A method and system for controlling the temperature of a polishing pad during a double-sided polishing process

Abstract

The application discloses a polishing disc temperature control method and system in a double-sided polishing process, and relates to the technical field of polishing disc temperature control.The method comprises the following steps: performing center-edge radial temperature difference analysis to generate a polishing disc thermal expansion offset result; performing contact pressure drift analysis on a glass wafer polishing contact area to obtain an edge contact pressure offset result; establishing an edge removal imbalance state matrix; establishing a dynamic coupling constraint model of edge collapse-TTV; performing radial temperature difference compensation iteration decision to obtain a radial partition temperature compensation strategy; and controlling a liquid temperature controller to perform differential temperature adjustment on upper and lower polishing discs according to the radial partition temperature compensation strategy.The application solves the technical problem that the uneven polishing disc surface temperature difference in the prior art leads to wafer edge collapse and thickness deviation exceeding the standard, and achieves the technical effects of realizing accurate polishing disc radial temperature difference regulation and control, and improving glass wafer double-sided polishing forming quality and thickness uniformity.

Description

Technical Field

[0001] This invention relates to the field of polishing disc temperature control technology, specifically to a method and system for controlling the temperature of a polishing disc during double-sided polishing. Background Technology

[0002] In the double-sided polishing process of glass wafers, radial temperature differences are easily generated when the upper and lower polishing pads are working. These temperature differences cause thermal expansion and displacement of the pad surfaces, resulting in abnormal shifts in wafer contact pressure. This leads to an imbalance in the material removal rate at the wafer edges, easily causing edge collapse defects and large overall thickness deviations. Traditional temperature control methods cannot combine deformation, pressure, and material removal status to regulate temperature, making it difficult to effectively constrain defect evolution. Ultimately, this results in insufficient stability of the finished wafer polishing quality, making it difficult to meet the requirements of high-precision processing and production.

[0003] Existing technologies suffer from uneven temperature differences on the polishing surface, leading to wafer edge collapse and excessive thickness deviation. Summary of the Invention

[0004] This application provides a method and system for controlling the temperature of the polishing pad during double-sided polishing, which is used to address the technical problem in the prior art where uneven temperature difference on the polishing pad surface leads to wafer edge collapse and excessive thickness deviation.

[0005] In view of the above problems, this application provides a method and system for controlling the temperature of the polishing disc during double-sided polishing.

[0006] A first aspect of this application provides a method for controlling the temperature of a polishing disc during double-sided polishing, the method comprising: Radial temperature difference analysis of the center edge is performed based on the temperature data of the upper and lower polishing pads to generate polishing pad thermal expansion offset results. Contact pressure drift analysis is then performed on the glass wafer polishing contact area based on the polishing pad thermal expansion offset results to obtain edge contact pressure offset results. Edge removal imbalance detection is performed using the edge contact pressure offset results and glass wafer surface material removal rate data to establish an edge removal imbalance state matrix. The glass wafer edge collapse evolution path is predicted based on the edge removal imbalance state matrix, and TTV offset correlation and edge collapse critical constraints are established based on the glass wafer edge collapse evolution path to establish an edge collapse-TTV dynamic coupling constraint model. Radial temperature difference compensation iterative decision-making is performed using the polishing pad thermal expansion offset results and the edge collapse-TTV dynamic coupling constraint model to obtain a radial partition temperature compensation strategy. Finally, a liquid temperature controller is used to perform differentiated temperature regulation on the upper and lower polishing pads based on the radial partition temperature compensation strategy.

[0007] A second aspect of this application provides a polishing disc temperature control system for double-sided polishing processes, the system comprising: The system includes the following modules: a temperature difference analysis module for performing radial temperature difference analysis at the center edge based on the temperature data of the upper and lower polishing pads, generating a polishing pad thermal expansion offset result; a contact pressure drift analysis module for performing contact pressure drift analysis on the glass wafer polishing contact area based on the polishing pad thermal expansion offset result, obtaining an edge contact pressure drift result; an edge removal imbalance detection module for detecting edge removal imbalance using the edge contact pressure drift result and glass wafer surface material removal rate data, establishing an edge removal imbalance state matrix; a constraint model establishment module for predicting the glass wafer edge collapse evolution path based on the edge removal imbalance state matrix, and establishing an edge collapse-TTV dynamic coupling constraint model based on the glass wafer edge collapse evolution path through TTV offset correlation and edge collapse critical constraints; a compensation strategy acquisition module for performing radial temperature difference compensation iterative decision-making based on the polishing pad thermal expansion offset result and the edge collapse-TTV dynamic coupling constraint model, obtaining a radial partition temperature compensation strategy; and a differentiated temperature adjustment module for controlling the liquid temperature controller to perform differentiated temperature adjustment on the upper and lower polishing pads based on the radial partition temperature compensation strategy.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: By analyzing the radial temperature difference between the center and edge of the upper and lower polishing pads, the thermal expansion offset of the polishing pads is generated. Contact pressure drift analysis is performed on the glass wafer polishing contact area to obtain the edge contact pressure offset. Edge removal imbalance detection is performed, and an edge removal imbalance state matrix is ​​established. Based on the edge removal imbalance state matrix, the glass wafer edge collapse evolution path is predicted, and TTV offset correlation and edge collapse critical constraints are established based on the glass wafer edge collapse evolution path to create a collapse-TTV dynamic coupling constraint model. Radial temperature difference compensation iterative decision-making is performed using the polishing pad thermal expansion offset results and the collapse-TTV dynamic coupling constraint model to obtain a radial partition temperature compensation strategy. Based on the radial partition temperature compensation strategy, a liquid temperature controller is used to perform differentiated temperature adjustment on the upper and lower polishing pads. This achieves precise control of the radial temperature difference of the polishing pads, improving the forming quality and thickness uniformity of the double-sided polishing of the glass wafer. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1A schematic flowchart of a method for controlling the temperature of a polishing disc during double-sided polishing is provided in an embodiment of this application. Figure 2 This is a schematic diagram of a temperature control system for a polishing disc during double-sided polishing, provided as an embodiment of this application.

[0011] Figure labeling: Temperature difference analysis module 10, contact pressure drift analysis module 20, edge removal imbalance detection module 30, constraint model establishment module 40, compensation strategy acquisition module 50, differential temperature adjustment module 60. Detailed Implementation

[0012] This application provides a method and system for controlling the temperature of the polishing pad during double-sided polishing, which addresses the technical problem in the prior art where uneven temperature difference on the polishing pad surface leads to wafer edge collapse and excessive thickness deviation.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] Example 1, as Figure 1 As shown, this application provides a method for controlling the temperature of the polishing disc during double-sided polishing, the method comprising: Step S100: Perform radial temperature difference analysis of the center edge based on the temperature data of the upper polishing pad and the temperature data of the lower polishing pad to generate the thermal expansion offset result of the polishing pad.

[0015] Specifically, the temperature difference between the center and edge of the upper and lower polishing pads is identified, and the radial temperature difference curves of the upper and lower polishing pads are constructed in sequence. Then, the thermal expansion offset fitting calculation is carried out based on the radial temperature difference curves of the upper and lower polishing pads, and finally the thermal expansion offset result of the polishing pad is generated.

[0016] Step S200: Based on the thermal expansion offset results of the polishing pad, perform contact pressure drift analysis on the glass wafer polishing contact area to obtain the edge contact pressure offset results.

[0017] Specifically, the height offset feature of the polishing pad is identified based on the thermal expansion offset result of the polishing pad. Based on the height offset feature, the pressure distribution of the glass wafer polishing contact area is mapped to obtain the contact pressure distribution result. Then, the edge pressure migration is identified based on the contact pressure distribution result to determine the edge contact pressure drift direction. Finally, the pressure offset is calculated by combining the contact pressure distribution result and the edge contact pressure drift direction, thereby obtaining the edge contact pressure offset result.

[0018] Step S300: Detect edge removal imbalance by using the edge contact pressure offset results and glass wafer surface material removal rate data, and establish an edge removal imbalance state matrix.

[0019] Specifically, the method involves adjusting the positioning based on the edge contact pressure offset results to obtain the wafer edge pressure concentration zone and wafer edge pressure attenuation zone. Combined with the glass wafer surface material removal rate data, the method identifies the gradient of the instantaneous removal rate change at the edge and the direction of the radial removal rate migration at the edge. Based on the pressure concentration zone and attenuation zone, the method divides the gradient of the instantaneous removal rate change at the edge into pressure response zones, identifies abnormal expansion areas of the edge removal rate, and extrapolates the edge erosion trend of the abnormal areas according to the direction of the radial removal rate migration at the edge. Finally, the method establishes an edge removal imbalance state matrix.

[0020] Step S400: Predict the glass wafer edge collapse evolution path based on the edge removal imbalance state matrix, and perform TTV offset correlation and edge collapse critical constraint based on the glass wafer edge collapse evolution path to establish an edge collapse-TTV dynamic coupling constraint model.

[0021] Specifically, TTV, or total wafer thickness variation, refers to the thickness difference between the highest and lowest points on the wafer surface. As a core indicator for measuring wafer flatness, its value directly affects the lithography accuracy and final device yield in semiconductor manufacturing. First, the evolution path of glass wafer edge collapse is predicted based on the edge removal imbalance state matrix, identifying the edge collapse expansion direction and the shrinkage trend of the effective contact area. Then, based on the shrinkage trend of the effective contact area, an edge material removal redistribution analysis is conducted to obtain the secondary offset result of the edge removal rate. Based on this result, the wafer radial thickness difference expansion process is further deduced to obtain the TTV dynamic offset trend. Subsequently, the TTV dynamic offset trend is used to identify the edge collapse instability criticality, and the critical constraint interval of edge collapse is determined. Finally, the edge collapse expansion direction, the secondary offset result of the edge removal rate, the TTV dynamic offset trend, and the critical constraint interval of edge collapse are spatiotemporally matched and mapped to establish a collapse-TTV dynamic coupling constraint model.

[0022] Step S500: Perform radial temperature difference compensation iterative decision-making based on the thermal expansion offset result of the polishing pad and the collapsed edge-TTV dynamic coupling constraint model to obtain the radial partition temperature compensation strategy.

[0023] Specifically, based on the thermal expansion offset results of the polishing disk, the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone are divided. The temperature compensation collaborative decision-making of the above three regions is carried out using the collapse-TTV dynamic coupling constraint model to obtain the temperature compensation decision sets of the center, edge, and transition zones. The three decision sets are randomly combined to generate the temperature compensation collaborative space. Then, TTV prediction is performed on each scheme in the collaborative space, and the temperature compensation candidate space is screened according to the preset TTV. Finally, the energy consumption minimization screening is performed on the candidate space to output the final radial partition temperature compensation strategy.

[0024] Step S600: Control the liquid temperature controller to perform differentiated temperature adjustment on the upper and lower polishing discs according to the radial partition temperature compensation strategy.

[0025] Specifically, based on the radial partition temperature compensation strategy, the liquid temperature controller performs radial partition differential temperature adjustment on the upper and lower polishing disks respectively. At the same time, it performs anomaly prediction based on the real-time status data of the liquid temperature controller to obtain abnormal characteristics of the controller. Based on the abnormal characteristics of the controller, the radial partition temperature compensation strategy is corrected, and a controller warning signal is output to ensure stable and reliable temperature control of the polishing disk.

[0026] In one possible implementation, step S100 further includes: Step S110: Identify the temperature difference between the center and edge of the upper polishing disk temperature data and construct the radial temperature difference curve of the upper polishing disk.

[0027] Step S120: Identify the temperature difference between the center and edge of the lower polishing pad temperature data and construct the radial temperature difference curve of the lower polishing pad.

[0028] Step S130: Perform thermal expansion offset fitting based on the radial temperature difference curve of the upper polishing disk and the radial temperature difference curve of the lower polishing disk to obtain the thermal expansion offset result of the polishing disk.

[0029] Specifically, along the radial direction of the upper polishing disk, the disk surface is divided into a central region, an intermediate transition region, and an edge region. Real-time temperature data of each region is collected. By calculating the temperature difference between the central reference point and each radial position point and extracting the radial temperature gradient, the temperature difference between the center and the edge is identified. Then, based on the temperature and difference data of each radial position, continuous fitting is performed to finally construct a radial temperature difference curve of the upper polishing disk that fully reflects the radial temperature distribution characteristics of the upper polishing disk.

[0030] Along the radial direction of the lower polishing pad, the pad surface is divided into a central region, an intermediate transition region, and an edge region. Real-time temperature data of each region is collected. By calculating the difference between the central reference temperature and the temperature at each radial position point by point and extracting the radial temperature gradient, the temperature difference between the center and the edge is identified. Then, based on the temperature and difference data at each radial position, continuous fitting is performed to finally construct a radial temperature difference curve of the lower polishing pad that fully reflects the radial temperature distribution characteristics of the lower polishing pad.

[0031] Based on the radial temperature difference curves of the upper and lower polishing pads, and combined with the thermal expansion coefficient of the polishing pad material, the thermal expansion deformation caused by uneven temperature at each radial measuring point of the upper and lower polishing pads is first calculated. Then, the temperature difference at each radial point and the corresponding thermal expansion deformation are fitted by a polynomial to obtain the thermal expansion offset distribution curve of the upper and lower polishing pads along the radial direction. Finally, the complete thermal expansion offset result of the polishing pads is formed by superimposing and aligning the offset of the upper and lower pad surfaces.

[0032] In one possible implementation, step S200 further includes: Step S210: Identify the height offset characteristics of the polishing pad based on the thermal expansion offset results.

[0033] Step S220: Map the pressure distribution of the glass wafer polishing contact area according to the disk height offset characteristics to obtain the contact pressure distribution results.

[0034] Step S230: Based on the contact pressure distribution results, perform edge pressure migration identification to determine the edge contact pressure drift direction.

[0035] Step S240: Calculate the pressure offset based on the contact pressure distribution result and the edge contact pressure drift direction to obtain the edge contact pressure offset result.

[0036] Specifically, based on the thermal expansion offset results of the polishing pad, the thermal expansion deformation values ​​of the upper and lower polishing pads are extracted point by point along the radial direction. By comparing the deformation differences between the central area and the edge area, the magnitude, gradient and radial distribution of the pad height offset are determined. At the same time, typical offset patterns such as central bulge / depression and edge warping are identified, thereby completely identifying the pad height offset characteristics.

[0037] The height offset characteristics of the disk surface are transformed into the contact gap distribution of the glass wafer polishing contact area. Combined with the loading load set by the polishing process and the contact stiffness between the polishing disk and the wafer, a height offset-contact pressure mapping relationship is established. The actual contact pressure between the wafer and the polishing disk is calculated point by point along the radial direction. Through pressure fitting and distribution reconstruction in the entire radial region, a complete contact pressure distribution result that characterizes the pressure change in the polishing contact area is obtained.

[0038] The contact pressure distribution results are divided radially into a central region, a transition region, and an edge region. The average contact pressure and pressure gradient of each region are calculated. By comparing the difference between the actual pressure in the edge region and the theoretical uniform pressure, it is determined whether the edge pressure shows a trend of gathering towards the center or expanding outward. Then, the drift direction of the edge contact pressure is determined by combining the radial pressure change rate, thus completing the edge pressure migration identification.

[0039] Based on the contact pressure distribution results, the actual contact pressure of multiple radial measuring points in the edge region of the glass wafer is first extracted. Combined with the determined edge contact pressure drift direction, the difference between the pressure of each measuring point on the edge and the ideal uniform pressure setting value is calculated to obtain the single-point pressure offset amplitude. Then, the offset amplitude of multiple points in the full circumference of the edge is weighted and gradient corrected. Finally, the overall offset magnitude and distribution pattern of the edge contact pressure are quantitatively calculated to obtain the complete edge contact pressure offset result.

[0040] In one possible implementation, step S300 further includes: Step S310: Based on the edge contact pressure offset results, perform position adjustments to obtain the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone.

[0041] Step S320: Based on the material removal rate data of the glass wafer surface, identify the gradient of the instantaneous removal rate change at the edge and the direction of radial removal rate migration at the edge.

[0042] Step S330: Based on the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone, perform pressure response partitioning on the instantaneous edge removal rate change gradient to identify abnormal expansion areas of edge removal rate.

[0043] Step S340: Based on the migration direction of the radial removal rate of the edge, perform edge erosion trend deduction on the abnormal expansion area of ​​the edge removal rate to obtain the edge removal imbalance state matrix.

[0044] Specifically, by comparing the actual contact pressure at each point on the edge with the standard pressure reference value, the specific radial and circumferential locations where the pressure values ​​increase or decrease are determined one by one. Boundaries are drawn based on the range of pressure increase or decrease, thereby distinguishing between the wafer edge pressure concentration zone with excessive pressure and the wafer edge pressure attenuation zone with insufficient pressure.

[0045] Extract the removal rate data of each radial point on the edge of the wafer, and calculate the change gradient by dividing the difference between the values ​​of two adjacent points by the radial distance. If the value continues to increase from the outside to the inside, it is determined that the removal rate is migrating towards the center of the wafer. If the value continues to decrease from the outside to the inside, it is determined that the removal rate is migrating towards the outside of the wafer. This completes the direction determination.

[0046] Using the edge pressure concentration zone and edge pressure attenuation zone as the basis for division, the instantaneous removal rate change gradient of the corresponding interval is matched and classified one by one to complete the pressure response partitioning. Then, the gradient value in the partition is compared with the reasonable threshold preset by the process. The area that exceeds the threshold range is the abnormal expansion area of ​​edge removal rate.

[0047] Based on the actual migration trend of the radial material removal rate at the edge, the disk area where the removal rate abnormally expands is identified. Combining the difference in removal rate at different points and the law of position change, the development trend and diffusion range of material edge erosion are deduced. Various removal deviation parameters and deformation correlation parameters within the region are statistically analyzed. The parameters representing the imbalance state are arranged in a dimensional order and integrated to construct the edge removal imbalance state matrix.

[0048] In one possible implementation, step S400 further includes: Step S410: Based on the glass wafer edge collapse evolution path, identify the edge collapse expansion direction and the shrinkage trend of the effective contact area at the edge.

[0049] Step S420: Perform edge material removal and redistribution analysis based on the shrinkage trend of the effective contact area of ​​the edge to obtain the secondary offset result of the edge removal rate.

[0050] Step S430: Based on the secondary offset result of the edge removal rate, perform wafer radial thickness difference expansion deduction to obtain the TTV dynamic offset trend.

[0051] Step S440: Based on the TTV dynamic offset trend, identify the critical collapse instability of the edge and determine the critical constraint range of the edge collapse.

[0052] Step S450: Based on the edge collapse expansion direction, the edge removal rate secondary offset result, the TTV dynamic offset trend, and the edge collapse critical constraint interval, perform spatiotemporal matching mapping to establish the collapse-TTV dynamic coupling constraint model.

[0053] Specifically, by tracking the deformation data at each point along the wafer collapse path, comparing the changes in edge contour morphology at different times, determining the radial inward or outward extension direction of the collapse, and simultaneously statistically analyzing the real-time change in the coverage area of ​​the contact surface, combined with the regional area reduction rate, the shrinkage trend of the effective contact area at the edge can be accurately identified.

[0054] Based on the shrinkage boundary and reduction range of the effective contact area at the edge, the actual force contact range is redefined, the total amount of material to be removed is redistributed according to the contact area ratio, the updated removal rate at each location is calculated, the new rate is compared with the standard process removal rate one by one to calculate the difference, the deviation values ​​at each location are summarized, and finally the secondary offset result of the edge removal rate is obtained.

[0055] Based on the secondary offset of the edge removal rate, the total amount of material removed from each radial point is accumulated layer by layer according to the processing time. The thickness difference between adjacent points is calculated in real time. The thickness deviation expansion change is iteratively deduced by combining the spatial diffusion law. The dynamic change trajectory of the deviation value is continuously recorded. Finally, the dynamic offset trend of the total thickness deviation of the wafer is obtained by fitting.

[0056] Based on the acquired TTV dynamic offset trend time series data, three core feature parameters—TTV offset, offset growth rate, and radial deviation diffusion slope—are extracted in real time. The real-time parameters are compared frame by frame with the system's preset wafer edge collapse instability threshold and critical slope threshold. When a sudden increase in TTV offset growth rate or a sudden inflection point in the deviation diffusion slope is detected, the current state is determined to have entered the critical state of edge collapse instability. At the same time, the upper and lower limit process parameter ranges corresponding to this inflection point are recorded. By fitting the parameter fluctuation range under the critical state, the critical constraint range of wafer edge collapse is accurately locked and determined.

[0057] Employing a multi-feature spatiotemporal fusion algorithm, a two-dimensional temporal-spatial sample dataset is first constructed. Features such as edge collapse propagation direction, edge removal rate secondary offset, and TTV dynamic offset trend corresponding to different processing sequences and wafer radial positions are used as the model input feature set. The critical constraint interval of edge collapse is used as the model boundary constraint label and the ground truth for instability judgment supervision. Spatiotemporal interpolation registration completes pixel-level and temporal-level accurate matching and mapping of all features. The model adopts a four-layer fully connected temporally coupled network structure, including an input layer, two feature fusion hidden layers, and a coupled constraint output layer. The input layer is... The model unifies the processing of four types of multidimensional feature data. The hidden layer learns the dynamic correlation and coupling relationship between edge collapse deformation and TTV offset through the weight matrix. During the training phase, the network weights and bias parameters are iteratively optimized using the mini-batch gradient descent algorithm. The critical constraint interval is used as the boundary penalty term of the loss function to suppress abnormal fitting bias beyond the threshold. After repeated iterations and convergence, the optimal model parameters are solidified to complete the training and construction of the edge collapse-TTV dynamic coupling constraint model. After the model is trained, process dynamic features can be input in real time to realize the linkage prediction, constraint correction and instability prediction of wafer edge collapse state and TTV offset trend.

[0058] In one possible implementation, step S500 further includes: Step S510: Based on the thermal expansion offset results of the polishing pad, determine the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone.

[0059] Step S520: Based on the collapsed edge-TTV dynamic coupling constraint model, perform temperature compensation collaborative decision-making for the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone to obtain the temperature compensation collaborative space.

[0060] Step S530: Perform TTV prediction based on each temperature compensation coordination scheme in the temperature compensation coordination space to obtain a TTV prediction set.

[0061] Step S540: Based on the TTV prediction set, filter the temperature compensation coordination space according to the predetermined TTV to obtain the temperature compensation candidate space.

[0062] Step S550: Perform energy consumption minimization screening on the temperature compensation candidate space and output the radial partition temperature compensation strategy.

[0063] Specifically, the thermal expansion offset values ​​of each point in the radial direction of the polishing pad are collected, and the boundary range is determined according to the magnitude of the offset and the gradient change threshold. The central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone between the two are divided in sequence.

[0064] Based on the established coupled constraint model of collapse edge thickness deviation, a linkage analysis was carried out for three thermal expansion regions. A suitable temperature adjustment scheme was formulated in a coordinated manner by combining the constraints of deformation and thickness change, and all feasible compensation parameter ranges were summarized to obtain the temperature compensation coordination space.

[0065] A lightweight temporal prediction neural network algorithm driven by partition features is used to complete TTV prediction and dataset construction. First, the combination of temperature compensation parameters for all partitions within the temperature compensation co-space is used as the model input features. Wafer radial partition collapse deformation features and thermal expansion offset features are used as constraint auxiliary features to construct a multi-dimensional input feature matrix. The prediction network adopts a four-layer dedicated structure: input layer, feature enhancement layer, temporal inference layer, and output layer. The input layer normalizes and reduces noise for various temperature compensation parameters. The feature enhancement layer performs cross-fusion of partition compensation features and collapse-TTV coupling features. The temporal inference layer relies on historical data... The weight parameters obtained from the processing sample training are used to deduce the evolution law of thickness deviation under the compensation effect. The output layer outputs the global TTV prediction value corresponding to a single compensation scheme. The model adopts the batch gradient descent algorithm in advance and completes offline training and convergence optimization based on massive process samples. The feature association weights and deviation correction coefficients are solidified. After training, the model is directly activated to traverse each temperature compensation collaborative scheme in the temperature compensation collaborative space. Each scheme is substituted into the model to complete the forward inference prediction and obtain the wafer TTV prediction results corresponding to each scheme. All prediction results are structured, summarized and archived to finally construct a complete TTV prediction set.

[0066] The preset thickness deviation acceptable threshold is retrieved, and each value in the prediction set is compared one by one. The compensation schemes that meet the standard are selected, and the combination of parameters that exceeds the standard is eliminated, thereby reducing the candidate space for temperature compensation.

[0067] Calculate the energy consumption of each compensation scheme in the candidate space, compare and select the parameter combination with the lowest energy consumption, and determine and output the temperature compensation execution strategy that is suitable for the radial partition of the disk.

[0068] In one possible implementation, step S520 further includes: Step S521: Make temperature compensation decisions for the central thermal expansion zone based on the collapsed edge-TTV dynamic coupling constraint model, and obtain the central temperature compensation decision set.

[0069] Step S522: Make temperature compensation decisions for the edge thermal expansion zone according to the collapsed edge-TTV dynamic coupling constraint model, and obtain the edge temperature compensation decision set.

[0070] Step S523: Based on the collapsed edge-TTV dynamic coupling constraint model, make temperature compensation decisions for the thermal expansion transition zone to obtain the transition temperature compensation decision set.

[0071] Step S524: Randomly combine the center temperature compensation decision set, the edge temperature compensation decision set, and the transition temperature compensation decision set to generate the temperature compensation collaborative space.

[0072] Specifically, the collapse-TTV dynamic coupling constraint model is retrieved, and the deformation, thickness deviation and thermal offset parameters of the central thermal expansion zone are substituted. The adaptive temperature adjustment parameters are solved by combining the constraint correlation, and all compliant compensation schemes are summarized to obtain the central temperature compensation decision set.

[0073] The pre-trained collapse-TTV dynamic coupling constraint model is invoked to collect the thermal expansion offset of the polishing pad corresponding to the edge thermal expansion zone, the wafer edge collapse expansion direction, the secondary offset value of the edge removal rate, and the TTV dynamic offset trend characteristic parameters. The multi-dimensional regional working condition parameters are completely input into the internal constraint equation of the model. The boundary conditions of the critical constraint interval of edge collapse are combined to perform a global traversal solution. All effective temperature compensation amplitude and adjustment parameter combinations that can suppress edge collapse defects and correct thickness deviations are selected. All compliant and feasible temperature compensation schemes are uniformly collected and organized to finally generate an edge temperature compensation decision set.

[0074] The pre-trained collapse-TTV dynamic coupling constraint model is invoked to collect the polishing pad thermal expansion offset parameters, wafer edge collapse expansion direction parameters, edge material removal rate secondary offset parameters, and total thickness deviation dynamic offset trend parameters corresponding to the thermal expansion transition zone. Simultaneously, combining the deformation transition characteristics specific to the thermal expansion transition zone and the critical constraint interval for edge collapse instability, all regional operating condition characteristic parameters are input into the collapse-total thickness deviation dynamic coupling constraint model for global iterative solution. Based on the dynamic coupling correlation law between collapse deformation and total thickness deviation within the model, all temperature compensation adjustment parameter combinations that can adapt to the processing characteristics of the thermal expansion transition zone and simultaneously meet the requirements of thickness uniformity control and collapse defect suppression are solved. All compliant and effective temperature compensation schemes are collected, organized, and structured for storage, ultimately generating the transition temperature compensation decision set corresponding to the thermal expansion transition zone.

[0075] The temperature compensation decision schemes corresponding to the central region, the edge region, and the transition region are combined and matched together. The compensation parameters of different zones are randomly matched and paired, and all the combinations are summarized to form a complete control scheme, thereby forming a temperature compensation coordination space.

[0076] In one possible implementation, step S600 further includes: Step S610: Based on the real-time status data of the liquid temperature controller, perform anomaly prediction and obtain the abnormal characteristics of the controller.

[0077] Step S620: Correct the radial partition temperature compensation strategy according to the abnormal characteristics of the controller.

[0078] Specifically, a stacked time-series feature isolated forest anomaly prediction machine learning algorithm is used to predict operational anomalies and extract anomaly characteristics of liquid temperature controllers. The algorithm is divided into an offline training phase and an online prediction phase. The main body of the model constructs a multi-layered nested structure for time-series feature extraction and anomaly discrimination. During offline training, firstly, full-dimensional time-series state data such as temperature control output value, response delay, load fluctuation, pipeline fluid flow stability, and equipment temperature rise under historical normal operating conditions of the liquid temperature controller are collected to construct a standardized normal sample training dataset. The original state data is then subjected to time-series interpolation, noise reduction normalization, and sliding window feature reconstruction to mine the inherent characteristic distribution patterns of the equipment's steady-state operation. This is achieved through unsupervised learning. The model is iteratively trained using a learning method, which refines the feature segmentation threshold and isolation discrimination weights, continuously optimizing the isolation degree calculation criteria for the sample space until the model converges and the optimal feature discrimination parameters are solidified. In the online prediction phase, real-time collected state data of the liquid temperature controller are input into the trained model. The model quantifies the deviation and isolation degree between the real-time samples and the normal steady-state feature space, accurately identifying abnormal operating conditions that deviate from the normal operating distribution. Simultaneously, based on the abnormal deviation dimension, deviation amplitude, and duration characteristics, various abnormal dimension information such as temperature drift anomaly, response lag anomaly, and control instability anomaly are extracted, and the anomaly intensity and evolution trend are quantified. Finally, the complete abnormal characteristics of the controller are integrated.

[0079] Based on the identified abnormal characteristics of the controller, the values ​​and control amplitudes of the temperature compensation parameters for each zone are adjusted accordingly, and the established radial zone temperature compensation strategy is adapted and modified to ensure that the temperature control effect is stable and meets the standards.

[0080] In one possible implementation, step S610 further includes: Step S611: Output a controller warning signal based on the abnormal characteristics of the controller.

[0081] Specifically, based on the identified abnormal characteristics of the controller, different abnormal categories such as temperature deviation, response stagnation, and unstable pipeline flow are distinguished. At the same time, the severity level of the fault is determined, and audible and visual prompts, text alarm messages, and status alarm codes are output according to the established standards to intuitively reflect the current abnormal situation of the equipment.

[0082] Example 2, based on the same inventive concept as the method for controlling the temperature of the polishing disc during double-sided polishing in the foregoing examples, such as... Figure 2 As shown, this application provides a polishing disc temperature control system during double-sided polishing. The system and method embodiments in this application are based on the same inventive concept. The system includes: The temperature difference analysis module 10 is used to perform radial temperature difference analysis of the center edge based on the temperature data of the upper polishing pad and the temperature data of the lower polishing pad, and generate the thermal expansion offset result of the polishing pad.

[0083] The contact pressure drift analysis module 20 is used to perform contact pressure drift analysis on the glass wafer polishing contact area based on the thermal expansion offset result of the polishing pad, and obtain the edge contact pressure offset result.

[0084] The edge removal imbalance detection module 30 is used to detect edge removal imbalance by means of the edge contact pressure offset result and the glass wafer surface material removal rate data, and to establish an edge removal imbalance state matrix.

[0085] The constraint model establishment module 40 is used to predict the glass wafer edge collapse evolution path based on the edge removal imbalance state matrix, and to perform TTV offset correlation and edge collapse critical constraint based on the glass wafer edge collapse evolution path, thereby establishing an edge collapse-TTV dynamic coupling constraint model.

[0086] The compensation strategy acquisition module 50 is used to perform radial temperature difference compensation iterative decision-making based on the thermal expansion offset result of the polishing disk and the collapsed edge-TTV dynamic coupling constraint model, and to obtain the radial partition temperature compensation strategy.

[0087] The differential temperature control module 60 is used to control the liquid temperature controller to perform differential temperature control on the upper and lower polishing discs according to the radial partition temperature compensation strategy.

[0088] Furthermore, the system is also used to implement the following functions: The temperature data of the upper polishing pad is used to identify the temperature difference between the center and the edge, and a radial temperature difference curve of the upper polishing pad is constructed. The temperature data of the lower polishing pad is also used to identify the temperature difference between the center and the edge, and a radial temperature difference curve of the lower polishing pad is constructed. Thermal expansion offset fitting is performed based on the radial temperature difference curves of the upper and lower polishing pads to obtain the thermal expansion offset result of the polishing pad.

[0089] Furthermore, the system is also used to implement the following functions: Based on the thermal expansion offset of the polishing pad, the pad height offset feature is identified; based on the pad height offset feature, the pressure distribution of the glass wafer polishing contact area is mapped to obtain the contact pressure distribution result; based on the contact pressure distribution result, edge pressure migration is identified to determine the edge contact pressure drift direction; based on the contact pressure distribution result and the edge contact pressure drift direction, the pressure offset is calculated to obtain the edge contact pressure offset result.

[0090] Furthermore, the system is also used to implement the following functions: Based on the edge contact pressure offset results, the positioning is adjusted to obtain the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone; based on the glass wafer surface material removal rate data, the gradient of instantaneous edge removal rate change and the direction of edge radial removal rate migration are identified; based on the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone, the gradient of instantaneous edge removal rate change is divided into pressure response zones to identify abnormal expansion areas of edge removal rate; based on the direction of edge radial removal rate migration, the edge erosion trend of the abnormal expansion areas of edge removal rate is deduced to obtain the edge removal imbalance state matrix.

[0091] Furthermore, the system is also used to implement the following functions: Based on the glass wafer edge collapse evolution path, the edge collapse propagation direction and the shrinkage trend of the effective contact area are identified; based on the shrinkage trend of the effective contact area, edge material removal and redistribution analysis is performed to obtain the secondary offset result of the edge removal rate; based on the secondary offset result of the edge removal rate, the wafer radial thickness difference propagation is extrapolated to obtain the TTV dynamic offset trend; based on the TTV dynamic offset trend, the edge collapse instability criticality is identified to determine the critical constraint interval of edge collapse; based on the edge collapse propagation direction, the secondary offset result of the edge removal rate, the TTV dynamic offset trend, and the critical constraint interval of edge collapse, a spatiotemporal matching mapping is performed to establish the edge collapse-TTV dynamic coupling constraint model.

[0092] Furthermore, the system is also used to implement the following functions: Based on the thermal expansion offset results of the polishing disk, the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone are determined. Temperature compensation collaborative decision-making is performed on the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone according to the collapsed edge-TTV dynamic coupling constraint model to obtain a temperature compensation collaborative space. TTV prediction is performed on each temperature compensation collaborative scheme within the temperature compensation collaborative space to obtain a TTV prediction set. Based on the TTV prediction set, the temperature compensation collaborative space is filtered according to a predetermined TTV to obtain a temperature compensation candidate space. Energy consumption minimization filtering is performed on the temperature compensation candidate space to output the radial partition temperature compensation strategy.

[0093] Furthermore, the system is also used to implement the following functions: Temperature compensation decisions are made for the central thermal expansion region based on the collapsed-TTV dynamic coupling constraint model to obtain a central temperature compensation decision set; temperature compensation decisions are made for the edge thermal expansion region based on the collapsed-TTV dynamic coupling constraint model to obtain an edge temperature compensation decision set; temperature compensation decisions are made for the thermal expansion transition region based on the collapsed-TTV dynamic coupling constraint model to obtain a transition temperature compensation decision set; the central temperature compensation decision set, the edge temperature compensation decision set, and the transition temperature compensation decision set are randomly combined to generate the temperature compensation collaborative space.

[0094] Furthermore, the system is also used to implement the following functions: Anomaly prediction is performed based on the real-time status data of the liquid temperature controller to obtain the abnormal characteristics of the controller; the radial partition temperature compensation strategy is then corrected based on the abnormal characteristics of the controller.

[0095] Furthermore, the system is also used to implement the following functions: Based on the abnormal characteristics of the controller, a controller warning signal is output.

[0096] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Specific embodiments of this specification have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0097] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0098] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for controlling the temperature of a polishing pad during double-sided polishing, characterized by, The method includes: Based on the temperature data of the upper and lower polishing pads, a radial temperature difference analysis of the center edge is performed to generate the thermal expansion offset result of the polishing pads. Based on the thermal expansion offset results of the polishing pad, a contact pressure drift analysis is performed on the glass wafer polishing contact area to obtain the edge contact pressure offset results. Edge removal imbalance is detected by using the edge contact pressure offset results and glass wafer surface material removal rate data, and an edge removal imbalance state matrix is ​​established. The glass wafer edge collapse evolution path is predicted based on the edge removal imbalance state matrix, and TTV offset correlation and edge collapse critical constraint are performed based on the glass wafer edge collapse evolution path to establish an edge collapse-TTV dynamic coupling constraint model. The radial temperature difference compensation iterative decision is made by using the thermal expansion offset result of the polishing pad and the collapsed edge-TTV dynamic coupling constraint model to obtain the radial partition temperature compensation strategy. The radial partition temperature compensation strategy is used to control the liquid temperature controller to perform differentiated temperature regulation on the upper and lower polishing discs.

2. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, Based on the temperature data of the upper and lower polishing pads, a radial temperature difference analysis of the center edge is performed to generate the thermal expansion offset results of the polishing pads, including: The temperature data of the upper polishing disk are used to identify the temperature difference between the center and the edge, and a radial temperature difference curve of the upper polishing disk is constructed. The temperature difference between the center and edge of the lower polishing pad is identified by analyzing the temperature data, and a radial temperature difference curve of the lower polishing pad is constructed. Thermal expansion offset fitting is performed based on the radial temperature difference curves of the upper and lower polishing discs to obtain the thermal expansion offset result of the polishing disc.

3. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, Based on the thermal expansion offset results of the polishing pad, a contact pressure drift analysis is performed on the glass wafer polishing contact area to obtain the edge contact pressure offset results, including: Based on the thermal expansion offset results of the polishing pad, identify the pad surface height offset characteristics; Based on the disk height offset characteristics, pressure distribution mapping is performed on the glass wafer polishing contact area to obtain the contact pressure distribution results; Based on the contact pressure distribution results, edge pressure migration is identified to determine the direction of edge contact pressure drift. The pressure offset is calculated based on the contact pressure distribution results and the edge contact pressure drift direction to obtain the edge contact pressure offset result.

4. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, Edge removal imbalance is detected using the edge contact pressure offset results and glass wafer surface material removal rate data, and an edge removal imbalance state matrix is ​​established, including: Based on the edge contact pressure offset results, the positioning is adjusted to obtain the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone; Based on the material removal rate data of the glass wafer surface, identify the gradient of the instantaneous removal rate change at the edge and the direction of radial removal rate migration at the edge; Based on the wafer edge pressure concentration zone and the wafer edge pressure attenuation zone, the pressure response partitioning of the instantaneous edge removal rate change gradient is performed to identify abnormal expansion regions of edge removal rate; Based on the migration direction of the radial removal rate of the edge, the edge erosion trend of the abnormal expansion area of ​​the edge removal rate is deduced, and the edge removal imbalance state matrix is ​​obtained.

5. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, Based on the glass wafer collapse evolution path, TTV offset correlation and collapse critical constraints are established to create a collapse-TTV dynamic coupling constraint model, including: Based on the glass wafer edge collapse evolution path, identify the edge collapse expansion direction and the shrinkage trend of the effective contact area at the edge; Based on the shrinkage trend of the effective contact area at the edge, an edge material removal and redistribution analysis is performed to obtain the secondary offset result of the edge removal rate. Based on the secondary offset results of the edge removal rate, the wafer radial thickness difference expansion is extrapolated to obtain the TTV dynamic offset trend; Based on the TTV dynamic offset trend, the critical collapse instability of the edge is identified, and the critical constraint range of edge collapse is determined. Based on the edge collapse expansion direction, the edge removal rate secondary offset result, the TTV dynamic offset trend, and the edge collapse critical constraint interval, a spatiotemporal matching mapping is performed to establish the collapse-TTV dynamic coupling constraint model.

6. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, Radial temperature difference compensation iterative decision-making is performed using the thermal expansion offset results of the polishing pad and the collapsed edge-TTV dynamic coupling constraint model to obtain a radial partition temperature compensation strategy, including: Based on the thermal expansion offset results of the polishing pad, the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone are determined; Based on the collapsed edge-TTV dynamic coupling constraint model, temperature compensation collaborative decision-making is performed on the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone to obtain the temperature compensation collaborative space; TTV prediction is performed based on each temperature compensation coordination scheme in the temperature compensation coordination space to obtain a TTV prediction set. Based on the TTV prediction set, the temperature compensation coordination space is filtered according to the predetermined TTV to obtain the temperature compensation candidate space; The energy consumption minimization screening is performed on the temperature compensation candidate space to output the radial partition temperature compensation strategy.

7. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 6, characterized in that, Based on the collapsed-TTV dynamic coupling constraint model, temperature compensation collaborative decision-making is performed on the central thermal expansion zone, the edge thermal expansion zone, and the thermal expansion transition zone to obtain the temperature compensation collaborative space, including: Based on the collapsed edge-TTV dynamic coupling constraint model, temperature compensation decisions are made for the central thermal expansion zone to obtain the central temperature compensation decision set. Based on the collapsed-TTV dynamic coupling constraint model, temperature compensation decisions are made for the edge thermal expansion zone to obtain the edge temperature compensation decision set. Based on the collapsed edge-TTV dynamic coupling constraint model, temperature compensation decisions are made for the thermal expansion transition zone to obtain the transition temperature compensation decision set. The temperature compensation collaborative space is generated by randomly combining the central temperature compensation decision set, the edge temperature compensation decision set, and the transition temperature compensation decision set.

8. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 1, characterized in that, The radial partition temperature compensation strategy controls the liquid temperature controller to perform differentiated temperature regulation on the upper and lower polishing discs, including: Anomaly prediction is performed based on the real-time status data of the liquid temperature controller to obtain abnormal characteristics of the controller; The radial partition temperature compensation strategy is corrected based on the abnormal characteristics of the controller.

9. The method for controlling the temperature of the polishing disc during double-sided polishing as described in claim 8, characterized in that, Obtain controller exception characteristics, including: Based on the abnormal characteristics of the controller, a controller warning signal is output.

10. A temperature control system for a polishing disc during double-sided polishing, characterized in that, The system is used to implement the polishing disc temperature control method during double-sided polishing as described in any one of claims 1-9, and the system comprises: The temperature difference analysis module is used to perform radial temperature difference analysis of the center edge based on the temperature data of the upper polishing pad and the temperature data of the lower polishing pad, and generate the thermal expansion offset result of the polishing pad. The contact pressure drift analysis module is used to perform contact pressure drift analysis on the glass wafer polishing contact area based on the thermal expansion offset result of the polishing pad, and to obtain the edge contact pressure offset result. The edge removal imbalance detection module is used to detect edge removal imbalance by using the edge contact pressure offset results and glass wafer surface material removal rate data, and to establish an edge removal imbalance state matrix. The constraint model establishment module is used to predict the glass wafer edge collapse evolution path based on the edge removal imbalance state matrix, and to perform TTV offset correlation and edge collapse critical constraint based on the glass wafer edge collapse evolution path to establish an edge collapse-TTV dynamic coupling constraint model. The compensation strategy acquisition module is used to make radial temperature difference compensation iterative decisions based on the thermal expansion offset result of the polishing disk and the collapse-TTV dynamic coupling constraint model, and to acquire the radial partition temperature compensation strategy. The differential temperature control module is used to control the liquid temperature controller to perform differential temperature control on the upper and lower polishing discs according to the radial partition temperature compensation strategy.

Images