Method, system for manufacturing silicon-on-insulator wafer and silicon-on-insulator wafer

Abstract

The present disclosure relates to the technical field of semiconductor manufacturing, and provides a manufacturing method and system of a silicon-on-insulator wafer and the silicon-on-insulator wafer. The manufacturing method comprises polishing a top layer of silicon of a silicon-on-insulator substrate, and obtaining three monitoring parameters respectively representing a remaining thickness of the top layer of silicon, a friction change in a rotation process of a polishing disc and a surface roughness of the top layer of silicon during the polishing; determining effectiveness of each monitoring parameter; when at least two monitoring parameters are effective, determining a monitoring parameter participating in endpoint determination and an endpoint determination rule corresponding to the monitoring parameter according to the effectiveness of each monitoring parameter; determining a polishing endpoint according to the monitoring parameter participating in endpoint determination and the endpoint determination rule, and stopping polishing when the polishing endpoint is reached.

Description

Technical Field

[0001] This disclosure relates to the field of semiconductor manufacturing technology, and in particular to a method, system, and silicon-on-insulator (SiO2) wafer manufacturing method. Background Technology

[0002] The top silicon layer of a silicon-on-insulator substrate can be thinned by chemical mechanical polishing. During the polishing process, endpoint detection is usually required to control the material removal status of the top silicon layer.

[0003] Existing methods for detecting the endpoint of chemical mechanical polishing (CMP) can utilize optical signals, friction signals, and other methods to monitor the polishing process. For example, optical detection can obtain monitoring information related to the film layer on the wafer surface, while friction detection can obtain monitoring information related to the polishing contact state. When the endpoint determination uses fixed monitoring criteria, and the corresponding monitoring parameters are unavailable or cannot be properly used for the current endpoint determination, the criteria used to determine the polishing endpoint will be limited. Summary of the Invention

[0004] This disclosure provides a method for manufacturing a silicon-on-insulator (SiI) wafer. The method involves polishing the top silicon layer of a SiI substrate, and acquiring three monitoring parameters during the polishing process to characterize the remaining thickness of the top silicon layer, the change in frictional force during the rotation of the polishing disc, and the surface roughness of the top silicon layer. The validity of each monitoring parameter is determined. When at least two monitoring parameters are valid, based on the validity of each monitoring parameter, a monitoring parameter participating in the endpoint determination is determined from the three monitoring parameters, and an endpoint determination rule corresponding to the monitoring parameter participating in the endpoint determination is determined. The polishing endpoint is determined based on the monitoring parameter participating in the endpoint determination and the corresponding endpoint determination rule, and polishing is stopped upon reaching the polishing endpoint to obtain a SiI wafer.

[0005] This disclosure also provides a manufacturing system for silicon-on-insulator (SiI) wafers. The SiI wafer manufacturing system includes a monitoring module and a controller. The monitoring module is used to acquire three monitoring parameters during the polishing process of the top silicon layer on a SiI substrate, characterizing the remaining thickness of the top silicon layer, the change in frictional force during the rotation of the polishing disk, and the surface roughness of the top silicon layer. The controller is used to determine the validity of each monitoring parameter; when at least two monitoring parameters are valid, it determines the monitoring parameters participating in the endpoint determination and the corresponding endpoint determination rules based on the validity of each monitoring parameter, and determines the polishing endpoint based on the monitoring parameters participating in the endpoint determination and the corresponding endpoint determination rules.

[0006] This disclosure also provides a silicon-on-insulator (SBI) wafer. The top silicon layer of the SBI wafer is formed by polishing; the thickness of the top silicon layer is 200 nm to 230 nm, and the thickness uniformity is less than 1.6%.

[0007] By determining the monitoring parameters involved in endpoint determination and the corresponding endpoint determination rules based on the validity of various monitoring parameters, the monitoring parameters and determination rules used for endpoint determination can be adjusted according to the validity of the monitoring parameters. This prevents invalid monitoring parameters from being used as the current determination basis, while allowing the remaining valid monitoring parameters to continue to be used to determine the polishing endpoint. When the remaining thickness, frictional force change, and surface roughness are involved in endpoint determination, the endpoint control can also be linked to the degree of material removal of the top silicon layer, the current frictional state, and the surface quality. Attached Figure Description

[0008] Figure 1 This is a schematic diagram illustrating the implementation scenario and device relationships of a silicon-on-insulator (SiO2) wafer manufacturing system.

[0009] Figure 2 This is a schematic diagram of the polishing endpoint determination process in the silicon-on-insulator wafer manufacturing method.

[0010] Figure 3 This is a schematic diagram showing the change in the remaining thickness of the top silicon layer.

[0011] Figure 4 This is a schematic diagram illustrating the changes in friction and the determination of plateau.

[0012] Figure 5 This is a schematic diagram of the online surface roughness variation.

[0013] Figure 6 A flowchart for determining the fault tolerance endpoint under abnormal parameter conditions.

[0014] Figure 7 This is a schematic diagram of the structure and product parameters of silicon-on-insulator wafers. Detailed Implementation

[0015] The technical solution of this disclosure will be described below with reference to the accompanying drawings. The drawings are arranged in the following order: implementation scenario, overall endpoint detection process, formation of different monitoring parameters, endpoint determination rules, fault tolerance handling under abnormal conditions, and processed wafer.

[0016] In chemical mechanical polishing (CMP), the polishing progress is typically assessed using monitoring information related to the state of the material being polished. For example, changes in the wafer surface film can be detected optically, or the polishing state can be determined based on changes in friction or friction-related operating parameters during the polishing process. Among existing endpoint detection schemes, both optical detection and friction force detection can be used to determine the endpoint of CMP.

[0017] For silicon-on-insulator substrates, the thickness of the top silicon layer can be evaluated optically; during chemical mechanical polishing, changes in friction can also reflect the polishing contact state. Different monitoring parameters correspond to different physical states, and in actual online monitoring, there may be situations where a certain monitoring parameter cannot be obtained or the obtained results cannot be used for the current endpoint determination. If the original parameter combination and determination method are still used, the currently available monitoring information may no longer match the endpoint determination criteria.

[0018] To address the above situation, this disclosure acquires three monitoring parameters during the chemical mechanical polishing process of the top silicon layer. First, it determines whether each monitoring parameter is currently valid. Then, based on the validity of each monitoring parameter, it determines the actual monitoring parameters involved in the judgment and the endpoint judgment rule corresponding to the parameter combination. Therefore, subsequent endpoint judgments can correspond to the currently available monitoring information. The following section will first combine... Figure 1 Explain the application scenarios, monitoring objects, and the relationship between the monitoring module and the controller for this endpoint detection scheme.

[0019] The thinning of the top silicon layer on a silicon-on-insulator substrate has a specific thickness target, while the frictional state and surface state near the polishing endpoint also change during the material removal process. Therefore, the three monitoring parameters in this disclosure are not simply collected in parallel, but are used to characterize the remaining thickness, frictional force change, and surface roughness, respectively, and in subsequent steps, the validity of the parameters determines whether they should be included in the current endpoint determination.

[0020] When all three monitoring parameters are obtainable, they are used together to determine the endpoint. However, if one parameter is unobtainable, requiring all parameters to participate will cause a mismatch between the current determination criteria and the actual available data. The participating parameters are selected based on their validity, and corresponding determination rules are configured for different combinations of two parameters, ensuring that normal joint determination and fault-tolerant determination when one parameter is invalid can be seamlessly integrated within the same overall process.

[0021] like Figure 1 As shown, Figure 1 This is a schematic diagram illustrating the implementation scenario and device relationships of a silicon-on-insulator (SiI) wafer fabrication system 100 in one embodiment. The SiI wafer fabrication system 100 is used in the process of chemical mechanical polishing (CMP) of the top silicon layer 21 of a SiI substrate 20 by a CMP apparatus 10. It is used to acquire three monitoring parameters related to the CMP process and determine the endpoint of the CMP based on the acquired monitoring parameters.

[0022] The chemical mechanical polishing apparatus 10 includes a polishing disc 11, a polishing pad 12 disposed on the working side of the polishing disc 11, a carrier head 13 for carrying a silicon-on-insulator substrate 20, and a rotary motor 14 for driving the polishing disc 11 to rotate. During chemical mechanical polishing of the top silicon layer 21, the carrier head 13 brings the silicon-on-insulator substrate 20 into contact with the polishing pad 12, and relative motion is formed between the polishing disc 11 and the silicon-on-insulator substrate 20 to remove material from the top silicon layer 21.

[0023] In one specific embodiment, the polishing pressure can be 2.0 psi, the rotation speed of the polishing disc 11 can be 87 rpm, the rotation speed of the carrier head 13 can be 93 rpm, the slurry flow rate can be 250 mL / min, the polishing pad 12 can be a polyurethane polishing pad, and the polishing fluid can be a colloidal silica slurry suitable for silicon layer polishing.

[0024] The aforementioned process parameters and monitoring parameters are independent of each other. Polishing pressure, grinding disc speed, carrier head speed, and slurry flow rate are used to define the material removal conditions, while the monitoring parameters characterize the current state of the top silicon 21 under these conditions. The controller 120 determines the endpoint based on the monitoring results, rather than directly replacing online endpoint detection with a fixed polishing time.

[0025] One example of thinning the top silicon layer 21 from 300nm to 200nm. This example is used to illustrate... Figure 1 The application scenario of the silicon-on-insulator wafer manufacturing system 100 in the top silicon thinning process is shown, and a unified process background is provided for the specific parameters of subsequent thickness changes, friction state changes and surface roughness changes.

[0026] The silicon-on-insulator (SiI) wafer fabrication system 100 includes a monitoring module 110 and a controller 120. The monitoring module 110 acquires three monitoring parameters during the chemical mechanical polishing (CMP) process and provides these parameters to the controller 120. The controller 120 performs CMP endpoint determination based on the three monitoring parameters provided by the monitoring module 110.

[0027] An optical monitoring module 111 is disposed inside the polishing disc 11 and is used to obtain optical measurement information related to the top silicon 21 during the chemical mechanical polishing process of the top silicon 21. Based on the optical measurement information obtained by the optical monitoring module 111, monitoring parameters for characterizing the remaining thickness state of the top silicon 21 can be formed.

[0028] The optical monitoring module 111 is also used to generate an online roughness estimate related to the surface condition of the top silicon 21. Thus, the same optical monitoring module 111 can provide monitoring parameters characterizing the remaining thickness of the top silicon 21 and monitoring parameters characterizing the surface condition of the top silicon 21, so that thickness information and surface condition information can be provided to the controller 120 in the same chemical mechanical polishing process.

[0029] The surface condition characterized by the online roughness estimate differs from the remaining thickness condition of the top silicon layer 21. The remaining thickness describes the degree of material removal along the thickness direction of the top silicon layer 21, while the roughness describes the deviation of various surface locations within the measurement area relative to a reference surface. Providing both types of information by the same optical monitoring module 111 allows for the maintenance of the surface condition without altering the surface texture. Figure 1 The overall system relationship shown creates two different monitoring dimensions.

[0030] The thickness-related signal of the optical monitoring module 111 can be acquired every 100ms, and the online roughness estimate can be output every 500ms; the friction monitoring unit 112 synchronously acquires friction change information. The controller 120 forms the monitoring status of the current polishing process based on the acquisition time of each data, so that data from different sampling frequencies can be used in the same endpoint detection process.

[0031] A friction monitoring unit 112 is located at the rotary motor 14 below the grinding disc 11 and is used to acquire friction-related information during the rotation of the grinding disc 11. The friction monitoring unit 112 provides the acquired friction-related information to the controller 120, and the continuously acquired friction-related information can form a friction change curve to characterize the friction change state during the rotation of the grinding disc.

[0032] Thus, during the chemical mechanical polishing process of the same top silicon 21, the monitoring module 110 can provide the controller 120 with three monitoring parameters to characterize the remaining thickness of the top silicon 21, the change in friction during the rotation of the polishing disc 11, and the surface roughness of the top silicon 21.

[0033] The controller 120 receives the above monitoring parameters and performs endpoint determination based on the current state of each monitoring parameter. Figure 1 This section primarily describes the implementation scenario of the silicon-on-insulator (SiI) wafer fabrication system 100, the sources of monitoring parameters, and the signal relationship between the monitoring module 110 and the controller 120. The specific endpoint determination process performed by the controller 120 for each monitoring parameter will be discussed later. Figure 2 The process shown is further explained below.

[0034] The controller 120 first uses the three monitoring parameters received to determine whether each monitoring dimension has a basis for judgment, and then proceeds to the endpoint condition judgment. This order avoids directly comparing unavailable monitoring parameters with the endpoint threshold, thus making the determination of the validity of monitoring parameters a preprocessing step before the selection of endpoint determination rules.

[0035] Figure 1 A signal relationship was established for monitoring information to enter the controller 120 from the polishing object and equipment side; Figure 2The above relationship is further transformed into a processing flow from S210 to S240, thereby illustrating how the monitoring parameters gradually form the current endpoint determination result from the input information.

[0036] like Figure 2 As shown, Figure 2 This is a schematic diagram of the polishing endpoint determination process in the silicon-on-insulator wafer manufacturing method. Figure 2 The process shown can be... Figure 1 The silicon-on-insulator (SiI) wafer fabrication system 100 performs chemical mechanical polishing (CMP) on the top silicon layer 21 of the SiI substrate 20 in a chemical mechanical polishing apparatus 10 to determine the CMP endpoint. The process includes steps S210 to S240.

[0037] In step S210, three monitoring parameters are obtained during the chemical mechanical polishing process.

[0038] The three monitoring parameters are used to characterize the remaining thickness of the top silicon layer 21, the change in friction during the rotation of the polishing disk 11, and the surface roughness of the top silicon layer 21, respectively. Different monitoring parameters correspond to different process information and can have different data formats and acquisition methods.

[0039] In step S220, the validity of each monitoring parameter is determined.

[0040] Validity is used to characterize whether the corresponding monitoring parameter can currently serve as the basis for endpoint determination. A monitoring parameter is considered valid if it can be obtained normally; it is considered invalid if it cannot be obtained. Validity is independent of whether the monitoring parameter has met the endpoint conditions; the former determines whether the monitoring parameter can participate in the determination, while the latter determines whether the monitoring parameter participating in the determination has reached the corresponding endpoint state.

[0041] Validity and endpoint status belong to different judgment levels. For example, remaining thickness. Even if the thickness can be obtained normally, if the target thickness range has not yet been reached, the thickness monitoring parameter is still a valid parameter, but the corresponding endpoint condition has not yet been met. Conversely, if the thickness monitoring result cannot be obtained, it cannot be judged whether it meets the standard based solely on the thickness endpoint condition, but should be treated as an invalid parameter first.

[0042] Similarly, the friction force change curve can be formed normally, but When the roughness is not below the threshold, the second monitoring parameter remains valid; the online roughness estimate can be output normally, but... When the roughness exceeds the roughness threshold, the third monitoring parameter remains a valid parameter. Therefore, validity is used to determine the judgment criteria, and the endpoint condition is used to determine the current state represented by the valid parameter.

[0043] If a monitoring parameter cannot be obtained, the corresponding monitoring parameter can be determined as invalid; if the monitoring parameter can be obtained, the corresponding monitoring parameter can be determined as valid. Validity is only used to determine whether a parameter can participate in the current endpoint determination, and is not equivalent to whether the monitoring parameter has reached the corresponding endpoint condition.

[0044] In step S230, when at least two monitoring parameters are valid, the monitoring parameters participating in the endpoint determination are determined from the three monitoring parameters based on the validity of each monitoring parameter, and the endpoint determination rule corresponding to the monitoring parameters participating in the endpoint determination is determined.

[0045] The monitoring parameters used for endpoint determination are determined by the currently valid monitoring parameters. When all three monitoring parameters can be used for endpoint determination, the endpoint determination is made based on all three monitoring parameters. When one of the three monitoring parameters cannot be used for endpoint determination, the determination is made based on the remaining two valid monitoring parameters and the endpoint determination rules corresponding to the combination of those two monitoring parameters.

[0046] When all three monitoring parameters are valid, the parameter combination involved in the endpoint determination consists of the first monitoring parameter, the second monitoring parameter, and the third monitoring parameter. The normal determination rule requires that the three monitoring parameters satisfy the corresponding endpoint conditions respectively. This rule enables the thickness state, current friction state, and surface state of the top silicon 21 to form a joint constraint at the same endpoint time.

[0047] When one of the three monitoring parameters is invalid, the controller 120 no longer requires that invalid parameter to meet the corresponding endpoint condition. Instead, it selects a fault-tolerance judgment rule based on the combination of the remaining two valid parameters. The fault-tolerance judgment rule can reconfigure the weights and comprehensive judgment thresholds of the two valid parameters and require the two valid parameters to meet their respective endpoint conditions; it does not confirm the endpoint based on the only remaining single monitoring parameter.

[0048] In step S240, the polishing endpoint is determined based on the monitoring parameters involved in the endpoint determination and the corresponding endpoint determination rules.

[0049] When the monitoring parameters involved in the endpoint determination meet the corresponding endpoint determination rules, the chemical mechanical polishing endpoint is determined to have been reached; if the corresponding endpoint determination rules are not met, chemical mechanical polishing continues and the corresponding monitoring parameters are obtained.

[0050] The endpoint result output in step S240 also serves as the triggering basis for subsequent control of the chemical mechanical polishing device 10. Upon reaching the endpoint, the device enters a low-pressure finishing state and stops polishing after a preset time; if the endpoint is not reached, the current polishing continues and the monitoring process is returned to update the three monitoring parameters. Thus, the endpoint detection result and the polishing process timing form a closed loop.

[0051] Based on steps S210 to S240, the monitoring parameters currently involved in the endpoint determination and the corresponding endpoint determination rules can be determined according to the validity of each monitoring parameter. The three monitoring parameters reflect the remaining thickness of the top silicon layer 21, the change in friction during the rotation of the polishing disk 11, and the surface roughness of the top silicon layer 21, respectively. To further illustrate the process state characterized by different monitoring parameters, the following first combines... Figure 3 This describes the monitoring parameters used to characterize the remaining thickness of the top silicon 21 layer.

[0052] like Figure 3 As shown, in Figure 2 Of the three monitoring parameters obtained in step S210, the monitoring parameter used to characterize the remaining thickness of the top silicon 21 is based on the remaining thickness of the top silicon 21 during the chemical mechanical polishing process. Sure. Figure 3 The remaining thickness is shown. With polishing time The relationship of change, where the horizontal axis represents polishing time. The vertical axis represents the remaining thickness of the top silicon 21.

[0053] As chemical mechanical polishing continues, the top silicon 21 material is gradually removed, leaving a thickness of... The thickness is reduced accordingly. During the polishing process, the optical monitoring module 111 acquires optical measurement information related to the top silicon layer 21 and determines the remaining thickness of the top silicon layer 21 based on the optical measurement information. This forms the first monitoring parameter used to characterize the remaining thickness state of the top silicon 21.

[0054] Set the target thickness for the predetermined processing thickness of the top silicon 21. and the target thickness The corresponding target thickness range. When the remaining thickness... Upon reaching the target thickness range, it can be determined that the first monitoring parameter meets the corresponding thickness endpoint condition. The target thickness range indicates that the current material removal degree of the top silicon 21 has reached the predetermined thickness state.

[0055] The fact that the first monitoring parameter meets the corresponding endpoint condition does not directly indicate that chemical mechanical polishing has reached the final endpoint. Figure 2 In steps S230 and S240, the endpoint of chemical mechanical polishing is determined jointly based on other monitoring parameters currently involved in the endpoint determination and the endpoint determination rules corresponding to the monitoring parameters involved in the endpoint determination.

[0056] The target thickness and target thickness range of the top silicon layer 21 are set according to the product specifications of the silicon-on-insulator substrate 20 to be processed. In the product state, the thickness of the top silicon layer 21 can be 200nm to 230nm; during online endpoint determination, the target thickness can be determined according to the specific product within this product specification range, and a corresponding target thickness range is set around the target thickness. During online monitoring, the optical monitoring module 111 acquires optical measurement information according to a predetermined sampling period and updates the remaining thickness accordingly. .

[0057] The thickness status can also be converted into a criterion for comprehensive evaluation. In a specific setting, when the remaining thickness of the top silicon 21... When the thickness falls within the target thickness range set according to the product specifications, Take the target status; when When outside this range, The status is not met.

[0058] The target thickness range can be set in combination with the product thickness specification of 200nm to 230nm and the material removal control precision of the process, so that the endpoint thickness condition corresponds to the specific product specification.

[0059] Figure 3 This section describes the remaining thickness state of the top silicon layer 21 and its corresponding thickness endpoint conditions. Besides the remaining thickness state, the frictional state during chemical mechanical polishing can also be used as monitoring information obtained in step S210. The following section combines this with... Figure 4 This further explains the relationship between changes in frictional force and the state of friction.

[0060] like Figure 4 As shown, the friction monitoring unit 112 acquires friction information during the rotation of the polishing disc 11, and forms a friction change curve based on the friction information continuously acquired over the polishing time. Friction force variation curve constitute Figure 2 The second monitoring parameter used to characterize the friction state in step S210 is used to monitor the friction force and the thickness of the top silicon 21 simultaneously.

[0061] As the top silicon layer 21 is gradually thinned, the frictional force change curve... The process involves a gradual stabilization of frictional force after significant initial changes. The rate of change in frictional force is relatively high in the initial polishing stage, then gradually decreases. As the polishing progresses towards the target thickness, the frictional force changes become more gradual, forming a stable region. This stable region characterizes the current frictional state as stabilizing.

[0062] To determine the state of frictional force change, the frictional force change curve is used. Determine its first derivative And the absolute value of the first derivative and the friction threshold are used. Comparison. When Below the friction threshold At that time, the change in frictional force per unit time decreases; combined with the frictional force change curve The changes within the corresponding time interval can be used to further determine whether the friction force change curve is plateaued.

[0063] The plateau state corresponds to a state where the friction force change curve remains relatively stable over a period of time. To avoid treating the instantaneous change at a single sampling moment as a steady state, a duration can be set. .when Below the friction threshold And continue to reach And the friction force change curve When the time interval is plateaued, the second monitoring parameter is determined to meet the endpoint condition corresponding to the friction state.

[0064] Friction threshold It can be set to 0.5 N / s, duration It can be 1.0s. When A frictional force of <0.5 N / s sustained for at least 1.0 s can be used as one of the criteria for judging that the frictional force change tends to stabilize. The stated value is a specific parameter setting used to determine the frictional state.

[0065] Historical data on the rate of change of friction force can further illustrate the selection of the friction state threshold. During the thinning of the top silicon layer 21 from 300nm to 260nm, the rate of change of friction force can range from 1.5N / s to 3.0N / s; in the 260nm to 220nm stage, it can range from 0.8N / s to 1.5N / s; and in the 220nm to 200nm stage, it can range from 0.3N / s to 0.8N / s. Based on the trend of the rate of change tending to be lower during the fine polishing stage, 0.5 N / s can be selected.

[0066] In the comprehensive evaluation, the friction determination result can be recorded as: .when When the velocity is <0.5 N / s for at least 1.0 s and plateaus occur, Take a stable state; if the corresponding conditions are not met, We consider an unsteady state. Therefore, the instantaneous value of the frictional force itself is not the endpoint condition; the continuous rate of change and the plateau state are used together to characterize the stabilization of the polishing process.

[0067] and Figure 3Similar to the thickness endpoint condition in [the text], the friction force change curve satisfying the above conditions indicates that the monitoring parameters used to characterize the friction state have reached the corresponding endpoint state, but does not independently determine whether chemical mechanical polishing ultimately reaches its endpoint. Figure 2 In steps S230 and S240, it is also necessary to make a judgment based on the monitoring parameters currently involved in the endpoint determination and the corresponding endpoint determination rules.

[0068] Figure 4 This explains the changes in frictional force and their corresponding frictional states. Besides the remaining thickness state and frictional state of the top silicon layer 21, the surface state of the top silicon layer 21 can also serve as another monitoring dimension in step S210, which will be discussed below. Figure 5 Explain the changes in online surface roughness with the chemical mechanical polishing process and the corresponding roughness endpoint conditions.

[0069] like Figure 5 As shown, the optical monitoring module 111 acquires optical measurement information related to the surface state of the top silicon 21 during the chemical mechanical polishing process, and obtains an online roughness estimate based on the optical measurement information. Online roughness estimate This constitutes a third monitoring parameter used to characterize the surface state of the top silicon 21 layer.

[0070] Roughness is characterized by the degree of deviation of each position on the surface within the measurement area from the reference plane. As chemical mechanical polishing proceeds, the surface condition of the top silicon layer 21 changes, and the online roughness estimate... Corresponding changes. To determine whether the current surface state has reached the predetermined state, a roughness threshold is set. When the online roughness estimate Not greater than the roughness threshold When the surface state reaches the corresponding endpoint state, it can be determined.

[0071] To reduce the impact of instantaneous measurement fluctuations on the judgment results, a duration can also be set. When the online roughness estimate Not greater than the roughness threshold And continue to reach At that time, it is determined that the third monitoring parameter satisfies the endpoint condition corresponding to the surface state. Roughness threshold It can be set to 0.4nm, duration It can be set to 2.0s, and the online roughness estimate can be updated every 500ms.

[0072] Online roughness estimates can also generate roughness judgment results for comprehensive evaluation. .when When ≤0.4nm and lasting for at least 2.0s, The target condition is met; when the online roughness estimate is higher than 0.4 nm, The non-compliant state is recorded. Continuous conditions ensure that surface condition determination is based on continuous monitoring results, rather than a single sampling point.

[0073] The roughness estimate obtained from online optical measurement can retain a margin of approximately 20% compared to the offline atomic force microscopy measurement. This margin is used to compensate for differences caused by factors such as vibration and slurry residue during online measurement, ensuring that when the online endpoint determination reaches 0.4 nm, the corresponding product roughness target can still be accepted as less than 0.5 nm.

[0074] The online roughness threshold and the final wafer product roughness specification use different values. When online optical measurements are affected by factors such as equipment vibration and polishing slurry residue, a measurement margin can be reserved relative to the product roughness specification. When the product roughness requirement for top silicon 21 is... When the roughness threshold is less than 0.5 nm, the online roughness threshold is... It can be set to 0.4nm, with a 20% margin to distinguish between online judgment thresholds and product acceptance specifications.

[0075] and Figure 3 and Figure 4 The monitoring parameters shown are the same, and the online roughness estimate reaching the corresponding endpoint does not independently determine the final chemical mechanical polishing (CMP) endpoint. Thus, Figures 3 to 5 The remaining thickness of the top silicon layer 21, the change in friction during the rotation of the polishing disc 11, and the surface roughness of the top silicon layer 21 have already been explained. The following section will combine... Figure 6 This section explains the judgment rules that can be used when the above monitoring parameters are used in the endpoint determination.

[0076] exist Figures 3 to 5 After obtaining the status corresponding to different monitoring parameters, Figure 2 The step S230 shown determines the corresponding endpoint determination rule based on the monitoring parameters currently involved in endpoint determination.

[0077] When all three monitoring parameters characterizing remaining thickness, frictional force change, and surface roughness are valid, a joint compliance rule can be used to determine the endpoint. The joint compliance rule determines whether the remaining thickness of the top silicon layer 21 is within the target thickness range, whether the absolute value of the first derivative of the frictional force change curve is below the frictional force threshold and plateaus, and whether the online roughness estimate is not greater than the roughness threshold. When all three monitoring parameters meet their respective endpoint conditions, the chemical mechanical polishing endpoint is determined to have been reached.

[0078] In addition to the joint compliance rule where each of the three monitoring parameters meets its corresponding endpoint condition, step S230 can also employ a comprehensive evaluation rule under appropriate process configurations or fault-tolerant conditions. The comprehensive evaluation rule obtains corresponding judgment results based on each monitoring parameter, and then calculates a comprehensive evaluation value based on the judgment results and their corresponding weights. The results of thickness determination, friction determination, and roughness determination are respectively denoted as... , and Each judgment result obtains its corresponding state value based on whether the corresponding endpoint conditions are met.

[0079]

[0080] in, , and These represent the weights of the thickness monitoring parameter, friction monitoring parameter, and roughness monitoring parameter, respectively. (Comprehensive evaluation value) This is used to combine multiple judgment results currently involved in the endpoint determination into an evaluation result that can be compared with the comprehensive judgment threshold.

[0081] When three monitoring parameters are used in the comprehensive evaluation It can be 0.5. It can be 0.3. It can be 0.2. When When the value is not less than 0.7 and the first monitoring parameter, the second monitoring parameter, and the third monitoring parameter respectively meet their corresponding endpoint conditions, the endpoint confirmation result can be output; when 0.4 ≤ When the pressure is less than 0.7, an endpoint warning result can be output, and the polishing pressure will be reduced to 50% of the current pressure to enter the fine polishing state while continuing monitoring; when When the value is less than 0.4, continue chemical mechanical polishing while maintaining the current process parameters.

[0082] Under normal circumstances, when all three monitoring parameters can be effectively obtained, the joint compliance rule where each of the three monitoring parameters meets its corresponding endpoint condition is preferred. That is, the endpoint is confirmed when the thickness, friction, and surface conditions simultaneously reach their respective endpoints. The comprehensive evaluation rule is mainly used for tolerance determination of two parameters after one monitoring parameter becomes invalid; if the process configuration allows for comprehensive evaluation, it can also be used based on the comprehensive evaluation value when all three monitoring parameters are valid. Determine the destination.

[0083] The applicability relationship between the joint compliance rules and the comprehensive evaluation rules is as follows: Under normal circumstances, the three monitoring parameters satisfying their respective endpoint conditions is used as the primary endpoint confirmation method; when one monitoring parameter cannot be obtained, the remaining two valid parameters form a new parameter combination, and the comprehensive evaluation rules corresponding to this combination are adopted, requiring the two valid parameters to satisfy their respective endpoint conditions; when pre-configured and allowed, the comprehensive evaluation value can also be used for judgment when all three parameters are valid.

[0084] Therefore, the weights are not fixed constants that replace the physical meaning of each monitoring parameter, but rather judgment configurations corresponding to the current combination of parameters involved in the endpoint determination. When the combination of monitoring parameters changes, the corresponding weights and comprehensive judgment thresholds can change synchronously, ensuring that the comprehensive evaluation results correspond to the actual available monitoring information.

[0085] After determining that the chemical mechanical polishing endpoint has been reached, the polishing pressure can be reduced for low-pressure finishing, and chemical mechanical polishing can be stopped after a preset time has elapsed during low-pressure finishing. For example, the low-pressure finishing time can be 2 seconds. The control process after this endpoint is a further application of the endpoint determination result and does not change the... Figure 2 The main process for endpoint detection in steps S210 to S240.

[0086] When all three monitoring parameters can participate in the endpoint determination, a joint compliance rule or a comprehensive evaluation rule can be used. When one of the monitoring parameters cannot participate normally in the current endpoint determination, the combination of participating parameters and the corresponding rules change accordingly. The following section discusses this in conjunction with... Figure 6 This section further explains the fault tolerance determination process.

[0087] When one of the three monitoring parameters is invalid Figure 2 Steps S220 and S230 further determine the remaining two valid monitoring parameters and their corresponding endpoint determination rules.

[0088] like Figure 6 As shown, this may specifically include steps S610 to S650.

[0089] In step S610, it is determined whether there is one of the three monitoring parameters that cannot be used for the current endpoint determination.

[0090] If there are no invalid monitoring parameters, the endpoint determination rules corresponding to the three monitoring parameters can be used for judgment, that is, step S650 is executed directly; if there is one invalid monitoring parameter, then step S620 is executed.

[0091] In step S620, invalid monitoring parameters are excluded from the current endpoint determination.

[0092] In step S630, the current combination of monitoring parameters participating in the endpoint determination is determined based on the remaining two valid monitoring parameters.

[0093] In step S640, the corresponding fault tolerance judgment rule is determined based on the current combination of monitoring parameters involved in the endpoint determination.

[0094] The fault tolerance judgment rule can adjust the weights of the two valid monitoring parameters and the comprehensive judgment threshold to allow the remaining two valid monitoring parameters to continue to participate in the endpoint judgment, and require the two monitoring parameters to meet their respective endpoint conditions.

[0095] In step S650, it is determined whether the chemical mechanical polishing has reached the endpoint based on the remaining valid monitoring parameters and the corresponding fault tolerance judgment rules.

[0096] Different fault-tolerance configurations can be formulated for the three types of monitoring parameters. For example:

[0097] Therefore, if one of the three monitoring parameters cannot be used normally for the current endpoint determination, it is not necessary to continue to make a judgment according to the original fixed parameter combination. Instead, the participating parameters and corresponding judgment rules are re-determined based on the other two valid monitoring parameters, and the endpoint determination in step S240 is completed.

[0098] When the thickness monitoring parameter is invalid, the fault tolerance rule uses the friction parameter and the roughness parameter, with weights that can be configured to 0.7 and 0.3 respectively; when the friction monitoring parameter is invalid, the thickness parameter and the roughness parameter are used, with weights that can be configured to 0.7 and 0.3 respectively; when the roughness monitoring parameter is invalid, the thickness parameter and the friction parameter are used, with weights that can be configured to 0.6 and 0.4 respectively. These configurations correspond to different combinations of remaining valid parameters.

[0099] All three combinations of two parameters mentioned above assume that each of the two monitored parameters meets its respective endpoint condition as a prerequisite for endpoint confirmation. Specifically, the thickness parameter and roughness parameter directly constrain the degree of material removal and surface quality, respectively; the thickness parameter and friction parameter constrain the material removal endpoint through target thickness and friction force plateauing, respectively, with surface roughness verified by post-processing measurement; the friction force parameter and roughness parameter characterize the stabilization of the material removal process through friction force plateauing and characterize the surface state through roughness, with the thickness and uniformity of the treated top silicon layer verified by product measurement.

[0100] The principle behind using the above two parameters for endpoint confirmation is that, although the three monitoring parameters correspond to different physical quantities, they all evolve with the removal of the top silicon material and exhibit corresponding endpoint characteristics near the endpoint. The remaining thickness entering the target thickness range directly characterizes the degree of top silicon material removal. As the top silicon gradually thins, the rate of change in friction gradually decreases, and the friction change curve tends to flatten and form a plateau segment near the target thickness. Therefore, this can be used to characterize that the current chemical mechanical polishing state has entered a stable stage near the endpoint. The online roughness estimate meets the roughness endpoint condition, indicating that the surface state of the top silicon has reached the predetermined state. Therefore, when the thickness parameter is combined with any state parameter, the thickness parameter can directly define the material removal endpoint, and another state parameter can further constrain the polishing state or surface state. In the combination of friction and roughness parameters without a thickness parameter, the friction plateau is used to characterize that the polishing process has entered a stable stage near the target thickness, and the roughness parameter is used to define the surface state. When both simultaneously meet the endpoint condition, they jointly form the endpoint determination basis. The thickness and uniformity of the processed top silicon are then verified through product measurement.

[0101] Figure 2 Step S220 is used to distinguish whether the monitoring parameters can be used for endpoint determination. If the monitoring parameters cannot be obtained, they are determined to be invalid; if the monitoring parameters can be obtained, they are determined to be valid. Whether the monitoring parameters meet the corresponding endpoint conditions is determined in the subsequent endpoint condition judgment.

[0102] Figure 6 This further clarifies the validity of the monitoring parameters, the monitoring parameters involved in endpoint determination, and the relationship between the corresponding determination rules. Based on the clearly defined methods for forming the monitoring parameters, the normal determination rules, and the anomaly tolerance rules, the following examples and comparative examples further illustrate the control or verification relationship between different endpoint control methods and the top silicon thickness and surface roughness of the processed silicon-on-insulator wafer. Unless otherwise specified, each example employs... Figure 1 The chemical mechanical polishing apparatus and process conditions are as follows: initial top silicon thickness is 300 nm, example target thickness is 200 nm, polishing pressure is 2.0 psi, polishing disc 11 rotation speed is 87 rpm, carrier head 13 rotation speed is 93 rpm, slurry flow rate is 250 mL / min, polishing pad 12 is a polyurethane polishing pad, and polishing fluid is colloidal silica slurry.

[0103] The thickness of the top silicon layer can be set from 200nm to 230nm according to product specifications, and the thickness of the top silicon layer at multiple measurement locations on the wafer surface can be statistically analyzed to evaluate thickness uniformity.

[0104] Example 1: Joint compliance determination when all three monitoring parameters are valid.

[0105] The optical monitoring module 111 updates the remaining thickness information of the top silicon 21 at a sampling period of 100ms, and the friction monitoring unit 112 synchronously obtains the friction change information. The online roughness estimate is updated at a period of 500ms. The controller 120 first confirms that the three monitoring parameters are valid, and then judges the thickness state, friction state and surface state respectively. When the remaining thickness enters the target thickness range, the absolute value of the first derivative of the friction change curve is less than 0.5N / s and lasts for at least 1.0s and the curve plateaus, and the online roughness estimate is no greater than 0.4nm and lasts for at least 2.0s, the controller 120 confirms that the endpoint has been reached, and puts the chemical mechanical polishing device 10 into a low-pressure finishing state for 2s before stopping.

[0106] In this endpoint control test, the material removal amount was controlled with an accuracy of ±2nm, and the surface roughness after treatment met the product acceptance requirements, i.e., less than 0.5nm. For product specifications with a top silicon thickness of 200nm to 230nm, the corresponding endpoint thickness range can be set according to the specific product target thickness, and the thickness and uniformity of the treated top silicon can be measured.

[0107] Example 2: Comprehensive evaluation and judgment when all three monitoring parameters are valid.

[0108] When all three monitoring parameters are valid, a comprehensive evaluation rule can also be used. When the judgment results corresponding to thickness, friction, and roughness all reach the corresponding state, the weighted comprehensive evaluation value of the three judgment results reaches the endpoint confirmation threshold. The controller 120 confirms the endpoint accordingly and enters low-pressure fine-tuning in the aforementioned manner. Product acceptance in this embodiment can be performed according to a top silicon thickness of 200nm to 230nm, thickness uniformity of less than 1.6%, and surface roughness of less than 0.5nm. Specific measured values ​​are obtained through product measurement.

[0109] Example 3: Fault tolerance determination when thickness monitoring parameters are invalid.

[0110] When thickness monitoring parameters are unavailable, the controller 120 excludes them, and friction and roughness monitoring parameters are used to determine the current endpoint. The weights of the friction and roughness monitoring parameters are configured as 0.7 and 0.3, respectively. When both parameters meet their corresponding endpoint conditions, the overall evaluation result meets the endpoint confirmation criteria. A plateau in frictional changes indicates that the material removal process is stabilizing, and a roughness threshold indicates that the surface condition meets the standard. Both are used to determine the endpoint. Since thickness monitoring parameters are not involved in the current determination, the thickness and uniformity of the processed top silicon layer are verified through product measurement. Product acceptance can be performed based on a thickness of 200nm to 230nm, thickness uniformity of less than 1.6%, and surface roughness of less than 0.5nm.

[0111] Example 4: Fault tolerance determination when friction monitoring parameters are invalid.

[0112] When friction monitoring parameters are unavailable, the controller 120 makes a judgment based on thickness and surface roughness monitoring parameters, with weights configured as 0.7 and 0.3, respectively. When the thickness falls within the target thickness range and the surface roughness meets the online judgment conditions, the comprehensive evaluation result reaches the corresponding endpoint confirmation condition. Thickness parameters directly constrain the degree of material removal, while surface roughness parameters directly constrain surface quality; therefore, the endpoint can be jointly determined when both meet their corresponding conditions.

[0113] Example 5: Fault tolerance determination when roughness monitoring parameters are invalid.

[0114] When online roughness monitoring parameters are unavailable, the controller 120 uses thickness monitoring parameters and friction monitoring parameters for further judgment, with weights configured as 0.6 and 0.4 respectively, and a comprehensive judgment threshold set to 0.6. When the thickness reaches the target thickness range and the friction reaches a stable state, the controller 120 confirms the endpoint and performs low-pressure finishing. The thickness parameter directly constrains the degree of material removal, while the friction plateau characterizes the material removal process as it tends to stabilize; therefore, both can jointly determine the endpoint when their corresponding conditions are met. Since the online roughness parameter is not involved in the current endpoint determination, the actual surface roughness of the processed wafer is confirmed through product measurement.

[0115] Comparative Example 1 uses only the optical thickness signal for endpoint determination.

[0116] As an existing single-parameter endpoint detection method, the endpoint is confirmed only by the fact that the top silicon thickness obtained by optical methods reaches the target thickness range.

[0117] Comparative Example 2 uses only the endpoint determination of the friction force signal.

[0118] As another existing single-parameter endpoint detection method, the polishing endpoint is determined solely based on the frictional force or changes related to the frictional force.

[0119] Comparative Example 3 uses only the endpoint of the roughness signal for judgment.

[0120] To highlight the impact of different endpoint control methods on product parameters, the above embodiments and comparative examples can be compared according to three product parameters: top silicon thickness, thickness uniformity, and surface roughness. The control or verification relationship of each scheme is shown in the table below.

[0121]

[0122] As can be seen from the comparison in the table, the method of this disclosure can establish a correspondence between the top silicon thickness, thickness uniformity requirements, and surface roughness and the endpoint control under normal conditions. When one monitoring parameter is invalid, it can switch the judgment rule according to the remaining two valid parameters. Compared with the single-parameter endpoint detection method that only uses thickness signal or only uses friction force signal, this disclosure establishes a correspondence between the top silicon thickness range of 200nm to 230nm, the thickness uniformity requirement of less than 1.6%, and the surface roughness requirement of less than 0.5nm at least in the endpoint control logic.

[0123] The above embodiments and comparative examples illustrate the differences in endpoint determination methods by showing the correspondence between endpoint control processes and product parameters. The silicon-on-insulator wafer structure obtained through the aforementioned chemical mechanical polishing process is as follows: Figure 7 As shown.

[0124] like Figure 7 As shown, the silicon-on-insulator substrate 20 includes a top silicon layer 21, a buried oxide layer 22, and a support substrate 23 arranged sequentially. The top silicon layer 21 is the object to be processed in the aforementioned chemical mechanical polishing process.

[0125] After chemical mechanical polishing, the top silicon layer 21 has a corresponding product thickness. The thickness of the top silicon layer 21 is 200nm to 230nm; the thickness uniformity can be evaluated by statistically analyzing the thickness of the top silicon layer at multiple measurement locations on the wafer surface, and the thickness uniformity in the product state is less than 1.6%.

[0126] The polished surface of top silicon 21 also has a surface roughness. In the product state, the surface roughness of top silicon 21 is less than 0.5 nm. This product roughness is comparable to... Figure 5 The online roughness judgment thresholds have different meanings. The online judgment threshold is used for endpoint control during the polishing process, while the product roughness is used to represent the surface parameters of the top silicon 21 after processing.

[0127] Figure 7 The top silicon thickness and surface roughness are product parameters that can be measured by the silicon-on-insulator wafer itself. The endpoint control test in Example 1 recorded a material removal control accuracy of ±2nm, and the surface roughness after treatment met the product requirement of less than 0.5nm. For product specifications with a thickness of 200nm to 230nm and a thickness uniformity of less than 1.6%, the endpoint thickness range can be set according to the specific product target thickness, and the top silicon thickness and thickness uniformity can be measured and verified after treatment.

[0128] The buried oxide layer 22 is located below the top silicon layer 21. By incorporating the remaining thickness state into the endpoint determination and reducing the polishing pressure for fine finishing after endpoint confirmation, the extent to which the top silicon layer 21 can be further removed near the endpoint can be limited, thereby reducing the risk of over-polishing towards the buried oxide layer 22.

[0129] Meanwhile, incorporating online roughness status into the endpoint determination means that the final stopping condition is not only related to the thickness status of the top silicon 21, but also to the surface status after processing. This helps the product wafer meet the predetermined top silicon thickness range, thickness uniformity requirements, and surface roughness requirements, and provides a process basis for the surface flatness and low damage status required for subsequent device fabrication processes.

[0130] Where there is no conflict, the above-mentioned normal joint judgment method, comprehensive evaluation method, abnormal fault tolerance method and low-voltage fine maintenance control can be used in combination according to their technical dependencies. The specific combination of each technical feature shall be based on the ability to form a corresponding relationship between the input, status judgment, endpoint judgment, subsequent control and product parameter results.

[0131] The above description is used to illustrate the specific implementation of the technical solutions described in this disclosure. When adjusting the monitoring parameter acquisition method, judgment parameter value, or process parameter based on the same technical relationship, the scope of protection of this disclosure should still be understood in conjunction with the technical facts recorded in the specification and the technical relationship defined by the claims.

Claims

1. A method for manufacturing a silicon-on-insulator wafer, characterized in that, include: The top silicon layer of a silicon-on-insulator substrate is polished, and three monitoring parameters are acquired during the polishing process. These three monitoring parameters are used to characterize the remaining thickness of the top silicon layer, the change in frictional force during the rotation of the polishing disc, and the surface roughness of the top silicon layer. Determine the validity of each of the monitoring parameters; When at least two of the monitoring parameters are valid, the monitoring parameters participating in the endpoint determination are determined from the three monitoring parameters according to the validity of each monitoring parameter, and the endpoint determination rule corresponding to the monitoring parameters participating in the endpoint determination is determined. Based on the monitoring parameters involved in the endpoint determination and the endpoint determination rules, the endpoint of polishing is determined, and polishing is stopped after the endpoint is reached to obtain the silicon-on-insulator wafer.

2. The method for manufacturing a silicon-on-insulator wafer according to claim 1, characterized in that, The three monitoring parameters include a first monitoring parameter, a second monitoring parameter, and a third monitoring parameter; The first monitoring parameter includes the remaining thickness of the top silicon layer obtained based on laser interferometry; The second monitoring parameter includes the friction force change curve during the rotation of the grinding disc; The third monitoring parameter includes the roughness estimate of the top silicon surface obtained based on optical measurements during the polishing process.

3. The method for manufacturing a silicon-on-insulator wafer according to claim 2, characterized in that, When the first monitoring parameter, the second monitoring parameter, and the third monitoring parameter are all valid, the monitoring parameters involved in the endpoint determination include the first monitoring parameter, the second monitoring parameter, and the third monitoring parameter. The endpoint determination rules include: When the first monitoring parameter, the second monitoring parameter, and the third monitoring parameter all meet the corresponding endpoint conditions, the polishing endpoint is determined to have been reached.

4. The method for manufacturing a silicon-on-insulator wafer according to claim 3, characterized in that, The endpoint condition corresponding to the first monitoring parameter includes that the remaining thickness of the top silicon layer is within the target thickness range; The endpoint conditions corresponding to the second monitoring parameter include that the absolute value of the first derivative of the friction force change curve is lower than the friction force threshold and that the friction force change curve is plateaued. The endpoint condition corresponding to the third monitoring parameter includes that the roughness estimate is not greater than the roughness threshold.

5. The method for manufacturing a silicon-on-insulator wafer according to claim 2, characterized in that, Determining the validity of each of the monitoring parameters includes: If one of the monitoring parameters cannot be obtained, the monitoring parameter is determined to be invalid; When one of the monitoring parameters is available, the monitoring parameter is determined to be valid.

6. The method for manufacturing a silicon-on-insulator wafer according to claim 5, characterized in that, If one of the three monitoring parameters is invalid and the other two are valid, the invalid monitoring parameter is excluded from the endpoint determination, and the other two valid monitoring parameters are determined as the monitoring parameters participating in the endpoint determination. The corresponding endpoint determination rule is determined based on the combination of the two monitoring parameters involved in the endpoint determination. The endpoint determination rule requires that the two monitoring parameters each meet their respective endpoint conditions.

7. The method for manufacturing a silicon-on-insulator wafer according to claim 6, characterized in that, The endpoint determination rules corresponding to the monitoring parameters involved in endpoint determination include comprehensive determination rules; The comprehensive judgment rules include: The weight of each monitoring parameter and the comprehensive judgment threshold are determined based on the combination of the two monitoring parameters involved in the endpoint determination. The comprehensive evaluation value is determined based on the judgment results and corresponding weights of the two monitoring parameters involved in the endpoint determination. When the comprehensive evaluation value meets the comprehensive judgment threshold and the two monitoring parameters involved in the endpoint determination meet their respective endpoint conditions, the polishing endpoint is determined to have been reached.

8. The method for manufacturing a silicon-on-insulator wafer according to claim 1, characterized in that, After determining that the polishing endpoint has been reached, the polishing pressure is reduced to perform low-pressure finishing, and the polishing is stopped after the low-pressure finishing has reached a preset time.

9. A system for manufacturing silicon-on-insulator wafers, characterized in that, Includes monitoring modules and controllers; The monitoring module is used to acquire three monitoring parameters during the polishing process of the top silicon layer of the silicon-on-insulator substrate. The three monitoring parameters are used to characterize the remaining thickness of the top silicon layer, the change in friction during the rotation of the polishing disc, and the surface roughness of the top silicon layer. The controller is connected to the monitoring module via a signal and is configured to: Determine the validity of each of the monitoring parameters; When at least two of the monitoring parameters are valid, the monitoring parameters participating in the endpoint determination are determined from the three monitoring parameters according to the validity of each monitoring parameter, and the endpoint determination rule corresponding to the monitoring parameters participating in the endpoint determination is determined. The endpoint of polishing is determined based on the monitoring parameters involved in the endpoint determination and the endpoint determination rules.

10. The silicon-on-insulator wafer manufacturing system according to claim 9, characterized in that, The monitoring module includes: The first monitoring unit is used to obtain the remaining thickness of the top silicon layer based on laser interferometry; The second monitoring unit is used to acquire the friction force change curve during the rotation of the grinding disc; and The third monitoring unit is used to obtain an estimate of the roughness of the top silicon surface based on optical measurements during the polishing process.

11. The manufacturing system for silicon-on-insulator wafers according to claim 10, characterized in that, The first monitoring unit and the third monitoring unit are implemented by the same optical monitoring module, which determines the remaining thickness of the top silicon layer and the roughness estimate based on the optical measurement results.

12. A silicon-on-insulator wafer, characterized in that, The silicon-on-insulator wafer is prepared by the manufacturing method according to any one of claims 1 to 8; The thickness of the top silicon layer is 200 nm to 230 nm, and the thickness uniformity of the top silicon layer is less than 1.6%.

13. The silicon-on-insulator wafer according to claim 12, characterized in that, The surface roughness of the top silicon layer is less than 0.5 nm.

Images