Wafer thinning system, method, device, electronic equipment and storage medium
By measuring and predicting the wafer morphology and optimizing the pressure response coefficient, high-precision pressure control of the polishing process was achieved, solving the problem of accurate pressure adjustment during polishing and improving the precision and efficiency of wafer thinning.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HWATSING (BEIJING) TECH CO LTD
- Filing Date
- 2023-12-22
- Publication Date
- 2026-07-31
AI Technical Summary
In existing technologies, it is difficult to achieve high precision in adjusting and controlling the pressure in different areas of the polishing head during wafer polishing, which affects the accuracy of wafer thinning.
By measuring the previous and target morphologies of the wafer, the desired applied pressure of the wafer to be polished is predicted using the optimized pressure response coefficient, and polishing is performed based on chemical mechanical polishing equipment. Combined with the iterative update of the pressure response coefficient, the polishing pressure control is optimized.
It improves the precision and efficiency of wafer polishing, reduces wafer material consumption, and achieves better polishing morphology and surface flatness.
Smart Images
Figure CN117697614B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of wafer fabrication technology, and in particular to a wafer thinning system, method, apparatus, electronic device, and computer-readable storage medium. Background Technology
[0002] Backside thinning of wafers refers to the high-precision grinding of semiconductor materials before packaging to reduce their thickness to a suitable shape, thereby reducing the packaging height, chip package size, and improving the chip's thermal diffusion efficiency, electrical performance, and mechanical performance.
[0003] In related technologies, wafer polishing (e.g., chemical mechanical polishing) can be performed after wafer grinding to further improve the flatness of the wafer surface, reduce the total thickness variation (TTV) of the wafer, and remove the damage layer caused by the grinding process.
[0004] During wafer polishing, the wafer is primarily pressed onto a polishing pad using a polishing head. Polishing is achieved through mechanical friction between the wafer and the polishing pad. Since the wafer thickness has already been ground to near a preset thickness, the polishable thickness may be less than 1 nm. This means that only one wafer polishing operation is possible, and the precision requirements are extremely high, resulting in a very low tolerance for error. Therefore, the accuracy of pressure adjustment and control in different areas of the polishing head becomes one of the most critical parameters, directly affecting the wafer thinning accuracy.
[0005] Therefore, how to achieve high-precision control of the pressure during the polishing process after grinding has become a technical problem that needs to be solved by existing technologies. Summary of the Invention
[0006] In view of this, embodiments of this application provide a wafer thinning solution to at least partially solve the above-mentioned problems.
[0007] According to a first aspect of the present application, a wafer thinning system is provided, comprising: a chemical mechanical polishing (CMP) apparatus and a controller applied in a wafer thinning apparatus; the controller is used to control the CMP apparatus to perform the following wafer thinning method: after the wafer is ground, the previous morphology of the wafer is measured, and the desired removal morphology of the wafer is obtained based on the previous morphology and the target morphology; prediction is performed based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished to obtain the desired applied pressure of the wafer to be polished; the CMP apparatus is controlled to perform one CMP polishing operation based on the desired applied pressure to obtain the thinned wafer; wherein the optimized pressure response coefficient is obtained by: estimating the removal morphology error of each reference data group based on each actual removal morphology and each desired applied pressure of each reference data group, and selectively iteratively updating the optimized pressure response coefficient based on each desired applied pressure and each removal morphology error to obtain the optimized pressure response coefficient; wherein the desired applied pressure of each reference data group is obtained by performing prediction based on the desired removal morphology of each reference data group using the optimized pressure response coefficient.
[0008] According to a second aspect of the embodiments of this application, a wafer thinning method is provided, the method comprising: estimating the removal morphology error of each reference data group based on each actual removal morphology and each desired applied pressure of each reference data group; selectively iteratively updating the pressure response coefficient to be optimized based on each desired applied pressure and each removal morphology error to obtain an optimized pressure response coefficient, wherein the desired applied pressure of each reference data group is predicted based on the desired removal morphology of each reference data group using the pressure response coefficient to be optimized; after the wafer is ground, measuring the previous morphology of the wafer, and obtaining the desired removal morphology of the wafer based on the previous morphology and the target morphology; performing prediction based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished to obtain the desired applied pressure of the wafer to be polished; controlling a chemical mechanical polishing (CMP) device to perform one CMP polishing based on the desired applied pressure to obtain a thinned wafer.
[0009] According to a third aspect of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the method described in the second aspect.
[0010] According to a fourth aspect of the embodiments of this application, a computer storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described in the second aspect.
[0011] According to a fifth aspect of the embodiments of this application, a computer program product is provided, including computer instructions that instruct a computing device to perform an operation corresponding to the method described in the second aspect.
[0012] According to the wafer thinning schemes provided in the embodiments of this application, the removal morphology error of each reference data group is estimated based on the actual removal morphology and the expected applied pressure of each reference data group. Based on this, the pressure response coefficient is selectively iteratively updated, which can achieve rapid convergence of the pressure response coefficient, reduce the wafer material loss generated in the early stage of wafer polishing during wafer thinning, improve the accuracy of wafer polishing pressure control, obtain a better wafer polishing morphology, and improve thinning accuracy and efficiency. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.
[0014] Figure 1 This is a schematic diagram of a wafer thinning apparatus applicable to embodiments of this application.
[0015] Figure 2 An application is shown Figure 1 A schematic diagram of the chemical mechanical polishing equipment in a wafer thinning machine.
[0016] Figure 3 This is a structural block diagram of a wafer thinning system as an exemplary embodiment of this application.
[0017] Figure 4 This is a flowchart illustrating a wafer thinning method as an exemplary embodiment of this application.
[0018] Figure 5 This is a schematic diagram of the data flow of a wafer thinning system as an exemplary embodiment of this application.
[0019] Figure 6 This is a flowchart illustrating a wafer thinning method as another exemplary embodiment of this application.
[0020] Figure 7 This is a flowchart illustrating a wafer thinning method as another exemplary embodiment of this application.
[0021] Figures 8A to 8B This is a comparison diagram of the convergence effect of the wafer thinning method implemented in the embodiments of this application.
[0022] Figure 9This is a structural block diagram of a wafer thinning apparatus as an exemplary embodiment of this application.
[0023] Figure 10 This is a schematic diagram of the structure of an electronic device that is an exemplary embodiment of this application. Detailed Implementation
[0024] To enable those skilled in the art to better understand the technical solutions in the embodiments of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art should fall within the protection scope of the embodiments of this application.
[0025] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Where there is no conflict between the embodiments, the following embodiments and features can be combined with each other. The steps in the following method embodiments are for illustrative purposes only and are not intended to limit the invention.
[0026] Chemical Mechanical Polishing (CMP) is one of the key technologies in integrated circuit manufacturing. CMP is a polishing method that combines chemical etching with mechanical polishing. Polishing pressure is a crucial parameter affecting material removal rate and uniformity. In practice, excessive polishing pressure can disrupt the uniform distribution of the polishing slurry on the material surface, leading to uneven removal rates. Conversely, insufficient polishing pressure reduces material removal rates and decreases wafer polishing efficiency.
[0027] To facilitate understanding, the following will combine... Figure 1 The wafer thinning apparatus of this application embodiment will be briefly described below.
[0028] The wafer thinning equipment provided in this embodiment is mainly used for back-side thinning of wafers. The back-side refers to the side of the wafer where no devices are laid, which is generally a substrate. The substrate material can be silicon, silicon oxide, silicon nitride, silicon carbide, sapphire, etc.
[0029] Figure 1 An embodiment of the present invention provides a wafer thinning device, including: a front-end module 1, a polishing module 2, and a grinding module 3.
[0030] Equipment front-end module 1 is used to realize the entry and exit of wafers. Equipment front-end module 1 is set at the front end of the wafer thinning equipment. Equipment front-end module 1 is a transition module that realizes the transfer of wafers from the outside to the inside of the equipment, so as to realize the "dry entry and dry exit" of wafers.
[0031] Grinding module 3 is used to grind the wafer, including rough grinding and fine grinding. Grinding module 3 is located at the end of the wafer thinning equipment.
[0032] Polishing module 2 is used to perform chemical mechanical polishing on the wafer after the grinding is completed. It also has the function of transferring the wafer between these three modules (equipment front-end module 1, grinding module 3, and polishing module 2). Polishing module 2 is located between equipment front-end module 1 and grinding module 3. Figure 1 As shown, polishing module 2 includes chemical mechanical polishing equipment 4.
[0033] Understandable Figure 1 The thinning device shown is merely an example. In other implementations, the grinding module 3 may include multiple grinding passes, such as 3, 4, or 5 passes. Any such variations that achieve wafer thinning by grinding should fall within the scope of this application.
[0034] Figure 2 An exemplary chemical mechanical polishing (CMP) apparatus applied to the aforementioned wafer thinning equipment is shown. As shown, the CMP apparatus 100 mainly includes a support device 102 for holding the wafer 200 to be polished, a polishing disk 104 covered with a polishing pad 106, and a supply device 108 for providing polishing slurry. During wafer polishing, the support device 102 can press the wafer 200 to be polished onto the polishing pad 106, and drive the wafer 200 to be polished to rotate relative to the polishing disk 106 (refer to indicator arrow F1) and move horizontally (refer to indicator arrow F2), while the polishing disk 106 also rotates simultaneously (refer to indicator arrow F3). Under the chemical action of the polishing slurry, the relative movement between the support device 102 and the polishing disk 104 causes the polishing pad 106 to rub against the surface of the wafer 200, thus polishing the wafer 200.
[0035] Based on the aforementioned chemical mechanical polishing equipment 100, various embodiments of this application provide a wafer thinning scheme that can achieve rapid convergence of model parameters, improve wafer polishing efficiency, and thus improve thinning efficiency. The following will provide a detailed description through several embodiments.
[0036] Figure 3 A structural diagram of a wafer thinning system according to an exemplary embodiment of this application is shown. As shown in the figure, the wafer thinning system 300 of this embodiment includes a chemical mechanical polishing device 302 and a controller 304.
[0037] For the composition, structure, and operating principle of the chemical mechanical polishing equipment 302, please refer to [link / reference needed]. Figure 1 and Figure 2 The relevant description of the chemical mechanical polishing equipment 100 is provided in the text.
[0038] The controller 304 is communicatively connected to the chemical mechanical polishing equipment 302 and stores control instructions for performing the wafer thinning method.
[0039] Figure 4 The processing flow of the wafer thinning method according to an exemplary embodiment of this application is described. Figure 5 This diagram illustrates the data flow during the implementation of a wafer thinning method in a wafer thinning system.
[0040] like Figure 4 As shown, the wafer thinning method 400 of this embodiment includes the following steps:
[0041] Step S402: After the wafer is ground, the previous morphology of the wafer is measured, and the desired removal morphology of the wafer is obtained based on the previous morphology and the target morphology. Based on the optimized pressure response coefficient and the actual removal morphology of the wafer to be polished, a prediction is performed to obtain the desired applied pressure of the wafer to be polished.
[0042] The wafer grinding process can be referred to in the above embodiments. The target morphology of the wafer can be the expected morphology after wafer thinning. The difference between the target morphology and the previous morphology can be the desired morphology to be removed.
[0043] refer to Figure 5 The desired removal topography of the wafer to be polished can be input into the pressure response model 306, so that the pressure response model 306 can perform prediction on the desired removal topography based on the pressure response coefficient (i.e., optimize the pressure response coefficient) to obtain the desired applied pressure on the wafer to be polished.
[0044] In this embodiment, the desired applied pressure on the wafer to be polished can be calculated using the following formula 1, based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished:
[0045] P = Y / A (Formula 1)
[0046] In Formula 1 above, P represents the desired applied pressure on the wafer to be polished, Y represents the desired removal of morphology from the wafer to be polished, and A represents the optimized pressure response coefficient.
[0047] In some embodiments, the optimized pressure response coefficients can be presented in matrix form, where each element in the matrix characterizes the relationship between the desired applied pressure and the desired removal of topography in different polishing regions of the wafer to be polished.
[0048] In some embodiments, the optimized pressure response coefficient can be obtained by estimating the removal morphology error of each reference data group based on the actual removal morphology and the expected applied pressure of each reference data group, and selectively iteratively updating the pressure response coefficient to be optimized based on the removal morphology error and the expected applied pressure of each reference data group to obtain the optimized pressure response coefficient.
[0049] In this embodiment, the expected applied pressure for each reference data group is obtained by performing a prediction on the expected removed topography of each reference data group using the pressure response coefficient to be optimized.
[0050] In some embodiments, the topography removal error for each reference data group includes local topography removal error and global topography removal error.
[0051] Specifically, by comparing the two global removal topography errors of two reference data sets in two adjacent optimization rounds (e.g., the current optimization round and the previous optimization round), it can be determined whether to utilize the local removal topography error of the current optimization round and the desired pressure to perform iterative updates on the pressure response coefficient of the current optimization round.
[0052] Step S404: Control the chemical mechanical polishing equipment to perform one chemical mechanical polishing based on the desired applied pressure to obtain a thinned wafer.
[0053] refer to Figure 5 The controller 304 can apply pressure according to the expected pressure of the pressure response model 306, and control each pressure chamber of the chemical mechanical polishing equipment 302 to apply corresponding pressure to each polishing area of the wafer 200 to be polished, so as to perform the polishing process of the wafer 200 to be polished.
[0054] In summary, the wafer thinning scheme provided in this embodiment estimates the expected applied pressure of the wafer to be polished based on the pressure response coefficient, accurately predicts the expected applied pressure of the wafer to be polished, and controls the chemical mechanical polishing equipment to perform the corresponding wafer polishing process accordingly. This can improve the wafer polishing effect and obtain better wafer polishing morphology and surface flatness, that is, the wafer morphology after thinning is better.
[0055] Furthermore, the wafer thinning scheme provided in this embodiment uses the actual removed topography and the expected applied pressure of each reference data set to estimate the removal topography error of the reference data set. Based on this, selective iterative optimization of the pressure response coefficient is performed using each expected applied pressure. This can achieve rapid convergence of the pressure response coefficient, shorten the model training cycle, reduce the model training cost, reduce wafer material loss, improve wafer polishing efficiency, and thus improve thinning efficiency.
[0056] Figure 6 The flowchart illustrating the optimization and update process of the pressure response coefficient in an exemplary embodiment of this application is shown. As shown in the figure, the method 600 of this embodiment mainly includes:
[0057] Step S602: Based on the current optimization round and the previous optimization round among multiple optimization rounds, randomly select the actual removal morphology and expected applied pressure of one reference data group from multiple reference data groups, and use it as the actual removal morphology and expected applied pressure of the current optimization round. Then, determine the optimization pressure response coefficient of the previous optimization round as the optimization pressure response coefficient of the current optimization round.
[0058] In this embodiment, the current optimization round is the optimization round that follows the previous optimization round among multiple optimization rounds. For example, if the current optimization round is the k-th optimization round, the previous optimization round is the (k-1)-th optimization round.
[0059] In some embodiments, multiple actual removal morphologies of multiple historical data groups can be obtained (wherein, the actual removal morphology characterizes the morphology of the wafer after polishing). Based on the given target removal morphology and the actual removal morphology of each historical data group, a similarity calculation is performed to obtain the similarity value between each historical data group and the target removal morphology.
[0060] refer to Figure 5 In some embodiments, after the wafer polishing process is completed, the desired removal morphology, the actual removal morphology, and the desired applied pressure of the wafer can be saved to dataset 308 as historical data for iterative optimization of the pressure response coefficient, thereby continuously improving the accuracy of the expected applied pressure prediction results of pressure response model 306.
[0061] In some embodiments, the similarity value between each historical data set and the target removed topography can be calculated using the following formula 2:
[0062] f(x i )=(x i -x) 2 (Formula 2)
[0063] In formula 2 above, x i Let f(x) represent the actual removed topography (e.g., actual topography removal rate) of the i-th historical data group, and let x represent the target removed topography (e.g., target topography removal rate). i ) represents the similarity value between the i-th historical data group and the target removed shape.
[0064] In some embodiments, each historical data group exceeding a similarity threshold can be identified as a reference data group based on the similarity value between each historical data group and the target removed morphology.
[0065] In some embodiments, based on the similarity value between each historical data group and the target removed shape, the historical data group with the highest similarity value can be sequentially determined as the reference data group until the number of reference data groups meets a preset reference quantity threshold. For example, the 10 historical data groups with the highest similarity values can be determined as the reference data groups.
[0066] Using the above scheme, reference data sets that are closest to the actual removed morphology and the target removed morphology can be selected from each historical data set to eliminate error interference in the sample data. This avoids the problem that error data in the historical data cannot be eliminated because it is always iterated in the pressure response coefficient, thereby improving the convergence speed of the pressure response coefficient.
[0067] In some embodiments, the actual removed topography and expected applied pressure of one reference data set can be randomly selected from each reference data set as the actual removed topography and expected applied pressure for the current optimization round (e.g., the kth optimization round).
[0068] In this embodiment, the same reference data set can be repeatedly extracted. That is, in different optimization rounds, the actual removed topography and expected applied pressure of the same reference data set can be used to perform iterative updates of the pressure response coefficient.
[0069] In this embodiment, the target pressure response coefficient of the previous optimization round can be determined as the initial pressure response coefficient of the current optimization round.
[0070] It should be noted that the initial pressure response coefficient and target pressure response coefficient mentioned above are used to characterize the pressure response coefficients at different update stages within the same optimization round. For example, the initial pressure response coefficient characterizes the pressure response coefficient that has not yet been updated in the current optimization round, while the target pressure response coefficient characterizes the pressure response coefficient that has been updated in the current optimization round. In this embodiment, the target pressure response coefficient of the (k-1)th optimization round is determined as the initial pressure response coefficient of the kth optimization round to achieve iterative updates of the pressure response coefficient in each optimization round.
[0071] Step S604: Based on the actual removed morphology, expected applied pressure, and initial pressure response coefficient of the current optimization round, obtain the global removed morphology error and iterative pressure response coefficient of the current optimization round.
[0072] In some embodiments, the removal morphology prediction can be performed based on the expected applied pressure and the initial pressure response coefficient of the current optimization round to obtain the predicted removal morphology of the current optimization round, and the difference calculation can be performed based on the predicted removal morphology and the actual removal morphology of the current optimization round to obtain the local removal morphology error of the current optimization round.
[0073] In this embodiment, the local removal topography error of the current optimization round can be expressed as the following formula 3:
[0074] ε k =y k -p k A k-1 (Formula 3)
[0075] In formula 3 above, ε k y represents the local removal topography error in the k-th optimization round. k p represents the actual removal topography in the k-th optimization round. k p represents the expected pressure applied in the k-th optimization round. k A k-1 Let A represent the predicted removed morphology in the k-th optimization round. k-1 This represents the initial pressure response coefficient in the k-th optimization round.
[0076] In some embodiments, the initial pressure response coefficient can be iteratively updated using a given pressure response coefficient update formula, based on the local removal of topographic error, the desired applied pressure, and the optimization weight coefficient of the current optimization round, to obtain the iterative pressure response coefficient of the current optimization round.
[0077] In this embodiment, the pressure response coefficient update formula is expressed as the following formula 4:
[0078] A k =A k-1 +w k p k ε k (Formula 4)
[0079] In formula 4 above, A k A represents the iterative pressure response coefficient in the k-th optimization round. k-1 w represents the initial pressure response coefficient for the k-th optimization round (i.e., the target pressure response coefficient for the (k-1)-th optimization round). k p represents the optimization weight coefficient for the k-th optimization round. k ε represents the expected pressure applied in the k-th optimization round. k This represents the local removal of topography error in the k-th optimization round.
[0080] In some embodiments, the optimization weight coefficients of the current optimization round can be obtained by using a given weight coefficient convergence formula, based on the optimization weight coefficients of the previous optimization rounds and the expected pressure applied in the current optimization round.
[0081] In this embodiment, the convergence formula for the weighting coefficients is expressed as the following formula 5:
[0082]
[0083] In formula 5 above, w k w represents the optimization weight coefficient for the k-th optimization round. k-1 p represents the optimization weight coefficient for the (k-1)th optimization round. K This represents the expected pressure applied in the k-th optimization round.
[0084] Experiments have shown that the convergence formula for the weight coefficients provided in this embodiment can achieve the best convergence effect for the pressure response coefficients, effectively shortening the training cycle of the pressure response model.
[0085] In some embodiments, each optimization round from the first optimization round to the current optimization round can be determined as the target round, and the local removal topography error corresponding to each of the multiple reference data groups can be calculated based on the optimization pressure response coefficient of each target round. The variance of the multiple local removal topography errors is calculated to obtain the global removal topography error of the current optimization round.
[0086] For example, when the current optimization round is the kth optimization round, the global removal shape error of the kth optimization round can be obtained by performing the mean calculation based on the k local removal shape errors from the 1st optimization round to the kth optimization round.
[0087] Step S606: Based on the global removal of topography error in the current optimization round and the global removal of topography error in the previous optimization round, determine the target pressure response coefficient for the current optimization round using either the initial pressure response coefficient or the iterative pressure response coefficient.
[0088] In this embodiment, the global topography removal error of the current optimization round can be compared with the global topography removal error of the previous optimization round. If the global topography removal error of the current optimization round is less than the global topography removal error of the previous optimization round, the iterative pressure response coefficient is determined as the target pressure response coefficient of the current optimization round. If the global topography removal error of the current optimization round is greater than or equal to the global topography removal error of the previous optimization round, the initial pressure response coefficient is determined as the target pressure response coefficient of the current optimization round.
[0089] For example, the global removal topography error E in the kth optimization round can be compared. k And the global removal topography error E in the (k-1)th optimization round k-1 If E k <E k-1 The iterative pressure response coefficient A of the current optimization round. k The target pressure response coefficient for the current optimization round is determined as E. k ≥E k-1The initial pressure response coefficient A of the current optimization round is set. k-1 The target pressure response coefficient for the current optimization round is determined (i.e., no iterative update is performed on the initial pressure response coefficient for the current optimization round).
[0090] Step S608: Determine whether the iterative update process of the pressure response coefficient to be optimized meets the given optimization termination condition. If not, proceed to step S610; if yes, proceed to step S612.
[0091] In some embodiments, when the actual number of executions of the optimization round reaches a given execution round threshold, a judgment result can be obtained that the iterative update process of the pressure response coefficient to be optimized meets the given optimization termination condition.
[0092] In some embodiments, when the iterative update change value of the pressure response coefficient tends to stabilize, a judgment result can be obtained that the iterative update process of the pressure response coefficient to be optimized meets the given optimization termination condition.
[0093] Step S610: Update the current optimization round to the previous optimization round, obtain the next optimization round from multiple optimization rounds as the new current optimization round, and return to execute step S602.
[0094] For example, if the current optimization round is the kth optimization round, the kth optimization round can be updated to the previous optimization round, and the (k+1)th optimization round can be determined as the new current optimization round.
[0095] Step S612: Determine the target pressure response coefficient of the current optimization round as the optimized pressure response coefficient.
[0096] Specifically, when the iterative update process of the pressure response coefficient to be optimized meets the given optimization termination condition, the current optimization round is the last optimization round. Therefore, the target pressure response coefficient that has been updated in the current optimization round can be determined as the optimized pressure response coefficient.
[0097] Figure 7 The following figure illustrates the optimization and update process of the pressure response coefficient according to another embodiment of this application. As shown, the method 700 of this embodiment mainly includes:
[0098] Step S702: Obtain the actual removal morphology, expected applied pressure, and initial pressure response coefficient for the current optimization round.
[0099] Specifically, multiple actual removal profiles and multiple expected applied pressures from multiple reference data sets can be obtained to form a dataset. The actual removal profile and expected applied pressure of a parameter data set can be randomly selected from the dataset as the actual removal profile and expected applied pressure of the current optimization round. The target pressure response coefficient of the previous optimization round is determined as the initial pressure response coefficient of the current optimization round.
[0100] For a specific implementation plan for this step, please refer to the detailed description of step S602 above.
[0101] Step S704: Based on the actual removed morphology, expected applied pressure, and initial pressure response coefficient of the current optimization round, obtain the global removed morphology error and iterative pressure response coefficient of the current optimization round.
[0102] In some embodiments, the removal morphology prediction can be performed based on the expected applied pressure and the initial pressure response coefficient of the current optimization round, and the removal morphology prediction result can be compared with the actual removal morphology to obtain the local removal morphology error of the current optimization round. Based on the local removal morphology error of the current optimization round and the expected applied pressure, the initial pressure response coefficient can be iteratively updated to obtain the iterative pressure response coefficient of the current optimization round.
[0103] For a specific implementation plan for this step, please refer to the detailed description of step S604 above.
[0104] Step S706: Determine whether the global topography removal error of the current optimization round is less than the global topography removal error of the previous optimization round. If yes, proceed to step S708; otherwise, proceed to step S710.
[0105] For example, comparing the global topography removal error E in the kth optimization round. k And the global removal topography error E in the (k-1)th optimization round k-1 If E k <E k-1 Execute step S708, if E k ≥E k-1 Execute step S710.
[0106] Step S708: Determine the iterative pressure response coefficient of the current optimization round as the target pressure response coefficient of the current optimization round, and continue to execute step S718.
[0107] Specifically, if the global topography removal error in the current optimization round is less than that in the previous optimization round, it means that the data fitting in the current optimization round is better than that in the previous optimization round. Therefore, the iterative pressure response coefficient A obtained in step S704 can be used. kThe initial pressure response coefficient A for the current optimization round k-1 (i.e., the target pressure response coefficient of the previous optimization round) is updated.
[0108] Step S710: Determine the initial pressure response coefficient of the current optimization round as the target pressure response coefficient of the current optimization round.
[0109] Specifically, if the global topography removal error in the current optimization round is greater than or equal to the global topography removal error in the previous optimization round, it means that the data fitting in the current optimization round has not improved compared to the previous optimization round. In this case, the pressure response coefficient is not updated, and the target pressure response coefficient in the current optimization round is the same as that in the previous optimization round, both being A. k-1 .
[0110] Step S712: Determine whether the actual number of execution rounds of the optimization round is greater than the execution round threshold. If yes, proceed to step S714; otherwise, proceed to step S718.
[0111] In some embodiments, the execution round threshold can be determined based on the initial number of reference data groups in the dataset. Typically, the execution round threshold can be set to 5 times the initial number of reference data groups. For example, if the initial number of reference data groups in the dataset is 10, the execution round threshold can be set to 50 rounds.
[0112] It should be noted that the setting of the execution round threshold is not limited to the above example. Those skilled in the art can make arbitrary adjustments according to the actual model training needs, and this application does not impose any restrictions on this.
[0113] Step S714: Determine whether the actual remaining number of reference data groups in the dataset is less than the remaining threshold of the data group. If yes, proceed to step S720; otherwise, proceed to step S716.
[0114] In some embodiments, the remaining data group threshold can be determined based on the initial number of reference data groups in the dataset. Generally, the execution round threshold can be set to 70% of the initial number of reference data groups. For example, if the initial number of reference data groups in the dataset is 10, the remaining data group threshold can be set to 7 groups.
[0115] Step S716: Remove the reference data group corresponding to the current optimization round from the dataset.
[0116] In this embodiment, if the global topography removal error of the current optimization round is not less than the global topography removal error of the previous optimization round, it indicates that there may be data errors in the reference data set of the current optimization round, which may interfere with the convergence of the pressure response coefficient. Therefore, if the actual remaining number of reference data sets in the dataset is greater than the remaining threshold of the data set, the reference data set of the current optimization round can be removed from the dataset to reduce the interference of data errors in the dataset.
[0117] Step S718: Update the current optimization round to the previous optimization round, obtain the next optimization round from multiple optimization rounds as the new current optimization round, and return to step S702.
[0118] For example, the kth optimization round can be updated to the previous optimization round, and the (k+1)th optimization round can be used as the new current optimization round. Then, return to step S702 to continue the iterative optimization process of the (k+1)th optimization round.
[0119] Step S720: Determine the target pressure response coefficient of the current optimization round as the optimized pressure response coefficient.
[0120] In this embodiment, when the actual number of execution rounds of multiple optimization rounds is greater than the execution round threshold, and the actual remaining number of reference data groups in the dataset is less than the remaining data group threshold, the iterative update process of the pressure response coefficient to be optimized can be judged to meet the optimization termination condition. Then, the loop processing of the optimization round ends, and the target pressure response coefficient of the current optimization round is determined as the optimized pressure response coefficient.
[0121] In summary, the pressure response coefficient optimization scheme provided in this embodiment improves the iterative optimization effect of the pressure response coefficient and enhances the accuracy of the expected applied pressure prediction results by estimating the local and global removal topography errors of the reference data set based on the actual removal topography and the expected applied pressure in each optimization round, and selectively iteratively updating the pressure response coefficient for each optimization round accordingly.
[0122] Furthermore, this embodiment uses minimum error estimation to filter the reference data set, effectively eliminating error interference from historical data sets. Simultaneously, by utilizing the weight coefficient convergence formula provided in this embodiment, the optimal convergence of the stress response coefficient can be achieved while ensuring iterative optimization, thereby effectively shortening the training cycle of the stress response model.
[0123] Tests showed that the traditional model optimization scheme, which incorporates data errors, yielded the following results: [The predicted expected applied pressure and its convergence performance are described below.] Figure 8AAs shown, the expected applied pressure and its convergence effect predicted by the wafer thinning method of this application are as follows: Figure 8B As shown. According to Figure 8B It can be seen that the optimized pressure response coefficient obtained by using the technical solution of this application clearly represents the mapping relationship between the expected applied pressure and the expected removed morphology, and the convergence speed and convergence effect are significantly improved.
[0124] Another embodiment of this application provides a wafer thinning method, the method comprising: estimating the removal morphology error of each reference data group based on each actual removal morphology and each desired applied pressure of each reference data group; selectively iteratively updating the pressure response coefficient to be optimized based on each desired applied pressure and each removal morphology error to obtain an optimized pressure response coefficient, wherein the desired applied pressure of each reference data group is predicted based on the desired removal morphology of each reference data group using the pressure response coefficient to be optimized; after the wafer is ground, measuring the previous morphology of the wafer, and obtaining the desired removal morphology of the wafer based on the previous morphology and the target morphology; performing prediction based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished to obtain the desired applied pressure of the wafer to be polished; controlling a chemical mechanical polishing (CMP) device to perform one CMP polishing based on the desired applied pressure to obtain a thinned wafer.
[0125] It should be noted that the specific implementation schemes of each step in the above wafer thinning method are all the same as those described above. Figures 4 to 7 The various method embodiments are based on the same concept; for details, please refer to [link / reference]. Figures 4 to 7 The relevant descriptions in the various method embodiments are not repeated here.
[0126] Figure 9 This is a schematic diagram of the structure of a wafer thinning apparatus according to an exemplary embodiment of this application. As shown in the figure, the wafer thinning apparatus 900 of this embodiment includes:
[0127] Training module 902 is used to estimate the removal shape error of each reference data group based on each actual removal shape and each expected applied pressure of each reference data group, and to selectively iteratively update the pressure response coefficient to be optimized based on each expected applied pressure and each removal shape error to obtain the optimized pressure response coefficient. The expected applied pressure of each reference data group is obtained by predicting based on the expected removal shape of each reference data group using the pressure response coefficient to be optimized.
[0128] The prediction module 904 is used to measure the previous morphology of the wafer after the wafer has been ground, and to obtain the desired removal morphology of the wafer based on the previous morphology and the target morphology. Based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished, a prediction is performed to obtain the desired applied pressure of the wafer to be polished.
[0129] The polishing module 906 is used to control the chemical mechanical polishing equipment to perform a chemical mechanical polishing operation based on the desired applied pressure to obtain a thinned wafer.
[0130] In some embodiments, the topography removal error for each reference data group includes local topography removal error and global topography removal error.
[0131] In some embodiments, the training module 902 is further configured to: perform a data acquisition step, the data acquisition step comprising: randomly acquiring the actual removed topography and expected applied pressure of a reference data group from the plurality of reference data groups based on the current optimization round and the previous optimization round, as the actual removed topography and expected applied pressure of the current optimization round, and determining the target pressure response coefficient of the previous optimization round as the initial pressure response coefficient of the current optimization round; and obtaining the global removed topography error and iteration of the current optimization round based on the actual removed topography, expected applied pressure, and initial pressure response coefficient of the current optimization round. Pressure response coefficient; based on the global topography removal error of the current optimization round and the global topography removal error of the previous optimization round, the target pressure response coefficient of the current optimization round is determined using either the initial pressure response coefficient or the iterative pressure response coefficient; the current optimization round is updated to the previous optimization round, and the next optimization round is obtained from the plurality of optimization rounds as the new current optimization round, and the data acquisition step is returned to be executed until the iterative update processing of the pressure response coefficient to be optimized meets the given optimization termination condition; the target pressure response coefficient of the current optimization round is determined as the optimized pressure response coefficient.
[0132] In some embodiments, the training module 902 is further configured to: acquire multiple actual removal morphologies from multiple historical data groups, wherein the actual removal morphology characterizes the morphology of the wafer after polishing; perform similarity calculation based on a given target removal morphology and the actual removal morphology of each historical data group to obtain a similarity value between each historical data group and the target removal morphology; and determine each historical data group exceeding a similarity threshold as a reference data group based on the similarity value between each historical data group and the target removal morphology.
[0133] In some embodiments, the training module 902 is further configured to: perform a removal morphology prediction based on the expected applied pressure and the initial pressure response coefficient of the current optimization round to obtain the predicted removal morphology of the current optimization round; and perform a difference calculation based on the predicted removal morphology and the actual removal morphology of the current optimization round to obtain the local removal morphology error of the current optimization round.
[0134] In some embodiments, the training module 902 is further configured to: use a given pressure response coefficient update formula, and iteratively update the initial pressure response coefficient according to the local removal of topography error, the expected applied pressure, and the optimization weight coefficient of the current optimization round, to obtain the iterative pressure response coefficient of the current optimization round;
[0135] The pressure response coefficient update formula is expressed as follows:
[0136] A k =A k-1 +w k p k ε k
[0137] Among them, A k A represents the iterative pressure response coefficient in the k-th optimization round. k-1 w represents the initial pressure response coefficient in the k-th optimization round. k p represents the optimization weight coefficient for the k-th optimization round. k ε represents the expected pressure applied in the k-th optimization round. k This represents the local removal of topography error in the k-th optimization round.
[0138] In some embodiments, the training module 902 is further configured to: obtain the optimization weight coefficients of the current optimization round by using a given weight coefficient convergence formula, based on the optimization weight coefficients of the previous optimization rounds and the expected pressure applied in the current optimization round.
[0139] The convergence formula for the weighting coefficients is expressed as follows:
[0140]
[0141] Among them, w k w represents the optimization weight coefficient for the k-th optimization round. k-1 p represents the optimization weight coefficient for the (k-1)th optimization round. K This represents the expected pressure applied in the k-th optimization round.
[0142] In some embodiments, the training module 902 is further configured to: determine each optimization round from the first optimization round to the current optimization round as a target round; and perform mean calculation based on the local removal topography error of each target round to obtain the global removal topography error of the current optimization round.
[0143] In some embodiments, the training module 902 is further configured to: compare the global topography removal error of the current optimization round with the global topography removal error of the previous optimization round; if the global topography removal error of the current optimization round is less than the global topography removal error of the previous optimization round, determine the iterative pressure response coefficient as the target pressure response coefficient of the current optimization round; if the global topography removal error of the current optimization round is greater than or equal to the global topography removal error of the previous optimization round, determine the initial pressure response coefficient as the target pressure response coefficient of the current optimization round.
[0144] In some embodiments, the training module 902 is further configured to: when the global topography removal error of the current optimization round is greater than or equal to the global topography removal error of the previous optimization round, the wafer thinning method further includes: when the actual number of execution rounds of the plurality of optimization rounds is greater than a given execution round threshold, removing the reference data group corresponding to the current optimization round from the dataset composed of the plurality of reference data groups.
[0145] In some embodiments, the training module 902 is further configured to: when the actual number of execution rounds of the plurality of optimization rounds is greater than the execution round threshold, and the actual remaining number of reference data groups in the dataset is less than the given data group remaining threshold, obtain a judgment result that the iterative update processing of the pressure response coefficient to be optimized satisfies the optimization termination condition.
[0146] This application provides a computer storage medium storing computer program code. When the computer program code is run by a processor, the processor executes the wafer thinning method described in various embodiments of this application.
[0147] An exemplary embodiment of this application provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform a wafer thinning method according to various exemplary embodiments of this application.
[0148] Please refer to Figure 10The present invention describes a structural block diagram of an electronic device 1000 that can serve as a server or client of this application, which is an example of a hardware device that can be applied to various aspects of this application. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the application described and / or claimed herein.
[0149] like Figure 10 As shown, the electronic device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1002 or a computer program loaded from a storage unit 1008 into a random access memory (RAM) 1003. The RAM 1003 may also store various programs and data required for the operation of the device 1000. The computing unit 1001, ROM 1002, and RAM 1003 are interconnected via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.
[0150] Multiple components in electronic device 1000 are connected to I / O interface 1005, including: input unit 1006, output unit 1007, storage unit 1008, and communication unit 1009. Input unit 1006 can be any type of device capable of inputting information to electronic device 1000. Input unit 1006 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 1007 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 1008 may include, but is not limited to, disk and optical disk. Communication unit 1009 allows electronic device 1000 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.
[0151] The computing unit 1001 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1001 performs the various methods and processes described above. For example, in some embodiments, the wafer thinning method described above can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 1008. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 1000 via ROM 1002 and / or communication unit 1009. In some embodiments, the computing unit 1001 can be configured to perform the wafer thinning method described above by any other suitable means (e.g., by means of firmware).
[0152] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0153] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0154] As used in this application, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.
[0155] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0156] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or middleware components (e.g., an application server), or frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0157] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other.
[0158] It should be noted that, depending on the implementation needs, the various components / steps described in the embodiments of this application can be broken down into more components / steps, or two or more components / steps or parts of the operation of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of this application.
[0159] The above embodiments are only used to illustrate the embodiments of this application, and are not intended to limit the embodiments of this application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of this application. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of this application, and the patent protection scope of the embodiments of this application should be defined by the claims.
Claims
1. A wafer thinning system, characterized in that, include: Chemical mechanical polishing equipment and controllers used in wafer thinning equipment; The controller is used to control the chemical mechanical polishing equipment to perform the following wafer thinning method: After the wafer is ground, the previous morphology of the wafer is measured, and the desired removal morphology of the wafer is obtained based on the previous morphology and the target morphology. Based on the optimized pressure response coefficient and the expected removal morphology of the wafer to be polished, a prediction is performed to obtain the expected applied pressure of the wafer to be polished; Control the chemical mechanical polishing equipment to perform a single chemical mechanical polishing operation based on the desired applied pressure, and obtain a thinned wafer; The optimized pressure response coefficient is obtained through the following method: Based on the actual removal morphology and the expected applied pressure of each reference data group, the removal morphology error of each reference data group is estimated. Based on the expected applied pressure and the removal morphology error, the pressure response coefficient to be optimized is selectively iteratively updated to obtain the optimized pressure response coefficient. The expected pressure applied to each reference data set is obtained by performing a prediction based on the expected morphology of each reference data set using the pressure response coefficient to be optimized.
2. The system according to claim 1, wherein, Multiple reference data sets correspond to global removal of topography errors; The step of estimating the removal morphology error of each reference data group based on the actual removal morphology and the expected applied pressure of each reference data group, and selectively iteratively updating the pressure response coefficient to be optimized based on the expected applied pressure and the removal morphology error to obtain the optimized pressure response coefficient includes: The data acquisition step includes: randomly acquiring the actual removal morphology and expected applied pressure of a reference data group from multiple reference data groups based on the current optimization round and the previous optimization round, and using it as the actual removal morphology and expected applied pressure of the current optimization round; and determining the target pressure response coefficient of the previous optimization round as the initial pressure response coefficient of the current optimization round. Based on the actual removed topography, expected applied pressure, and initial pressure response coefficient of the current optimization round, the global removed topography error and iterative pressure response coefficient of the current optimization round are obtained. Based on the global topography removal error of the current optimization round and the global topography removal error of the previous optimization round, the target pressure response coefficient of the current optimization round is determined using one of the initial pressure response coefficient and the iterative pressure response coefficient. The current optimization round is updated to the previous optimization round, and the next optimization round is obtained from the multiple optimization rounds as the new current optimization round. The data acquisition step is then returned to be executed until the iterative update process of the pressure response coefficient to be optimized meets the given optimization termination condition. The target pressure response coefficient of the current optimization round is determined as the optimized pressure response coefficient.
3. The system according to claim 2, characterized in that, Multiple reference data sets were obtained using the following methods: Multiple actual removal morphologies from multiple historical data sets are acquired, and the actual removal morphologies characterize the morphology of the wafer after polishing. Based on the given target removal shape and the actual removal shape of each historical data group, a similarity calculation is performed to obtain the similarity value between each historical data group and the target removal shape; Based on the similarity value between each historical data group and the target removed morphology, each historical data group exceeding the similarity threshold is determined as a reference data group.
4. The system according to claim 2, characterized in that, The topography removal error for each reference data group includes a local topography removal error. Based on the actual topography removal, expected applied pressure, and initial pressure response coefficient of the current optimization round, the local topography removal error for the current optimization round is obtained, including: Based on the expected applied pressure and initial pressure response coefficient of the current optimization round, perform topography removal prediction to obtain the predicted topography removal for the current optimization round; The local removal morphology error of the current optimization round is obtained by performing a difference calculation between the predicted removal morphology and the actual removal morphology of the current optimization round.
5. The system according to claim 4, wherein, Based on the actual removed morphology, expected applied pressure, and initial pressure response coefficient of the current optimization round, the iterative pressure response coefficient of the current optimization round is obtained, including: Using the given pressure response coefficient update formula, the initial pressure response coefficient is iteratively updated based on the local removal of topography error, the expected applied pressure, and the optimization weight coefficient of the current optimization round, to obtain the iterative pressure response coefficient of the current optimization round. The pressure response coefficient update formula is expressed as follows: ,in, Indicates the first The iterative pressure response coefficient of each optimization round. Indicates the first The initial pressure response coefficients for each optimization round. Indicates the first The optimization weight coefficients for each optimization round. Indicates the first The expected pressure of each optimization round, Indicates the first Local removal of topography errors in each optimization round.
6. The system according to claim 5, characterized in that, The optimization weight coefficients for the current optimization round are obtained as follows: Using the given convergence formula for the weight coefficients, and based on the optimization weight coefficients of the previous optimization rounds and the expected pressure applied in the current optimization round, the optimization weight coefficients for the current optimization round are obtained: The convergence formula for the weighting coefficients is expressed as follows: ,in, Indicates the first The optimization weight coefficients for each optimization round. Indicates the first The optimization weight coefficients for each optimization round. Indicates the first The expectation of each optimization round puts pressure on the system.
7. The system according to claim 2, characterized in that, The global removal topography error for the current optimization round is obtained as follows: Each optimization round from the first optimization round to the current optimization round is determined as the target round. Based on the optimized pressure response coefficient for each target round, the local removal topography error corresponding to each of the multiple reference data groups is calculated. The variance of the multiple local removal topography errors is calculated to obtain the global removal topography error for the current optimization round.
8. The system according to any one of claims 2, 4 to 7, characterized in that, The step of determining the target pressure response coefficient for the current optimization round based on the global removal topography error of the current optimization round and the global removal topography error of the previous optimization round, using either the initial pressure response coefficient or the iterative pressure response coefficient, includes: Compare the global topography removal error of the current optimization round with the global topography removal error of the previous optimization round; If the global topography removal error of the current optimization round is less than the global topography removal error of the previous optimization round, the iterative pressure response coefficient is determined as the target pressure response coefficient of the current optimization round. If the global topography removal error of the current optimization round is greater than or equal to the global topography removal error of the previous optimization round, the initial pressure response coefficient is determined as the target pressure response coefficient of the current optimization round.
9. The system according to claim 8, characterized in that, If the global topography removal error of the current optimization round is greater than or equal to the global topography removal error of the previous optimization round, the wafer thinning method further includes: When the actual number of execution rounds of the multiple optimization rounds exceeds the given execution round threshold, the reference data group corresponding to the current optimization round is removed from the dataset consisting of multiple reference data groups.
10. The system according to claim 9, characterized in that, The determination result that the iterative update process of the pressure response coefficient to be optimized satisfies the optimization termination condition is obtained by the following method: When the actual number of execution rounds of the multiple optimization rounds is greater than the execution round threshold, and the actual remaining number of reference data groups in the dataset is less than the given remaining data group threshold, the result of the iterative update process of the pressure response coefficient to be optimized satisfies the optimization termination condition is obtained.
11. A wafer thinning method, characterized in that, The method includes: Based on the actual removed morphology and the expected applied pressure of each reference data group, the removal morphology error of each reference data group is estimated. Based on the expected applied pressure and the removal morphology error, the pressure response coefficient to be optimized is selectively iteratively updated to obtain the optimized pressure response coefficient. The expected applied pressure of each reference data group is predicted based on the expected removed morphology of each reference data group using the pressure response coefficient to be optimized. After the wafer is ground, the previous morphology of the wafer is measured, and the desired removal morphology of the wafer is obtained based on the previous morphology and the target morphology. Based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished, a prediction is performed to obtain the desired applied pressure of the wafer to be polished; The chemical mechanical polishing equipment is controlled to perform a single chemical mechanical polishing operation based on the desired pressure, resulting in a thinned wafer.
12. A wafer thinning apparatus, characterized in that, The device includes: The training module is used to estimate the removal shape error of each reference data group based on each actual removal shape and each expected applied pressure of each reference data group, and to perform selective iterative updates on the pressure response coefficient to be optimized based on each expected applied pressure and each removal shape error to obtain the optimized pressure response coefficient. The expected applied pressure of each reference data group is obtained by performing prediction based on the expected removal shape of each reference data group using the pressure response coefficient to be optimized. The prediction module is used to measure the previous morphology of the wafer after the wafer has been ground, and to obtain the desired removal morphology of the wafer based on the previous morphology and the target morphology. Based on the optimized pressure response coefficient and the desired removal morphology of the wafer to be polished, a prediction is performed to obtain the desired applied pressure of the wafer to be polished. The polishing module controls the chemical mechanical polishing equipment to perform a single chemical mechanical polishing operation based on the desired applied pressure, resulting in a thinned wafer.
13. An electronic device, comprising: The processor, the communication interface, the memory, and the bus are connected, and the processor, the communication interface, and the memory communicate with each other via the bus. The memory is used to store at least one executable instruction that causes the processor to perform an operation corresponding to the method of claim 11.
14. A computer-readable storage medium storing computer instructions that, when executed by a processor, cause the processor to perform the method of claim 11.