wheel is mainly in the form of microscopic disintegration. That is, during continuous grinding, the abrasive grain surface is subjected to the alternating effects of periodic thermal stress and mechanical load, resulting in microscopic material peeling and recrystallization on the abrasive grain surface, forming a fresh cutting edge. This self-sharpening characteristic allows the grinding wheel to maintain a relatively stable grinding capability and sharpness throughout its service life, effectively avoiding the problems of increased grinding force and surface quality deterioration caused by abrasive grain passivation in traditional grinding wheels. Compared to the traditional PW06 type grinding wheel, which can only process 7,600 wafers continuously, the SD6000-vbc14 type grinding wheel can process more than 10,000 wafers continuously, with a service life increase of more than 40% and a cost reduction of about 74% per grinding wheel. The equipment downtime and cost losses caused by the replacement of grinding wheels at the end of their service life are significantly reduced.

[0040] Regarding the configuration and parameter settings of the precision grinding equipment, this invention uses the DFG-8560 precision grinding equipment as the bearing and motion control carrier for the grinding wheel. This equipment has a high-precision spindle drive system and a highly stable table rotation mechanism, which can meet the stringent requirements of the SD6000-vbc14 grinding wheel for speed control and position accuracy. Based on the grinding characteristics of the SD6000-vbc14 grinding wheel, this invention, through systematic research and development of process parameters, has determined the following key process parameter configurations: the cut depth is set to 15μm to ensure a smooth transition of the wafer from its free state to the grinding area, avoiding edge chipping or hidden damage caused by direct collision between the wafer edge and the high-speed rotating grinding wheel; the spindle speed is set to 2330rpm to drive the grinding wheel to rotate at high speed, forming a stable grinding linear velocity, ensuring that the abrasive grains can cut into the silicon wafer surface with sufficient kinetic energy and achieve effective material removal; the table speed is set to 234rpm, forming a specific proportional relationship with the spindle speed, creating a complex relative motion trajectory on the wafer surface, allowing the grinding wheel abrasive grains to contact the wafer surface layer by layer with micro-angle differences, avoiding surface periodic texture defects caused by fixed trajectory grinding. The design principle of the cut-off distance parameter is as follows: before entering the effective grinding area, the wafer needs to undergo a transition stroke of a safe distance. During this period, the wafer and the grinding wheel remain in a non-contact state. Driven by the table, the wafer gradually accelerates to a surface linear velocity that matches the linear velocity of the grinding wheel, thereby achieving a smooth transition at the grinding start point. If this parameter is set too small, the wafer edge may enter the grinding zone before reaching a stable rotational speed, generating impact loads that cause edge chipping. If this parameter is set too large, it will increase the processing time and ineffective stroke of a single wafer, reducing equipment utilization. Taking into account both grinding quality and production efficiency, this invention optimizes the cut-off distance to 15μm.

[0041] Regarding the implementation of multi-stage feed strategies Figure 3 As shown, this invention overcomes the limitations of traditional single-speed or dual-speed feed modes by employing a three-stage decreasing feed strategy to achieve efficient material removal and surface finishing. The design principle of this strategy is based on the inherent contradiction between material removal rate and surface integrity in grinding: faster feed speeds can improve material removal efficiency but easily introduce deeper damage layers and larger residual stresses into the processed surface; slower feed speeds can achieve excellent surface quality but significantly reduce processing efficiency and increase wheel wear. Addressing the dual requirements of rapidly removing residual damage layers from previous processes and ensuring final surface integrity in the reclaimed silicon wafer finishing process, this invention designs a progressively decreasing three-stage feed strategy, achieving dynamic optimization of the grinding process through staged speed changes.

[0042] The first-stage feed rate is set to 0.35 μm / s. The core task of this stage is to rapidly remove macroscopic undulations and residual damage layers from previous processes on the wafer surface. At this feed rate, the grinding wheel contacts the wafer surface with a relatively high material removal rate, and a significant thickness of silicon material can be removed in a single grinding cycle. After undergoing previous processes such as recycling inspection, demolding, and cleaning, reclaimed silicon wafers typically have damage of varying sources and depths on their surfaces, including scratches caused by customer equipment, residual particulate contamination, and stress concentration areas formed during previous use. The first-stage grinding, with its high removal rate parameters, can essentially eliminate these macroscopic defects in a short time, creating a good surface foundation for subsequent finishing processes. The total removal amount in this stage typically accounts for 60% to 70% of the total removal amount in the entire finishing process, with the specific value dynamically adjusted based on the degree of damage to the incoming material. For reclaimed silicon wafers with severe surface damage, the total removal amount in the first stage can be appropriately increased to ensure complete removal of all damaged layers; for silicon wafers with relatively good surface conditions, the removal amount can be appropriately reduced to minimize ineffective consumption of the substrate material.

[0043] The second-stage feed rate is set at 0.30 μm / s. The core task of this stage is to precisely adjust the thickness distribution and total thickness variation of the silicon wafer. After the first-stage roughing, the wafer surface is basically flat, but the thickness distribution may still have some macroscopic fluctuations, requiring further optimization of thickness uniformity through precise feed control. The feed rate in the second stage is reduced by approximately 14% compared to the first stage, significantly improving feed positioning accuracy and thickness control accuracy while maintaining a certain material removal efficiency. This stage employs the equipment's high-precision closed-loop position control system, which dynamically adjusts the feed position of the grinding wheel by monitoring the vertical displacement changes of the table in real time, ensuring that each wafer achieves a consistent thickness distribution when processed to the target thickness. The total removal volume in this stage typically accounts for 20% to 30% of the total removal volume in the entire finishing process, forming a reasonable removal volume distribution ratio with the first stage.

[0044] The third-stage feed rate is set at 0.2 μm / s. The core task of this stage is to complete the final surface finishing, ensuring uniform release of surface stress and providing an ideal starting point for subsequent polishing processes. This third-stage feed rate is the lowest of the three stages, allowing the abrasive grains to contact the wafer surface with an extremely fine depth of cut. This primarily achieves microplastic flow and elastic recovery of the surface layer material, rather than significant chip formation. The mechanism of this stage can be explained at the atomic scale: under extremely low feed rates, the interaction forces between the abrasive grains and the atoms on the silicon wafer surface are in a near-equilibrium state. Material removal is mainly characterized by atomic-level rearrangement of the surface layer, rather than the fracture removal of bulk material. This processing mode can form a dense amorphous or nanocrystalline surface layer on the wafer, effectively sealing the microcrack tips introduced by previous processing and reducing the residual tensile stress level of the surface layer. After the third-stage finishing, the stress state on the wafer surface tends to be uniformly distributed, providing a good starting point for subsequent chemical mechanical planarization processes.

[0045] The total removal amount of the three-stage feed strategy is determined comprehensively based on the product's technical requirements and the incoming material's condition. In a preferred embodiment of the invention, for a typical 12-inch recycled silicon wafer, the total removal amount across the three stages is controlled within the range of 4μm to 6μm. Insufficient removal may not completely eliminate the damaged layer from the preceding process, resulting in hidden defects in the final product; excessive removal will increase processing time and accelerate grinding wheel wear, reducing the economic efficiency of the process.

[0046] Regarding the finishing and stress control, this invention introduces a finishing process after the third stage of fine grinding. The finishing speed is set to 5 rev, meaning that after the grinding wheel completes the material removal action, it continues to maintain contact with the wafer for 5 revolutions. The design principle of this process is based on the establishment of dynamic equilibrium in grinding: in the conventional grinding stage, there is significant relative motion and material removal between the grinding wheel and the wafer, and the interaction force between them is in a dynamic state. Under the action of periodic loads, the wafer surface accumulates a certain degree of micro-stress and lattice distortion. In the finishing stage, by reducing or pausing the feed motion, a stable dynamic equilibrium is gradually established between the abrasive grains of the grinding wheel and the wafer surface. At this time, the contact pressure between the abrasive grains and the surface is reduced to near zero, and the wafer surface completes stress release and lattice relaxation in continuous low-energy contact. The effects of this process are twofold: first, it fully releases the surface stress accumulated during grinding, reduces the density of microcracks in the surface layer, and improves the structural integrity of the surface; second, it brings the wafer surface roughness to a dynamically stable state, providing a repeatable processing benchmark for subsequent inspection and polishing processes. The setting of the grinding revolutions needs to comprehensively consider the stress release effect and processing time cost: too few revolutions result in insufficient stress release, leaving high residual stress on the surface; too many revolutions prolong the processing time per wafer and increase unnecessary wear on the grinding wheel. Through process optimization and verification, this invention has determined that 5rev is the optimal parameter configuration for the grinding process.

[0047] Experimental verification of a polishing speed of 5 rev: During the process development phase, this invention conducted comparative experiments on different polishing speeds (3 rev, 5 rev, 8 rev). The residual stress on the silicon wafer surface was measured using an X-ray diffraction residual stress analyzer. The results showed that at a polishing speed of 3 rev, the surface residual stress was approximately 35 MPa; at 5 rev, the surface residual stress decreased to 12 MPa; and at 8 rev, the surface residual stress further decreased to 10 MPa, but the processing time per wafer increased by 25%. Considering both stress relief effect and production efficiency, 5 rev was determined to be the optimal polishing speed parameter.

[0048] Regarding escape velocity control and edge protection, this invention employs a controlled escape velocity to detach the grinding wheel from the wafer surface after the finishing process. The escape velocity is set at 0.2 μm / s, a value consistent with the feed rate in the third stage, ensuring that the grinding wheel gradually detaches from the wafer surface at a stable and controllable rate after finishing. The design principle of this parameter is to eliminate the adverse effects of impact loads generated by high-speed retraction on the integrity of the machined surface. In traditional grinding processes, if the grinding wheel is lifted vertically from the wafer surface at a relatively high rate after finishing, stress concentration may occur in the wafer edge region, leading to edge chipping or microcrack propagation. Furthermore, a momentary negative pressure environment may be formed between the grinding wheel and the wafer during rapid retraction, attracting particulate contaminants from the air to settle on the freshly machined surface, affecting surface cleanliness. This invention, by setting a lower escape velocity parameter, makes the separation process between the grinding wheel and the wafer gradual and stable, effectively mitigating the stress state in the edge region and avoiding edge damage caused by sudden stress changes. The escape velocity parameter and the aforementioned polishing parameter work synergistically: the polishing process ensures that the surface stress is fully released, and the escape velocity control ensures that the surface after stress release does not introduce new stress concentration points during the detachment process. Together, they constitute a complete surface protection mechanism.

[0049] The rationale for setting both the escape velocity and the third-stage feed rate at 0.2 μm / s is as follows: This invention conducted comparative experiments on different escape velocities (0.1 μm / s, 0.2 μm / s, and 0.5 μm / s). Experimental results show that: when the escape velocity is 0.5 μm / s, the wafer edge chipping rate increases to 1.2%; when the escape velocity is 0.1 μm / s, the single-wafer processing time increases by 8%, and the edge chipping rate is not significantly different from that at 0.2 μm / s (both are below 0.05%); when the escape velocity is 0.2 μm / s, the edge chipping rate is the lowest (0.02%), and the processing time is moderate. Therefore, the escape velocity is set to the same 0.2 μm / s as the third-stage feed rate to balance edge protection and production efficiency.

[0050] In terms of closed-loop monitoring and adaptive parameter adjustment, this invention establishes a quality monitoring and parameter optimization system for the entire fine grinding process. Figure 2As shown, the core functional modules of the system include a real-time data acquisition module, a data analysis and processing module, and a parameter adaptive adjustment module. These three modules interact and transmit commands through the communication interface of the equipment control system. After each wafer completes the precision grinding process, the real-time data acquisition module automatically collects key quality parameters of that wafer, including thickness tolerance data, total thickness variation, and surface roughness data. Thickness tolerance data is acquired using a high-precision non-contact displacement sensor installed inside the equipment, with a measurement accuracy on the order of 0.1 μm, accurately reflecting the thickness distribution of the wafer at different radii. The total thickness variation is obtained by collecting thickness data at multiple measurement points on the same wafer surface and calculating the difference between the maximum and minimum values, used to evaluate the wafer's ability to maintain thickness uniformity during the precision grinding process. Surface roughness data is acquired using an online optical measurement device, with measurement wavelengths covering the visible to near-infrared band, capable of characterizing the micron to nanometer-scale undulations of the wafer surface.

[0051] The specific configuration of the displacement sensors in the closed-loop monitoring system is as follows: Thickness tolerance data is acquired using a KEYENCE LK-G5000 series laser displacement sensor with a measurement accuracy of ±0.1μm. It is installed on the side of the spindle of the precision grinding equipment, at a vertical distance of 50mm from the wafer surface, with a sampling frequency of 50Hz. Surface roughness data is acquired using a ZYGO NewView 9000 series white light interferometer, installed at the equipment's outlet, with a measurement field of view of 0.5mm × 0.5mm, a lateral resolution of 0.5μm, and a vertical resolution of 0.1nm.

[0052] The data analysis and processing module preprocesses, extracts features, and performs standard comparison analysis on the collected raw data. The preprocessing stage includes data filtering, outlier removal, and measurement error compensation to ensure the reliability of the data used in subsequent analysis. The feature extraction stage calculates core indicators characterizing wafer processing quality from the raw data, including the mean and standard deviation of thickness distribution, and the arithmetic mean roughness and root mean square roughness of surface roughness. The standard comparison analysis stage compares the calculated quality indicators with preset product specification standards to determine whether the currently processed wafer meets the quality requirements. Product specification standards are set according to the technical specifications of different customers, including thickness tolerance range (typically ±2μm), upper limit of total thickness variation (typically 5μm), and upper limit of surface roughness (typically Ra 0.2μm). If any indicator exceeds the standard range, the system automatically marks the wafer and classifies it as an anomaly, providing a basis for subsequent quality traceability and process diagnosis.

[0053] The adaptive parameter adjustment module dynamically corrects the subsequent wafer process parameter configuration based on the output of the data analysis and processing module. The core algorithm of this module employs an incremental PID control strategy from feedback control theory. Based on the deviation between the current processing result and the target value, it calculates the parameter adjustment amount and sends it to the equipment control system for execution. Specifically, when a systematic shift in the thickness tolerance of multiple consecutive wafers is detected, the module automatically calculates the correction amount for the feed rates in stages P1, P2, and P3, eliminating thickness deviations by fine-tuning the feed endpoint position of subsequent wafers. When an increasing trend in the total thickness variation is detected, the module automatically adjusts the ratio between the spindle speed and the table speed, improving thickness uniformity by optimizing the relative motion mode between the grinding wheel and the wafer. When surface roughness deteriorates, the module automatically reduces the feed rate in the third stage and appropriately increases the polishing revolutions, restoring surface quality by optimizing the processing parameters in the finishing stage. This adaptive adjustment mechanism ensures that the process system can automatically maintain the stability of processing quality when facing fluctuations in incoming material characteristics, equipment status drift, and changes in environmental conditions, achieving online optimization of process parameters without manual intervention.

[0054] The specific parameters of the incremental PID control strategy are as follows: the proportional gain P is initially set to 0.8, the integral time I is initially set to 0.5 seconds, and the derivative time D is initially set to 0.1 seconds. The trigger condition for adaptive parameter adjustment is: the thickness tolerance of three consecutive wafers exceeds ±1.5 μm, or the total thickness variation of a single wafer exceeds 2.0 μm. The control outputs are the spindle speed correction, the table speed correction, and the feed rate correction for each stage. The mapping relationship is: spindle speed correction = thickness deviation × P; feed rate correction = total thickness variation deviation * I. The upper limit of the adjustment step size is: each spindle speed correction should not exceed 50 rpm, and each feed rate correction should not exceed 0.02 μm / s.

[0055] The parameter adaptive control mechanism of this invention also supports independent parameter setting for individual wafers to adapt to differences in the characteristics of recycled silicon wafers and special processing requirements of customers. Recycled silicon wafers from different sources may have significant differences in film type, damage level, and geometric accuracy, and the consistency of silicon wafers within the same batch is also difficult to guarantee completely. To address this characteristic, this invention adds an incoming material characteristic identification interface to the closed-loop monitoring system, allowing operators to set processing parameters individually for each wafer based on the incoming material inspection results. For example, for silicon wafers with severe surface damage, the first-stage feed rate and total removal amount can be appropriately increased to ensure complete removal of all damaged layers; for silicon wafers with good surface condition but large thickness deviations, the second-stage feed parameters can be adjusted to optimize thickness uniformity; for orders with special customer requirements, specific values ​​for various parameters can be set to meet customized processing needs. This function significantly improves the flexibility and market adaptability of the process system, enabling this invention to meet the comprehensive requirements of high precision, high efficiency, and low cost in the wafer manufacturing industry for the fine grinding process.

[0056] To verify the technical effectiveness of the process scheme disclosed in this invention, a set of specific application examples are provided below. This example uses an SD6000-vbc14 diamond grinding wheel on a DFG-8560 precision grinding machine to verify the batch processing of 12-inch reclaimed silicon wafers.

[0057] During the equipment configuration phase, the operator installs the grinding wheel on the spindle and performs dynamic balancing to ensure that the vibration amplitude of the grinding wheel at a speed of 2330 rpm is controlled within the allowable range. Subsequently, the standard process parameters for fine grinding are set through the equipment's operating interface: empty cut 15 μm, spindle speed 2330 rpm, table speed 234 rpm, finishing speed 5 rev, and escape velocity 0.2 μm / s. The three-stage feed rates are set as follows: first stage 0.35 μm / s, second stage 0.30 μm / s, and third stage 0.2 μm / s. This parameter set is stored as a standard reference in the equipment control system and can be called upon for subsequent processing or modified according to specific needs.

[0058] During the incoming material preparation stage, the recycled silicon wafers to be processed undergo incoming material inspection to obtain data on the film type, surface condition, and initial thickness of each wafer. The inspection results showed that this batch of silicon wafers was mainly composed of 12-inch monitor wafers, with scratches and stress damage layers generated during the use of the client's FAB equipment, with a damage depth of approximately 2μm to 3μm. Based on the inspection results, this batch of silicon wafers was divided into three groups: the first group had poor surface condition, and a higher first-stage removal rate was set to ensure complete elimination of the damage layer; the second group had moderate surface condition and was processed using standard parameter sets; the third group consisted of custom orders with special customer requirements, and the parameters were adjusted accordingly.

[0059] In the fine grinding stage, the silicon wafer is fixed to the equipment table, and the automatic processing cycle is started. The grinding wheel, driven by the spindle, accelerates to a stable speed of 2330 rpm, while the table rotates the silicon wafer to a process speed of 234 rpm. During the air-cutting stage, the grinding wheel quickly moves to the starting position on the outer side of the wafer surface, and then guides the wafer smoothly into the grinding area with a set air-cutting amount of 15 μm. The first stage feed is executed at a speed of 0.35 μm / s, with a total removal amount set at 3.5 μm, used to quickly remove scratches and coarse damage layers from the wafer surface. During processing, the equipment control system monitors the spindle current and feed force feedback signals in real time. When the monitored values ​​exceed the normal range, an alarm is automatically triggered and processing is paused to prevent the abnormal situation from worsening. After the first stage of processing is completed, the equipment automatically switches to the second stage feed, with the speed reduced to 0.30 μm / s. The total removal amount for this stage is set at 1.0 μm, used to precisely adjust the thickness distribution and total thickness variation. After the second stage of processing, the equipment automatically switches to the third stage feed, with the speed further reduced to 0.20 μm / s. The total removal amount in this stage is set to 0.5 μm to complete the final surface finishing. After the third stage of processing, the equipment performs a polishing process, maintaining the grinding wheel in contact with the wafer for 5 revolutions, and then smoothly removing the grinding wheel from the wafer surface at an escape velocity of 0.2 μm / s. The entire processing cycle is now complete. The polishing time for a single wafer is approximately 2.5 to 3 minutes, with the specific value determined based on the initial thickness and target thickness of the silicon wafer.

[0060] During the quality inspection phase, each wafer automatically enters the online inspection station after fine grinding. Inspection items include thickness distribution measurement, total thickness variation calculation, and surface roughness measurement. Measurement data is transmitted in real-time to a closed-loop monitoring system for storage and analysis via a data communication interface. For the first group of silicon wafers, the inspection results show that the average thickness tolerance after fine grinding is ±1.0μm, the average total thickness variation is 0.9μm, and the surface roughness meets customer requirements. For the second group of silicon wafers, the inspection results show that the average thickness tolerance after fine grinding is ±1.2μm, the average total thickness variation is 1.1μm, and the surface roughness meets customer requirements. For the third group of custom-ordered silicon wafers, the inspection results show that all quality indicators after fine grinding meet the customer's specific requirements. All three groups of silicon wafers showed no obvious mechanical damage layer after fine grinding, with a good surface condition, providing an ideal starting point for subsequent chemical mechanical planarization processes.

[0061] During the batch verification phase, continuous processing life tests were conducted on the SD6000-vbc14 grinding wheel. The tests used standard parameter sets to continuously process reclaimed silicon wafers, recording quality data and checking the wheel's condition every 500 wafers. Test results showed that after processing 10,000 wafers continuously, the SD6000-vbc14 grinding wheel maintained stable processing quality, with no significant deterioration in thickness tolerance or total thickness variation. Visual inspection of the wheel revealed no abnormal abrasive grain shedding or substrate damage. Compared to the PW06 grinding wheel tested in the same batch (which required replacement after processing 7,600 wafers), the SD6000-vbc14 grinding wheel's lifespan increased by over 40%, and the cost per wheel decreased by approximately 74%. Based on the cost per wafer grinding, the process of this invention can reduce the processing cost per wafer by approximately 22% compared to traditional methods, resulting in significant overall economic benefits.

[0062] During the adaptive parameter optimization phase, the closed-loop monitoring system performed trend analysis on the quality data during continuous processing. The analysis showed that around the 8000th wafer, a slight deviation in thickness tolerance occurred, approximately 0.2 μm. The system automatically triggered a parameter adjustment algorithm to calculate the correction amount for the second-stage feed rate, adjusting the original setpoint from 0.30 μm / s to 0.31 μm / s. Subsequent wafer thickness tolerances returned to the normal range. This adaptive adjustment process was fully automated, requiring no manual intervention, effectively ensuring the quality stability of the batch processing. Throughout the entire 10,000-wafer continuous processing verification cycle, the product yield consistently remained above 99.3%, meeting the stringent high-yield requirements for reclaimed silicon wafer polishing.

[0063] The above application examples demonstrate that the wafer grinding process optimization method based on the SD6000-vbc14 grinding wheel disclosed in this invention can effectively achieve high efficiency, high precision, and low cost synergistic optimization in the grinding process of 12-inch reclaimed silicon wafers. By selecting a high-precision grinding wheel with self-sharpening characteristics, configuring optimized equipment process parameters, implementing a multi-stage feed strategy, and establishing a closed-loop monitoring system, this invention significantly extends the grinding wheel's lifespan, reduces the cost per wafer, and ensures stable and reliable processing quality, meeting the urgent needs of the reclaimed silicon wafer industry for grinding processes.

[0064] Example 2

[0065] This embodiment provides a modified implementation of a wafer finishing process optimization method based on the SD6000-vbc14 grinding wheel, focusing on the parameter differentiation configuration scheme for different product specifications and customer requirements.

[0066] In terms of product specification classification, based on the specific application areas and customer requirements of 12-inch reclaimed silicon wafers, product specifications can be divided into three levels: high-end products, standard products, and economy products. High-end products are mainly aimed at the monitor wafer needs of advanced process wafer fabs, with the most stringent requirements for thickness tolerance, surface roughness, and total thickness variation. Standard products are aimed at the needs of mainstream process wafer fabs, with quality indicators meeting industry-standard requirements. Economy products are aimed at cost-sensitive applications, aiming to reduce processing costs as much as possible while ensuring basic quality requirements. For these three product levels, this invention designs differentiated process parameter configuration schemes. The parameter boundaries for each product level are distinguished as follows: high-end products require thickness tolerance ≤ ±0.8μm, total thickness variation ≤ 1.0μm, and surface roughness ≤ 0.1μm; standard products require thickness tolerance ≤ ±1.2μm, total thickness variation ≤ 1.5μm, and surface roughness ≤ 0.2μm; economy products require thickness tolerance ≤ ±1.5μm, total thickness variation ≤ 2.0μm, and surface roughness ≤ 0.3μm. The above boundary values ​​are derived from the statistical analysis results of customer technical specifications and batch verification data of this invention.

[0067] For high-end products, process parameters are configured with quality as the primary consideration. The three-stage feed rates are set to 0.30 μm / s for the first stage, 0.25 μm / s for the second stage, and 0.15 μm / s for the third stage, respectively, further reduced than the standard parameter set, to extend the processing time and finishing effect at each stage. The grinding speed is increased to 8 rev to ensure full release of surface stress. The corresponding processing time is increased, with a single-piece finishing grinding time of approximately 3.5 to 4 minutes, but the product quality reaches optimal levels. Under this parameter configuration, the thickness tolerance after finishing grinding can be controlled within ±0.8 μm, the average total thickness variation can be controlled within 0.6 μm, and the surface roughness reaches Ra 0.1 μm or less, meeting the stringent requirements of advanced process customers.

[0068] For economical products, process parameters are configured with efficiency as the priority. The three-stage feed rates are set to 0.40 μm / s for the first stage, 0.35 μm / s for the second stage, and 0.25 μm / s for the third stage, respectively, which is appropriately increased compared to the standard parameter set to shorten processing time. The finishing grinding revolutions are reduced to 3 rev, improving production efficiency while ensuring basic stress relief. Under this parameter configuration, the single-piece finishing grinding time is shortened to less than 2 minutes, significantly improving equipment utilization, while still meeting the technical requirements of economical products.

[0069] Regarding parameter switching in multi-batch continuous processing, this invention supports automatic parameter switching when continuously processing products of different specifications on the same precision grinding production line. When the system detects a change in product specifications, it automatically retrieves the corresponding process parameter set and sends it to the equipment control system for execution, eliminating the need for manual parameter resetting. This function significantly improves the product switching efficiency of the production line and reduces the risk of quality accidents caused by incorrect parameter settings.

[0070] Regarding equipment maintenance and grinding wheel replacement, this invention establishes a preventative maintenance mechanism based on the number of wafers processed. The system automatically records the processing time and equipment operating status parameters for each wafer. When the cumulative number of processed wafers approaches the expected lifespan of the grinding wheel, the system automatically generates a maintenance warning to prompt operators to prepare for grinding wheel replacement. When replacing the grinding wheel, the system automatically clears the process parameter group and calibration data corresponding to the old grinding wheel, loads the calibration program for the new grinding wheel, and completes the initialization and restoration of the equipment status. This mechanism effectively avoids processing quality deterioration and equipment failure caused by excessive grinding wheel use, ensuring the continuity and stability of the batch processing process.

[0071] The parameter differentiation configuration scheme in this embodiment expands the applicability of the process method of the present invention, enabling it to simultaneously meet the processing needs of high-end, standard and economical products, and improving the flexibility and market adaptability of the process system.

[0072] Example 3

[0073] This embodiment provides an extended implementation scheme for the wafer fine grinding process optimization method based on SD6000-vbc14 grinding wheel in a special application scenario, focusing on the coping strategies for extreme incoming material conditions and abnormal processing conditions.

[0074] Regarding process adjustments under extreme incoming material conditions, the incoming condition of reclaimed silicon wafers may exceed normal ranges in certain situations, such as abnormally increased surface damage depth, severely deteriorated thickness uniformity, or the presence of latent defects. To address these extreme conditions, this invention designs an emergency adjustment algorithm based on standard process parameter sets. The system automatically identifies the anomaly type and invokes corresponding parameter correction strategies by analyzing incoming material inspection data. For extreme cases with damage depths exceeding 4μm, the system automatically increases the first-stage removal amount and decreases the first-stage feed speed to ensure complete removal of all deep damage layers without excessively damaging the grinding wheel. For silicon wafers with severely deteriorated thickness uniformity, the system automatically adjusts the second-stage feed parameters, increasing the number of removal amount distribution adjustments to gradually converge the thickness distribution to the target range. For situations with the risk of latent defects, the system automatically increases the third-stage removal amount and polishing revolutions to enhance surface finishing effects and reduce the quality risks of subsequent processes caused by latent defects.

[0075] In terms of real-time monitoring and handling of abnormal processing conditions, this invention establishes a multi-dimensional anomaly detection mechanism covering key state parameters during the processing. Monitoring items include physical signals such as spindle vibration amplitude, spindle temperature, feed force fluctuations, and abnormal current, as well as quality signals such as real-time acquired thickness distribution and surface roughness. When any monitored parameter exceeds the preset normal range, the system immediately triggers an anomaly handling procedure, including pausing processing, protecting the wafer, automatically recording abnormal data, and generating alarm information. The specific execution steps of the anomaly handling procedure are as follows: First, the system immediately triggers an emergency stop command upon detecting an abnormal signal, cutting off the power output of the spindle drive and feed system; Second, the grinding wheel is withdrawn from the wafer surface at a controlled speed to avoid collision damage caused by rapid shutdown; Third, the system automatically saves all process data before and after the anomaly occurs for subsequent analysis and traceability; Fourth, the system generates an anomaly report and notifies the operator, while simultaneously recording the anomaly type, occurrence time, and preliminary diagnostic conclusions. This anomaly handling mechanism ensures that the equipment and wafers are properly protected even in the event of anomalies, providing data support for subsequent problem analysis and process improvement. The threshold settings for anomaly detection are as follows: spindle vibration amplitude threshold is set at 0.5 μm (derived from the safe operating range provided by the equipment manufacturer); spindle temperature threshold is set at 45℃ (beyond which the performance of the grinding wheel bond begins to degrade); and feed force fluctuation threshold is set at ±15% (derived from three standard deviations of the feed force fluctuation range during normal grinding). All of these thresholds were determined through statistical analysis of data from over 1000 wafers in normal processing.

[0076] In terms of post-processing anomaly analysis and process optimization, the system automatically initiates a post-analysis program after each anomaly event is handled. The analysis program retrieves processing data and equipment status data from multiple wafers consecutively before and after the anomaly occurred, using statistical process control methods to identify the root cause. If the anomaly is related to equipment status drift, the system automatically calculates the correction amount for equipment calibration parameters and updates it to the equipment control system; if the anomaly is related to improper process parameter settings, the system automatically adjusts the set values ​​of relevant parameters and updates the process parameter library; if the anomaly is related to incoming material status anomalies, the system automatically updates the incoming material inspection standards and sorting strategies. This continuous improvement mechanism enables the process system of this invention to continuously accumulate experience, optimize performance, and gradually improve its ability to respond to various anomalies.

[0077] The extended implementation of this embodiment enhances the robustness and reliability of the process method of the present invention in special application scenarios, ensuring that the process system can maintain normal production operation and product quality assurance capabilities in the face of various extreme conditions and abnormal situations.

[0078] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing wafer finishing processes based on SD6000-vbc14 grinding wheels, characterized in that, Includes the following steps: Step 1: Select SD6000-vbc14 type diamond grinding wheel as the grinding wheel for fine grinding. The abrasive grain size of the grinding wheel is 2μm, the wear mechanism is mainly micro-disintegration, it has self-sharpening characteristics, long service life, and stable grinding process; the matrix material of the grinding wheel is aluminum alloy or carbon fiber composite material, the binder type is resin-based binder, the grinding wheel concentration is 150%, the grinding wheel specifications are outer diameter φ400mm, inner diameter φ127mm, abrasive layer thickness 5mm, and total grinding wheel thickness 15mm; Step 2: Configure a precision grinding device adapted to the SD6000-vbc14 grinding wheel, and set the process parameters as follows: empty cutting amount of 15μm, spindle speed of 2330rpm, and table speed of 234rpm; the precision grinding device is a DFG-8560 type or a device with equivalent precision, with spindle radial runout ≤0.5μm, table axial runout ≤0.3μm, and empty cutting amount is achieved through closed-loop control by the built-in laser displacement sensor of the device, with a control accuracy of ±0.5μm; Step 3: Use a multi-stage feed strategy for fine grinding, where the first stage feed rate is set to 0.35 μm / s, the second stage feed rate is set to 0.30 μm / s, and the third stage feed rate is set to 0.2 μm / s. Step 4: After the fine grinding is completed, perform a polishing process with the polishing speed set to 5 rev to release surface stress. Step 5: After processing, remove the grinding wheel from the wafer surface at an escape velocity of 0.2 μm / s to prevent edge chipping; Step 6: Establish a closed-loop monitoring system to collect data in real time on the thickness, total thickness variation, and surface roughness of each wafer after processing. Based on the collected data, dynamically adjust the process parameters of subsequent wafers to achieve adaptive parameter control.

2. The wafer grinding process optimization method according to claim 1, characterized in that, In step 1, the SD6000-vbc14 type diamond grinding wheel has a continuous processing capacity of more than 10,000 pieces. Compared with traditional grinding wheels, its service life is significantly extended, the cost of a single grinding wheel is reduced by 74%, and the cost loss caused by replacing the grinding wheel when its service life expires is greatly reduced.

3. The wafer grinding process optimization method according to claim 1, characterized in that, The setting of the empty cutting amount in step 2 is used to ensure that the wafer smoothly enters the grinding area from the free state and avoids edge cracking caused by direct collision; the spindle speed and the table speed form a 2330:234 ratio relationship to form a stable relative motion force field, so that the grinding wheel abrasive grains contact the wafer surface layer by layer with micro-angle differences, avoiding surface texture defects caused by fixed trajectory grinding.

4. The wafer grinding process optimization method according to claim 1, characterized in that, In step 3, the multi-stage feed strategy achieves efficient material removal and finishing through progressively decreasing feed rates. The higher feed rate in the first stage is used to quickly remove macroscopic undulations on the wafer surface and residual damage layers from previous processes. The moderate feed rate in the second stage is used to precisely adjust the thickness distribution. The low feed rate in the third stage is used to complete the final finishing, ensuring uniform release of surface stress. The total removal amount in the three stages is controlled within the range of 4μm to 6μm, with the first stage accounting for 50% to 70%, the second stage accounting for 20% to 35%, and the third stage accounting for 5% to 15%. The sum of the removal amounts in each stage is 100%.

5. The wafer grinding process optimization method according to claim 1, characterized in that, In step 4, the polishing process establishes a stable dynamic equilibrium between the abrasive grains of the grinding wheel and the wafer surface through continuous grinding contact at a preset number of revolutions. This fully releases the surface stress accumulated during the grinding process, reduces the density of microcracks in the surface layer, and provides a good starting point for the subsequent chemical mechanical planarization process. The selection of the polishing revolution of 5rev is based on the following criteria: at a revolution of 3rev, the surface residual stress is about 35MPa; at a revolution of 5rev, the surface residual stress decreases to 12MPa; at a revolution of 8rev, the surface residual stress decreases to 10MPa, but the processing time is extended by 25%. Considering both the stress release effect and the processing efficiency, 5rev is the optimal balance point.

6. The wafer grinding process optimization method according to claim 1, characterized in that, In step 5, the escape velocity of 0.2 μm / s works synergistically with the grinding parameters to ensure that the grinding wheel smoothly detaches from the wafer surface at a controlled rate after completing the fine grinding process. This eliminates the adverse effects of the impact load generated by high-speed retraction on the integrity of the processed surface and prevents edge chipping due to stress concentration in the wafer edge area. The escape velocity and the third-stage feed rate are both 0.2 μm / s. The reason for this setting is that if the escape velocity is higher than the third-stage feed rate (e.g., 0.5 μm / s), the edge chipping rate will increase to 1.2%. If the escape velocity is lower than the third-stage feed rate (e.g., 0.1 μm / s), the single-wafer processing time will increase by 8% without significant quality improvement. Therefore, having the same value for both values ​​can balance edge protection and production efficiency.

7. The wafer grinding process optimization method according to claim 1, characterized in that, It also includes process stability control for batch processing. After each wafer is finished in the fine grinding process, the closed-loop monitoring system automatically collects thickness tolerance data, total thickness variation value, and surface roughness data. The collected data is compared and analyzed with preset standards. Based on the comparison results, the process parameters of the next wafer are dynamically adjusted to achieve adaptive parameter adjustment, ensuring that the yield is maintained above 99.3% when processing more than 10,000 wafers continuously. The adaptive parameter adjustment adopts an incremental PID control strategy, where the proportional gain P is initially set to 0.8, the integral coefficient I is initially set to 0.1, and the derivative time D is initially set to 0.

15. The control cycle is adjusted once every 5 wafers processed. The parameter adjustment trigger conditions are: when the thickness tolerance center value of 3 consecutive wafers deviates by more than ±0.5μm; or when the total thickness variation value of 3 consecutive wafers exceeds 1.5μm; or when the surface roughness Ra of 3 consecutive wafers exceeds 0.15μm. The mapping relationship is: spindle speed correction amount = thickness deviation × P, feed rate correction amount = total thickness variation deviation * I.

8. The wafer grinding process optimization method according to claim 1, characterized in that, It supports setting processing parameters independently for each wafer to adapt to the differences in the characteristics of recycled silicon wafers and the special processing needs of customers, thereby improving the flexibility of the process and market adaptability, and meeting the requirements of the wafer manufacturing industry for high precision, high efficiency and low cost of the grinding process.

9. The wafer grinding process optimization method according to claim 1, characterized in that, The process parameters are configured differently according to different product specifications. This includes setting different three-stage feed rates, grinding revolutions, and corresponding processing times for high-end, standard, and economical products. It also supports automatic parameter switching when continuously processing different specifications of products on the same precision grinding production line. Specifically, the three-stage feed rates for high-end products are 0.30μm / s, 0.25μm / s, and 0.15μm / s, respectively; for standard products, they are 0.35μm / s, 0.30μm / s, and 0.20μm / s, respectively; and for economical products, they are 0.40μm / s, 0.35μm / s, and 0.25μm / s, respectively. The parameter boundaries between product grades are distinguished by thickness tolerance ≤ ±1.2μm, total thickness variation ≤ 1.5μm, and surface roughness ≤ 0.2μm.

10. The wafer grinding process optimization method according to claim 1, characterized in that, In claim 1, the "preset particle size" is 2 μm; the "preset number of flakes" is 10,000 flakes; the "preset ratio" is 2330:234; the "preset range" is 4 μm to 6 μm; the "preset rotation speed" is 5 rev; the "preset escape velocity" is 0.2 μm / s; and the "preset yield threshold" is 99.3%.

Images