A dust control method, system, device and storage medium
By constructing an algorithm model for sticking rod pressure control based on three-dimensional morphology parameters, the problems of pressure overshoot and dust morphology differences in sticking rod dust removal control were solved, achieving flexible contact and precise alignment, thereby improving dust removal efficiency and wafer yield.
Patent Information
- Application Number
- CN202611363753.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-04
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2046-09-04
AI Technical Summary
In the existing technology, the sticking rod dust removal control scheme has problems such as pressure overshoot causing wafer damage, failure to consider the differences in dust morphology leading to inconsistent contact pressure, and lack of precise horizontal positioning control.
By acquiring the attribute parameters of the adhesive rod and the three-dimensional geometric morphology of dust on the semiconductor wafer surface, an adhesive rod pressure control algorithm model is constructed to calculate the target displacement and achieve flexible contact and precise alignment.
Eliminate the risk of pressure overshoot, achieve adaptive and differentiated control of dust morphology, improve dust removal efficiency and accuracy, avoid secondary damage, and enhance yield management precision.
Smart Images

Figure CN122872005B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of semiconductor manufacturing technology, specifically to a dust removal control method, system, device, and storage medium. Background Technology
[0002] In semiconductor manufacturing processes, controlling particulate contamination on the wafer surface is one of the key factors affecting product yield. As semiconductor process nodes continue to evolve, even extremely small dust particles on the wafer surface can cause short circuits, open circuits, or photolithographic defects, thus requiring increasingly higher precision in wafer surface cleaning.
[0003] Currently, contact cleaning solutions using adhesive sticks are gaining increasing attention. These sticks work by using an adhesive material at the tip to contact dust particles on the wafer surface, removing the dust through adhesion. However, because wafer surfaces are extremely sensitive to contact pressure, excessive pressure can cause irreversible damage to the wafer's surface structure. Therefore, precise control of the contact between the adhesive stick and the wafer surface is crucial.
[0004] In existing technologies, the position control of the adhesive rod is typically achieved using the following two methods:
[0005] One approach is a closed-loop control scheme based on pressure feedback. This scheme uses a pressure sensor to monitor the contact pressure between the adhesive rod and the wafer surface in real time. When the detected pressure value reaches a preset threshold, the adhesive rod is controlled to stop moving forward. However, this scheme has inherent drawbacks: the pressure sensor has limited detection accuracy at small ranges, and the inherent signal acquisition, transmission, and processing delays in the control system mean that the adhesive rod often exceeds the predetermined position when it receives the stop command, resulting in an "overshoot" phenomenon. The actual contact pressure far exceeds the preset threshold, thus damaging the precision product.
[0006] Secondly, a feedforward control scheme based on a pressure-displacement model is proposed. This scheme establishes a pressure-displacement relationship function by pre-calibrating the pressure applied to the test carrier by the test stick at different displacements. During the cleaning process, the required displacement is calculated based on the preset target contact pressure value, and then the stick is controlled to move to the position corresponding to the target displacement. This scheme eliminates the overshoot risk caused by feedback delay in the displacement control logic. However, this scheme only considers the influence of the stick's own property parameters on the pressure-displacement relationship, without considering the influence of the morphological differences of the dust particles on the contact process. In actual cleaning scenarios, the mechanical behavior of dust particles of different heights and sizes when in contact with the tip of the stick varies significantly. If all dust particles are controlled with a uniform initial zero point position and the same displacement, when the dust size is large, the stick may contact the dust too early, resulting in excessive pressure; when the dust size is small, the stick may not be able to fully contact the dust, resulting in incomplete cleaning.
[0007] Furthermore, existing technologies lack precise horizontal positioning control when aligning the stick with the dust location, and usually rely on two-dimensional images for coarse positioning, which cannot accurately align with the geometric center of the dust.
[0008] Therefore, there is an urgent need for a dust control method that can adaptively adjust the displacement of the sticking rod according to the three-dimensional geometric morphology of dust to achieve flexible contact. Summary of the Invention
[0009] This application provides a dust removal control method, system, device, and storage medium, aiming to solve the technical problems of existing dust removal control schemes based on pressure feedback causing wafer damage due to signal delay leading to overshoot of the sticking rod; existing feedforward control schemes based on pressure-displacement models failing to consider the differences in dust morphology resulting in inconsistent contact pressures for dust of different sizes; existing technologies lacking precise horizontal positioning control based on dust morphology when the sticking rod is aligned with the dust position; and existing technologies failing to distinguish between dust and defects, potentially causing secondary damage by blindly performing dust removal actions on irreversible defect types.
[0010] To achieve the above objectives, this application provides the following technical solution:
[0011] In a first aspect, this application provides a dust control method applied to an adhesive stick cleaning device, comprising the following steps:
[0012] S1: Obtain the sticky rod property parameters and the three-dimensional geometric morphology features of the dust on the semiconductor wafer surface. The three-dimensional geometric morphology features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane.
[0013] S2: Construct a sticky rod pressure control algorithm model based on the sticky rod property parameters and the three-dimensional geometric features;
[0014] S3: During the cleaning process, obtain the real-time parameters of the current sticky stick and the three-dimensional geometric features of the current dust.
[0015] S4: Input the real-time parameters and the current three-dimensional geometric features of the dust into the sticky rod pressure control algorithm model to calculate the target displacement of the sticky rod in the horizontal and vertical directions;
[0016] S5: Control the sticking rod to move to the target position according to the target displacement, so that the tip of the sticking rod comes into contact with the dust to achieve dust removal.
[0017] Specifically, in S1, the height extreme value H and the spread size D are obtained through the following steps:
[0018] S11. Define the reference plane of the abnormal area where the dust is located as O1, define the highest point of the abnormal area as O2, and define the height extreme value H as the vertical distance of O2 relative to O1.
[0019] S12. Establish a height section Z1 at a preset height offset △H above the reference plane, extract the outline of the abnormal area, and calculate the maximum span between any two points on the outline as the display size D.
[0020] Specifically, the height offset ΔH is 30% to 70% of the height extreme value H.
[0021] Specifically, in step S2, constructing the sticky rod pressure control algorithm model includes:
[0022] S21: Provide a test stick and a test carrier, record the pressure value applied by the test stick to the test carrier at different displacement values, and generate a pressure-displacement relationship function;
[0023] S22: Acquire three-dimensional topographic data of multiple semiconductor wafer surfaces, extract the height extreme value H and the distribution size D of each dust, establish the mapping relationship between the dust topographic parameters and the initial zero point displacement, and generate a topographic-zero point mapping function.
[0024] S23: Provide the test sticky rod with different attribute parameters and the dust with different three-dimensional geometric features, repeat S21 and S22, and statistically analyze the attribute parameters, the three-dimensional geometric features, the pressure-displacement relationship function and the morphology-zero point mapping function to construct the sticky rod pressure control algorithm model.
[0025] Specifically, in step S4, the target displacement of the adhesive rod in the horizontal and vertical directions is calculated, including:
[0026] S41: Based on the current spread size D of the dust, calculate the horizontal displacement required for the front end of the adhesive stick to align with the geometric center of the dust in the horizontal direction;
[0027] S42: Determine the initial zero point position of the sticky rod based on the height extreme value H of the current dust using the topography-zero point mapping function;
[0028] S43: Based on the initial zero point position, combined with the real-time parameters of the sticky rod and the pressure-displacement relationship function, calculate the target displacement of the sticky rod in the vertical direction.
[0029] Specifically, in step S43, the pressure-displacement relationship function is preset with a target contact pressure value, which is determined according to the material type of the semiconductor wafer. By inputting the target contact pressure value and the real-time parameters of the sticky rod into the pressure-displacement relationship function, the vertical displacement of the sticky rod from the initial zero point position is calculated. The sum of the initial zero point position and the vertical displacement constitutes the target displacement in the vertical direction.
[0030] Specifically, in the morphology-zero point mapping function, dust with different height extrema H corresponds to different initial zero point positions; the initial zero point position increases as the height extrema H increases; if the height extrema H is positive, it is determined to be a removable attachment type, and S3 to S5 are executed; if the height extrema H is negative, it is determined to be an irreversible defect type, the position is skipped and the defect information is recorded.
[0031] Specifically, the adhesive rod attribute parameters include the external dimensions and / or expiration date information of the adhesive rod; the three-dimensional geometric features are acquired by a three-dimensional detection module, which is any one of a white light interferometer sensor, a spectral confocal sensor, a laser triangulation sensor, a structured light projection sensor, or a time-of-flight sensor.
[0032] Secondly, this application provides a dust removal control system, including a parameter acquisition module, a model construction module, a real-time detection module, a displacement calculation module, and an execution control module. The parameter acquisition module is used to acquire the attribute parameters of the sticky rod and the three-dimensional geometric features of the dust on the semiconductor wafer surface. The three-dimensional geometric features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane. The model construction module is used to construct a sticky rod pressure control algorithm model based on the sticky rod attribute parameters and the three-dimensional geometric features. The real-time detection module is used to acquire the real-time parameters of the current sticky rod and the three-dimensional geometric features of the current dust during the cleaning process. The displacement calculation module is used to input the real-time parameters and the three-dimensional geometric features of the current dust into the sticky rod pressure control algorithm model to calculate the target displacement of the current sticky rod in the horizontal and vertical directions. The execution control module is used to control the current sticky rod to move to the target position according to the target displacement, so that the tip of the current sticky rod contacts the current dust to achieve dust removal.
[0033] Specifically, the displacement calculation module includes a horizontal displacement calculation unit, a zero-point determination unit, and a vertical displacement calculation unit. The horizontal displacement calculation unit calculates the horizontal displacement required for the tip of the adhesive rod to align with the geometric center of the dust in the horizontal direction, based on the current dust distribution size D. The zero-point determination unit determines the initial zero-point position of the adhesive rod based on the current dust height extreme value H using the topography-zero-point mapping function. The vertical displacement calculation unit calculates the target vertical displacement of the adhesive rod based on the initial zero-point position, combined with the real-time parameters of the adhesive rod and the pressure-displacement relationship function.
[0034] Thirdly, this application provides a dust removal device, including a machine base, a motion component, a support platform, a three-dimensional detection module, an adhesive rod cleaning component, and a dust removal control system. The motion component is mounted on the device frame assembly; the support platform is mounted on the motion component and is used to support a semiconductor wafer. The three-dimensional detection module is located above the motion component and is used to acquire the three-dimensional geometric morphology features of dust on the surface of the semiconductor wafer. The adhesive rod cleaning component is located on the motion component and is used to clean the dust on the surface of the semiconductor wafer. The dust removal control system is electrically connected to the motion component, the three-dimensional detection module, and the adhesive rod cleaning component. The dust removal control system includes a parameter acquisition module, a model construction module, a real-time detection module, a displacement calculation module, and an execution control module. The parameter acquisition module is used to acquire adhesive rod attribute parameters and the three-dimensional geometric morphology features of dust on the surface of the semiconductor wafer. The three-dimensional geometric morphology features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane. The model construction module is used to construct an adhesive rod pressure control algorithm model based on the adhesive rod attribute parameters and the three-dimensional geometric morphology features. The real-time detection module is used to acquire real-time parameters of the current sticky rod and the three-dimensional geometric features of the current dust during the cleaning process. The displacement calculation module is used to input the real-time parameters and the three-dimensional geometric features of the current dust into the sticky rod pressure control algorithm model to calculate the target displacement of the current sticky rod in the horizontal and vertical directions. The execution control module is used to control the current sticky rod to move to the target position according to the target displacement, so that the tip of the current sticky rod contacts the current dust to achieve dust removal.
[0035] Fourthly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a dust removal control method. The dust removal control method includes the following steps:
[0036] S1: Obtain the sticky rod property parameters and the three-dimensional geometric morphology features of the dust on the semiconductor wafer surface. The three-dimensional geometric morphology features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane.
[0037] S2: Construct a sticky rod pressure control algorithm model based on the sticky rod property parameters and the three-dimensional geometric features;
[0038] S3: During the cleaning process, obtain the real-time parameters of the current sticky stick and the three-dimensional geometric features of the current dust.
[0039] S4: Input the real-time parameters and the current three-dimensional geometric features of the dust into the sticky rod pressure control algorithm model to calculate the target displacement of the sticky rod in the horizontal and vertical directions;
[0040] S5: Control the sticking rod to move to the target position according to the target displacement, so that the tip of the sticking rod comes into contact with the dust to achieve dust removal.
[0041] Compared with the prior art, the dust removal control method, system, equipment and storage medium provided in this application have the following beneficial effects:
[0042] Eliminating the risk of pressure overshoot and protecting wafer safety. This application fundamentally eliminates the overshoot problem caused by pressure sensor accuracy limitations and signal transmission delays by transforming stick displacement control from "real-time control based on pressure feedback" to "feedforward displacement control based on three-dimensional topography parameters." The stick moves directly to the target position according to the pre-calculated target displacement. Even with control signal delays, the stick will not exceed the predetermined position, ensuring that the wafer surface contact pressure is always within a safe range.
[0043] This application achieves adaptive and differentiated dust morphology control. By acquiring the height extreme value H and distribution size D of the dust, a mapping relationship (morphology-zero point mapping function) is established between the dust morphology parameters and the initial zero-point displacement, so that dust of different heights and sizes corresponds to different initial zero-point positions. For dust with greater height, the initial zero-point position is set further away from the wafer surface to avoid premature contact of the adhesive rod leading to excessive pressure; for dust with less height, the initial zero-point position is set closer to the wafer surface to avoid insufficient contact of the adhesive rod leading to incomplete cleaning. Thus, differentiated and flexible dust removal control with a "one dust, one policy" approach is achieved.
[0044] Precise horizontal alignment improves dust removal efficiency and cleaning effect. This application calculates the target displacement of the stick tip in the horizontal direction based on the dust distribution size D, ensuring that the center of the stick tip is precisely aligned with the geometric center of the dust. This maximizes the utilization of the effective adhesion area of the stick tip, completing dust removal in a single contact. This avoids repeated dust removal caused by eccentric contact, improving dust removal efficiency. At the same time, it reduces unnecessary contact between the stick and the wafer surface, lowering the risk of secondary contamination and mechanical damage.
[0045] This application achieves precise separation of dust and defects, avoiding ineffective dust removal and secondary damage. It determines the anomaly type based on the sign of the extreme value H. When H is positive, it is determined to be removable dust, and dust removal is performed; when H is negative, it is determined to be an irreversible defect, the location is skipped, and defect information is recorded. This mechanism avoids secondary damage caused by blindly performing dust removal when defects are misidentified as dust, and also avoids wafer erroneous scrapping caused by misidentifying dust as defects, thus improving yield management accuracy.
[0046] The model possesses self-evolution capabilities. This application constructs a composite algorithm model by statistically analyzing the pressure-displacement relationship function and morphology-zero point mapping function corresponding to different adhesive rod attribute parameters (shape and shelf life) and different dust three-dimensional morphological features (H, D). As data from different batches of adhesive rods and dust with different morphologies accumulate, the model can be continuously iterated and updated, and the control accuracy continues to improve with the increase of usage time. Attached Figure Description
[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein:
[0048] Figure 1 This is a flowchart of the dust removal method in the embodiments of this application.
[0049] Figure 2 for Figure 1 The flowchart of S1 is shown below.
[0050] Figure 3 This is a first-angle schematic diagram of the three-dimensional geometric morphology of dust on the surface of a semiconductor wafer.
[0051] Figure 4 This is a second-angle schematic diagram of the three-dimensional geometric morphology of dust on the surface of a semiconductor wafer.
[0052] Figure 5 for Figure 1 The flowchart of S2 is shown below.
[0053] Figure 6 for Figure 1 The flowchart of S4 is shown.
[0054] Figure 7 This is a schematic diagram of the three-dimensional geometric morphology of dust on the surface of a semiconductor wafer after dust removal, taken from the first angle.
[0055] Figure 8 This is a structural block diagram of the dust removal control system in the embodiments of this application.
[0056] Figure 9 This is a structural block diagram of another embodiment of the dust removal control system in this application.
[0057] Figure 10 This is a three-dimensional structural diagram of the dust removal equipment in the embodiments of this application. Detailed Implementation
[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0059] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.
[0060] Furthermore, the use of terms such as "first," "second," etc., in this application is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least three, such as three, four, etc., unless otherwise explicitly specified.
[0061] In this application, unless otherwise expressly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.
[0062] Furthermore, the technical solutions of the various embodiments of this application can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this application.
[0063] Please refer to Figure 1 This application provides a dust control method applied to sticky rod cleaning equipment, comprising the following steps:
[0064] Step S1: Obtain the adhesive rod property parameters and the three-dimensional geometric morphology features of the dust on the semiconductor wafer surface. The three-dimensional geometric morphology features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane.
[0065] The adhesive rod refers to a rod-shaped structure with an adhesive substance at one end. This substance contacts dust on the wafer surface, using adhesion to remove dust particles. This application fundamentally eliminates the overshoot problem caused by pressure sensor accuracy limitations and signal transmission delays by transforming the adhesive rod displacement control from "real-time control based on pressure feedback" to "feedforward displacement control based on three-dimensional topography parameters." The adhesive rod moves directly to the target position according to the pre-calculated target displacement. Even with a delay in the control signal, the adhesive rod will not exceed the predetermined position, ensuring that the contact pressure on the wafer surface remains within a safe range.
[0066] Please refer to the reference. Figure 2 , Figure 3 and Figure 4 In step S1, the height extreme value H and the spread size D are obtained through the following steps:
[0067] S11. Define the reference plane of the abnormal area where the dust is located as O1, define the highest point of the abnormal area as O2, and define the height extreme value H as the vertical distance of O2 relative to O1.
[0068] S12. Establish a height section Z1 at a preset height offset △H above the reference plane, extract the outline of the abnormal area, and calculate the maximum span between any two points on the outline as the display size D.
[0069] like Figure 3 As shown, this illustrates the principle for extracting the height extremum H and the spread dimension D. The abnormal region (dust) is convex relative to the wafer surface reference plane O1, and the vertical distance of its highest point O2 relative to O1 is the height extremum H (H>0). A height section Z1 is established above the reference plane O1 at a preset height offset ΔH. This section intersects with the dust profile to obtain a closed section line. The maximum span between any two points on this section line (i.e., the maximum diameter of the section profile) is the spread dimension D.
[0070] In this application, the height offset ΔH is 30% to 70% of the height extreme value H. Preferably, ΔH is 50% of H. This ratio range ensures that the cross-section is located in the upper-middle region of the dust body, while avoiding excessively small spread size measurements due to an excessively high cross-section (close to the top of the dust) or excessively large spread size measurements due to an excessively low cross-section (close to the bottom of the dust), thereby obtaining a stable and repeatable spread size D. To verify the technical effect of ΔH being 30% to 70% of H, this application conducted a comparative experiment using a 12-inch unpatterned silicon wafer, testing standard PSL polystyrene microspheres with a particle size of 0.5μm to 10μm. The experimental data are as follows:
[0071] Test metrics △H=20%H △H=30%H △H=50%H △H=70%H △H=80%H Repeatability of fabric dimension D measurement (relative standard deviation) 8.7% 3.2% 1.5% 2.8% 7.9% Dust recognition accuracy 91.2% 97.4% 98.7% 97.1% 90.8% Deviation between diameter and SEM measurement ±0.82μm ±0.31μm ±0.15μm ±0.28μm ±0.73μm Marginal false alarm rate 6.3% 2.1% 0.3% 1.8% 5.9%
[0072] As shown in the table above, when ΔH is within the range of 30% to 70% of H, the repeatability of the spread dimension D measurement is significantly better than outside this range, with a dust identification accuracy of over 97%, a deviation from SEM measurements controlled within ±0.31 μm, and an edge false alarm rate of less than 2.1%. The optimal performance is achieved when ΔH = 50%H. When ΔH is too small (e.g., 20%H), the cross-sectional position is too low, causing the bottom of the dust layer to be included in the measurement, amplifying individual differences between different dust particles and reducing measurement repeatability. When ΔH is too large (e.g., 80%H), the cross-sectional position is too high, approaching the top of the dust layer, and the differences in the top morphology of different dust particles lead to increased fluctuations in the measured values. This application limits ΔH to the range of 30% to 70%H, ensuring the stability and repeatability of the spread dimension D measurement and providing reliable input parameters for subsequent model construction.
[0073] In another embodiment, if the abnormal area where the dust is located is concave relative to the reference surface (i.e., H<0), a depth section Z1' is established at a preset depth offset △H' below the reference surface, the outline of the concave area is captured, and the maximum span between any two points on the outline is calculated as the spreading dimension D. For the concave abnormal shape, △H' is also preferably 30% to 70% of |H|.
[0074] The adhesive rod attribute parameters include the external dimensions of the adhesive rod (such as the diameter of the tip and the shape of the end face) and / or the expiration date information (such as the production date and storage time). Since the adhesive material at the tip of the adhesive rod may solidify or age to some extent during storage, adhesive rods with different expiration dates may have different mechanical properties. Therefore, incorporating the expiration date information into the attribute parameters helps to improve the accuracy of subsequent models.
[0075] The three-dimensional geometric features are acquired by a three-dimensional detection module, which can be any one of a white light interferometer, a spectral confocal sensor, a laser triangulation sensor, a structured light projection sensor, or a time-of-flight sensor. White light interferometers offer the highest vertical resolution (up to 0.1 nm), making them suitable for detecting nanoscale dust particles; spectral confocal sensors are suitable for detecting transparent or highly reflective wafer surfaces; laser triangulation sensors and structured light projection sensors offer fast scanning speeds, suitable for large-area rapid detection; time-of-flight sensors offer the fastest speed but relatively lower accuracy, suitable for coarse inspection scenarios where high precision is not required. Those skilled in the art can select the appropriate 3D sensor type based on the actual application scenario.
[0076] Please refer to the reference. Figure 1 and Figure 5 Step S2: Based on the adhesive rod attribute parameters and the three-dimensional geometric features, construct an adhesive rod pressure control algorithm model. Specifically, S21 to S23 are included in S22.
[0077] S21: Provide a test stick and a test carrier, record the pressure value applied by the test stick to the test carrier at different displacement values, and generate a pressure-displacement relationship function.
[0078] Specifically, the initial zero point is first determined: the position where the test stick is about to contact the surface of the test carrier but the contact pressure is zero is defined as the initial zero point. Then, the test stick is driven to gradually press down from the initial zero point in small steps (e.g., 1 μm to 5 μm), and the displacement d corresponding to each step position is recorded. i The pressure value F fed back by the pressure sensor i According to multiple groups (d) i F iThe data is used to generate a pressure-displacement relationship function F=f(d, k) through curve fitting, where k is a coefficient related to the properties of the test sticky rod.
[0079] In another embodiment, the pressure-displacement relationship function can also be generated using piecewise linear interpolation or cubic spline interpolation methods to more accurately describe the nonlinear mechanical behavior of the sticky rod during the initial contact stage and the deep compression stage.
[0080] S22: Acquire three-dimensional topographic data of multiple semiconductor wafer surfaces, extract the height extreme value H and distribution size D of each dust particle, establish the mapping relationship between dust topographic parameters and initial zero-point displacement, and generate a topographic-zero-point mapping function.
[0081] Specifically, for dust particles with different height extremes H and different distribution sizes D, the "optimal initial zero-point position" was determined experimentally. This is the position where the adhesive rod is about to contact the dust surface but the contact pressure is zero. The dust morphology parameters (H, D) were correlated with the corresponding optimal initial zero-point position Z0, and a morphology-zero-point mapping function Z0 = g(H, D) was generated by fitting a function using multiple regression analysis or machine learning algorithms.
[0082] In this embodiment, the topography-zero mapping function can take the following form:
[0083] Z0 = α·H + β·D + γ
[0084] Where α, β, and γ are regression coefficients obtained by fitting experimental data. This function reflects the differences in the initial zero-point position corresponding to dust with different morphologies. The larger H or D is, the larger the Z0 value, meaning the farther the initial zero-point position is from the wafer surface.
[0085] In another embodiment, the topography-zero mapping function can also be in a nonlinear form:
[0086] Z0 = α·H^p + β·D^q + γ, where p and q are power exponents used to describe the nonlinear effects of H and D on the initial zero point position.
[0087] In another embodiment, the topography-zero mapping function may also use only the height extreme value H as the independent variable, that is:
[0088] Z0 = g(H)
[0089] At this point, the influence of the fabric size D is compensated for by adjusting the subsequent target contact pressure value.
[0090] S23: Provide test sticky rods with different attribute parameters and dust with different three-dimensional geometric features. Repeat S21 and S22, and statistically analyze the attribute parameters, the three-dimensional geometric features, the pressure-displacement relationship function, and the morphology-zero point mapping function to construct the sticky rod pressure control algorithm model.
[0091] Specifically, by providing test adhesive sticks with different dimensions (e.g., different diameters, different end-face shapes) and different shelf lives (e.g., 3 minutes, 5 minutes, or 10 minutes), as well as dust samples with different H and D values, the calibration and mapping processes S21 and S22 are repeatedly performed to obtain a complete dataset covering various adhesive stick properties and various dust morphologies. Based on this dataset, a composite algorithm model is constructed using multivariate regression analysis, neural network fitting, or other machine learning algorithms.
[0092] In this embodiment, the sticky rod pressure control algorithm model can be expressed as:
[0093] ΔZ = Φ(k, H, D, F target )
[0094] Where ΔZ is the total displacement of the sticky rod from its initial position to its target position, k is the sticky rod attribute parameter vector, H and D are the dust morphology parameters, and F... target Φ represents the target contact pressure value. Φ is a composite mapping function, internally solved by simultaneously solving the pressure-displacement relationship function F=f(d, k) and the topography-zero point mapping function Z0=g(H, D), outputting the target contact pressure value F. target The required total displacement ΔZ.
[0095] In another embodiment, the sticky rod pressure control algorithm model can also be implemented using a neural network model. The sticky rod attribute parameters and dust morphology parameters are used as input layers, and the target displacement is used as the output layer. The network weight parameters are obtained through training with a large amount of experimental data. This model can output the corresponding pressure-displacement relationship function and morphology-zero point mapping function based on the input sticky rod attribute parameters and the three-dimensional geometric morphology features of the dust.
[0096] In another embodiment, the construction of the sticky rod pressure control algorithm model can be completed on the sticky rod cleaning device itself, or it can be completed on other computing devices and then the model can be imported into the storage module of the sticky rod cleaning device. The former can better adapt to the mechanical characteristics of the device itself (such as the response characteristics of the drive motor, the friction characteristics of the transmission mechanism, etc.), while the latter can use higher performance computing resources to perform more complex model training.
[0097] Step S3: During the cleaning process, obtain the real-time parameters of the current sticky stick and the three-dimensional geometric features of the current dust.
[0098] The real-time parameters include the attribute parameters of the current adhesive stick (shape and size information, expiration date information) and the relative position parameters of the current adhesive stick to the surface of the wafer to be cleaned (such as the coordinates of the tip of the current adhesive stick in the device coordinate system). The three-dimensional geometric features of the current dust are acquired in real time by the three-dimensional detection module, including the height extreme value H and the distribution size D of the current dust.
[0099] Please refer to the reference. Figure 1 and Figure 6 Step S4: Input the real-time parameters and the current three-dimensional geometric features of the dust into the sticky rod pressure control algorithm model to calculate the target displacement of the sticky rod in the horizontal and vertical directions. Figure 6 As shown, in step S4, the target displacement of the adhesive rod in the horizontal and vertical directions is calculated, specifically including steps S41 to S43:
[0100] S41: Based on the current dust distribution size D, calculate the horizontal displacement required for the front end of the adhesive stick to align with the geometric center of the dust in the horizontal direction.
[0101] Specifically, based on the current dust distribution size D and the position coordinates (X, Y, F, Z) of the dust geometric center in the current coordinate system obtained by the 3D detection module... c Y c ), calculate the movement of the current sticky tip center from its current position (X0, Y0) to the dust geometry center (X). c Y c The required horizontal displacement ΔX = X c - X0, ΔY = Y c - Y0. This horizontal displacement ensures that the tip of the adhesive stick is precisely aligned with the geometric center of the dust in the horizontal direction, allowing the adhesive material at the tip of the stick to contact the dust over the largest possible area, thus improving the success rate of dust removal in a single pass. In another embodiment, when the dust distribution area D is large, a multi-point contact strategy can be adopted—that is, the tip of the adhesive stick sequentially contacts different areas of the dust multiple times to cover the entire dust area, avoiding incomplete cleaning due to insufficient contact area in a single pass.
[0102] S42: Based on the current height extreme value H of the dust, determine the initial zero-point position of the adhesive rod using the topography-zero-point mapping function. Specifically, input the current height extreme value H of the dust (and its spreading size D) into the topography-zero-point mapping function Z0 = g(H, D) established in step S22 to calculate the initial zero-point position Z0 corresponding to the current dust. This initial zero-point position is the position where the adhesive rod is about to contact the surface of the current dust but the contact pressure is zero. Figure 6 and Figure 7As shown, for dust particles with a smaller height (H1), their initial zero-point position Z01 is closer to the wafer surface; for dust particles with a larger height (H2, H2>H1), their initial zero-point position Z02 is farther from the wafer surface. That is, the initial zero-point position increases as the height increases towards the extreme value H. Since different heights of dust correspond to different initial zero-point positions, this step achieves differentiated control of "one dust particle, one zero point".
[0103] S43: Based on the initial zero-point position, combined with the real-time parameters of the adhesive rod and the pressure-displacement relationship function, the target displacement of the adhesive rod in the vertical direction is calculated. In some embodiments of this application, the pressure-displacement relationship function is preset with a target contact pressure value F. target The target contact pressure value is determined based on the material type of the semiconductor wafer. For example, for bare silicon wafers, the target contact pressure value can be set to 5g to 10g; for wafers that have undergone patterning, the target contact pressure value can be set to 1g to 3g; for wafers with fragile structures (such as wafers with MEMS structures or low-k dielectric layers), the target contact pressure value can be set to 0.5g to 1g. This target contact pressure value F... target The real-time parameters of the adhesive rod (such as its dimensions and expiration date) are input into the pressure-displacement relationship function F = f(d, k), and the vertical displacement ΔZ required for the adhesive rod to continue pressing down from the initial zero point position is calculated. press Specifically, by solving equation F target = f(ΔZ press , k) to obtain ΔZ press The initial zero point position Z0 and the vertical displacement ΔZ of the continued downward pressure. press The sum of these constitutes the target displacement Z in the vertical direction. target = Z0 + ΔZ press .
[0104] In another embodiment, the target contact pressure value can also be dynamically determined based on the three-dimensional geometric morphology of the dust. For dust with a larger height or larger spread size, the target contact pressure value can be appropriately increased; for dust with a smaller height or smaller spread size, a lower target contact pressure value is used, thereby achieving more refined differentiated control. For example, the target contact pressure value can be set to F. target = F base + λ·H + μ·D, where F base The base pressure value is λ, and μ are coefficients adjusted according to the dust morphology.
[0105] Step S5: Control the sticky rod to move to the target position according to the target displacement, so that the tip of the sticky rod comes into contact with the dust to achieve dust removal.
[0106] In step S5, the current sticky rod is controlled to move according to the horizontal displacement (ΔX, ΔY) calculated in step S41, so that the center of the front end of the sticky rod is aligned with the geometric center of the dust in the horizontal direction; then the current sticky rod is controlled to move according to the vertical target displacement Z calculated in step S43. target The stick is moved to the target position, bringing its tip into contact with the dust. Once at the target position, the adhesive material at the tip of the stick comes into contact with the dust, using adhesion to remove the dust from the wafer surface, completing a single dust removal operation. Since the target displacement is calculated using a feedforward model based on the current three-dimensional morphological characteristics (H, D) of the dust and the real-time attribute parameters of the stick, rather than relying on the real-time feedback signal from a pressure sensor, the risk of overshoot due to signal delay is fundamentally eliminated.
[0107] In another embodiment, S5 further includes a safety monitoring step: real-time monitoring of the actual contact pressure value between the adhesive rod and the wafer surface; when the actual contact pressure value exceeds a preset safety threshold, a protection mechanism is triggered, the protection mechanism including issuing an alarm signal and / or controlling the adhesive rod to retract. This safety monitoring step serves as a last line of defense to address unexpected situations such as unknown hardware failures, abnormal product placement, or model deviations.
[0108] Please refer to the following: Figure 7 In another embodiment, after dust removal is completed, a re-inspection and verification step is also included:
[0109] Obtain the three-dimensional topographic data of the location after dust removal, and extract the second height extreme value H' and the second distribution size D' of the location;
[0110] If the second height extreme value H' recovers to near the reference plane (i.e., H'≈0), it is determined that the dust has been removed, and the corresponding marker information is updated to a clean state; if the second height extreme value H' still deviates significantly from the reference plane (i.e., H' is still much greater than 0), it is determined that the dust has not been removed, and the dust removal action is triggered again or the position is recorded as an anomaly.
[0111] In some embodiments of this application, after step S1, a step is further included to determine whether the dust can be removed based on the three-dimensional geometric features. If the height extreme value H is positive, it is determined to be a type of removable deposit, and the dust removal steps S3 to S5 are continued; if the height extreme value H is negative, it is determined to be an irreversible defect type, the position is skipped, and the defect information is recorded. This diversion mechanism avoids secondary damage caused by misjudging defects as dust and blindly performing dust removal actions.
[0112] Please refer to the reference. Figure 8 and Figure 10 This application also provides a dust removal control system, including a parameter acquisition module, a model building module, a real-time detection module, a displacement calculation module, and an execution control module. The parameter acquisition module is used to acquire the attribute parameters of the sticky rod and the three-dimensional geometric features of the dust on the semiconductor wafer surface. The three-dimensional geometric features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane. The model building module is used to construct a sticky rod pressure control algorithm model based on the sticky rod attribute parameters and the three-dimensional geometric features. The real-time detection module is used to acquire the real-time parameters of the current sticky rod and the three-dimensional geometric features of the current dust during the cleaning process. The displacement calculation module is used to input the real-time parameters and the three-dimensional geometric features of the current dust into the sticky rod pressure control algorithm model to calculate the target displacement of the current sticky rod in the horizontal and vertical directions. The execution control module is used to control the current sticky rod to move to the target position according to the target displacement, so that the tip of the current sticky rod contacts the current dust to achieve dust removal.
[0113] Please refer to the reference. Figure 9 and Figure 10 In some embodiments of this application, the displacement calculation module includes a horizontal displacement calculation unit, a zero-point determination unit, and a vertical displacement calculation unit. The horizontal displacement calculation unit calculates the horizontal displacement required for the tip of the adhesive rod to align with the geometric center of the dust in the horizontal direction, based on the current dust distribution size D. The zero-point determination unit determines the initial zero-point position of the adhesive rod based on the current dust height extreme value H using the topography-zero-point mapping function. The vertical displacement calculation unit calculates the target vertical displacement of the adhesive rod based on the initial zero-point position, combined with the real-time parameters of the adhesive rod and the pressure-displacement relationship function.
[0114] Please refer to Figure 10 This application also provides a dust removal device, including a machine base 1, a motion component 2, a support platform 3, a three-dimensional detection module 5, an adhesive stick cleaning component 4, and a control component 6. The motion component 2 is mounted on the machine base 1. The support platform 3 is mounted on the motion component 2 and is used to support a semiconductor wafer 10. The three-dimensional detection module 5 is located above the motion component 2 and is used to acquire the three-dimensional geometric morphology features of dust on the surface of the semiconductor wafer 10. The adhesive stick cleaning component 4 is mounted on the motion component 2 and is used to clean the dust on the surface of the semiconductor wafer 10. The control component 6 is electrically connected to the motion component 2, the three-dimensional detection module 5, and the adhesive stick cleaning component 4, respectively, and is used to execute the aforementioned dust removal control method.
[0115] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described dust removal control method.
[0116] The above description is merely an embodiment of this application. It should be noted that those skilled in the art can make improvements without departing from the inventive concept of this application, but these improvements all fall within the protection scope of this application.
Claims
1. A dust control method applied to an adhesive stick cleaning device, characterized in that, Includes the following steps: S1: Obtain the attribute parameters of the stick and the three-dimensional geometric features of the dust on the semiconductor wafer surface. The attribute parameters include the external dimensions of the stick and the three-dimensional geometric features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane. S2: Construct a sticky rod pressure control algorithm model based on the sticky rod property parameters and the three-dimensional geometric features; S3: During the cleaning process, obtain the real-time parameters of the current stick and the three-dimensional geometric features of the current dust. The real-time parameters include the attribute parameters of the current stick and the relative position parameters of the current stick to the surface of the wafer to be cleaned. S4: Input the real-time parameters and the current three-dimensional geometric features of the dust into the sticky rod pressure control algorithm model to calculate the target displacement of the sticky rod in the horizontal and vertical directions; S5: Control the sticking rod to move to the target position according to the target displacement, so that the tip of the sticking rod comes into contact with the dust to achieve dust removal; Specifically, in step S2, constructing the sticky rod pressure control algorithm model includes: S21: Provide a test stick and a test carrier, record the pressure value applied by the test stick to the test carrier at different displacement values, and generate a pressure-displacement relationship function; S22: Acquire three-dimensional topographic data of multiple semiconductor wafer surfaces, extract the height extreme value H and the distribution size D of each dust, establish the mapping relationship between the dust topographic parameters and the initial zero point displacement, and generate a topographic-zero point mapping function. S23: Provide the test sticky rod with different attribute parameters and the dust with different three-dimensional geometric features, repeat S21 and S22, and statistically analyze the attribute parameters, the three-dimensional geometric features, the pressure-displacement relationship function and the shape-zero point mapping function to construct the sticky rod pressure control algorithm model; In step S4, the target displacement of the adhesive rod in the horizontal and vertical directions is calculated, specifically including: S41: Based on the current spread size D of the dust, calculate the horizontal displacement required for the front end of the adhesive stick to align with the geometric center of the dust in the horizontal direction; S42: Determine the initial zero point position of the sticky rod based on the height extreme value H of the current dust using the topography-zero point mapping function; S43: Based on the initial zero point position, combined with the real-time parameters of the sticky rod and the pressure-displacement relationship function, calculate the target displacement of the sticky rod in the vertical direction.
2. The dust control method according to claim 1, characterized in that, In step S1, the height extreme value H and the spread size D are obtained through the following steps: S11. Define the reference plane of the abnormal area where the dust is located as O1, define the highest point of the abnormal area as O2, and define the height extreme value H as the vertical distance of O2 relative to O1. S12. Establish a height section Z1 at a preset height offset △H above the reference plane, extract the outline of the abnormal area, and calculate the maximum span between any two points on the outline as the display size D.
3. The dust control method according to claim 2, characterized in that, The height offset ΔH is 30% to 70% of the height extreme value H.
4. The dust control method according to claim 1, characterized in that, In step S43, the pressure-displacement relationship function has a preset target contact pressure value, which is determined according to the material type of the semiconductor wafer. By inputting the target contact pressure value and the real-time parameters of the sticky rod into the pressure-displacement relationship function, the vertical displacement of the sticky rod from the initial zero point position is calculated. The sum of the initial zero point position and the vertical displacement of the continued downward pressure constitutes the target displacement in the vertical direction.
5. The dust control method according to claim 1, characterized in that, In the morphology-zero point mapping function, the dust with different height extrema H corresponds to different initial zero point positions; the initial zero point position increases as the height extrema H increases; if the height extrema H is positive, it is determined to be a removable attachment type, and S3 to S5 are executed; if the height extrema H is negative, it is determined to be an irreversible defect type, the position where the height extrema H is negative is skipped and the defect information is recorded.
6. The dust control method according to claim 1, characterized in that, The adhesive rod attribute parameters include the external dimensions and expiration date information of the adhesive rod; the three-dimensional geometric features are acquired by a three-dimensional detection module, which is any one of a white light interferometer sensor, a spectral confocal sensor, a laser triangulation sensor, a structured light projection sensor, or a time-of-flight sensor.
7. A dust removal control system, characterized in that, A method for implementing dust control as described in any one of claims 1 to 6, comprising: The parameter acquisition module is used to acquire the attribute parameters of the sticky rod and the three-dimensional geometric morphology features of the dust on the surface of the semiconductor wafer. The three-dimensional geometric morphology features include the height extreme value H of the dust relative to the reference plane of the wafer surface and the distribution size D of the dust in the direction parallel to the reference plane. The model building module is used to build a sticky rod pressure control algorithm model based on the sticky rod property parameters and the three-dimensional geometric features. The real-time detection module is used to acquire the real-time parameters of the current sticky stick and the three-dimensional geometric features of the current dust during the cleaning process. The displacement calculation module is used to input the real-time parameters and the current three-dimensional geometric features of the dust into the sticky rod pressure control algorithm model to calculate the target displacement of the sticky rod in the horizontal and vertical directions. The execution control module is used to control the current sticky rod to move to the target position according to the target displacement, so that the front end of the current sticky rod comes into contact with the current dust to achieve dust removal.
8. A dust removal device, characterized in that, include: Machine base; A motion component, wherein the motion component is disposed on the machine base; A support platform, mounted on the motion component, is used to support semiconductor wafers; A three-dimensional detection module, located above the motion component, is used to acquire the three-dimensional geometric morphology features of dust on the surface of the semiconductor wafer; an adhesive stick cleaning component, located on the motion component, is used to clean the dust on the surface of the semiconductor wafer. and The dust removal control system as described in claim 7 is electrically connected to the motion component, the three-dimensional detection module, and the adhesive stick cleaning component, respectively.
9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the dust removal control method as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Semiconductor device surface dust removal method and system based on visual inspection
CN117470104A
Flexible contact control method of semiconductor element, cleaning device and medium
CN122284349A
