A wafer processing equipment for semiconductor automated production
Through data analysis and clustering technology, the cleaning frequency and time settings of semiconductor wafer cleaning equipment are optimized, which solves the problem of poor cleaning effect in the existing technology, achieves more efficient particulate matter removal, and improves the quality of the wafer surface.
Patent Information
- Application Number
- CN202411195589.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-29
AI Technical Summary
In the prior art, in semiconductor wafer cleaning, the setting of cleaning frequency and time is unreasonable, resulting in poor removal of particulate matter on the wafer surface and affecting the performance of the circuit structure.
The data of the historical cleaning process is obtained through the data acquisition module, the clustering module analyzes the difference in the number of particles, and the cleaning control module clusters according to the optimal cleaning effect parameters and process parameters, and sets a reasonable cleaning frequency and time to improve the cleaning effect.
It improves the rationality of the frequency and time setting of the wafer cleaning process, improves the cleaning effect, ensures the sufficient removal of particles on the wafer surface, and improves the performance of the circuit structure.
Smart Images

Figure CN118712103B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wafer processing, and in particular to wafer processing equipment for semiconductor automated production. Background Art
[0002] Semiconductor wafers are the core materials in semiconductor manufacturing, usually made of silicon, and used to manufacture integrated circuits and microelectronic devices. During the production and processing of wafers, they are easily affected by tiny particles on equipment and materials, causing particles to adhere to the surface of the wafer, affecting the surface flatness of the wafer surface, which may lead to performance defects in the circuit structure. In order to prevent particles on the wafer surface from affecting the quality and performance of the wafer, particles on the wafer surface need to be removed.
[0003] Megasonic cleaning is a widely used method for cleaning semiconductor wafers. The existing technology cleans the particles on the surface of the wafer by setting the cleaning frequency and cleaning time of megasonic cleaning. Setting the cleaning frequency and cleaning time can meet the cleaning needs of most wafers. However, due to the different sizes of particles attached to the surface of the wafer, the requirements for cleaning frequency and cleaning time are different. If the set cleaning frequency and cleaning time do not meet the needs of the wafer, it will affect the rationality of the wafer cleaning control, and ultimately lead to poor wafer cleaning effect. Summary of the invention
[0004] In order to solve the technical problem that the prior art has a poor effect on wafer surface cleaning, the purpose of the present invention is to provide a wafer processing device for semiconductor automated production, and the technical solution adopted is as follows:
[0005] In order to solve the above technical problems, the present invention provides a wafer processing device for semiconductor automated production, the wafer processing device includes a processing device body, the wafer processing device also includes a cleaning processing device, the cleaning processing device includes:
[0006] A data acquisition module is used to acquire a process data set of each historical cleaning process in a historical database and a wafer image of a cleaning process to be processed; the process data set includes: an image of a wafer to be cleaned, an image of a wafer after cleaning, a degree of damage, and cleaning process parameters;
[0007] A clustering module is used to obtain the number of each type of particle in each wafer image; obtain a distance measurement value between any two historical cleaning processes according to the difference in the number of particles between the wafer image to be cleaned and the wafer image after cleaning in any two of the historical cleaning processes; cluster all the historical cleaning processes according to the distance measurement value to obtain each particle cleaning category;
[0008] A cleaning control module is used to obtain the best cleaning effect parameters and the best cleaning process parameters of the particle cleaning category according to the damage degree and cleaning process parameters of each historical cleaning process in the particle cleaning category; obtain the overall matching degree of the particle cleaning category according to the difference in quantity between the wafer image of the cleaning process to be processed and the wafer image to be cleaned in the particle cleaning category, the difference in quantity between the wafer image to be cleaned and the wafer image after cleaning in the historical cleaning process in the particle cleaning category, and the best cleaning effect parameters of the particle cleaning category; and perform cleaning control on the cleaning process to be processed according to the overall matching degree of the cleaning category and the best cleaning process parameters.
[0009] Furthermore, the method for obtaining the distance metric value includes:
[0010] Taking all the particle types corresponding to any two of the historical cleaning processes as all the particle types to be analyzed in the two historical cleaning processes;
[0011] Obtaining a category removal particle index of the type of particle to be analyzed in each of any two of the historical cleaning processes according to a difference in the number of particles in the type of particle to be analyzed corresponding to the wafer image to be cleaned and the wafer image after cleaning in each of any two of the historical cleaning processes;
[0012] According to the difference between the particle removal indexes of the types of particles to be analyzed in any two of the historical cleaning processes, obtaining the difference index of the removal effect of the types of particles to be analyzed in the two historical cleaning processes;
[0013] According to the difference in the number of particles in the type of particles to be analyzed corresponding to the two wafer images to be cleaned in any two of the historical cleaning processes, obtaining the category initial condition confidence index of the type of particles to be analyzed in the two historical cleaning processes;
[0014] According to the category initial condition confidence index and the category removal effect difference index, obtaining the category distance index of the type of particulate matter to be analyzed in the two historical cleaning processes;
[0015] The category distance indicators of all types of particulate matter to be analyzed in the two historical cleaning processes are combined to obtain the distance measurement value of the two historical cleaning processes.
[0016] Furthermore, the method for obtaining the category particle removal index includes:
[0017] By calculating the absolute value of the difference in the number of particles in the type of particle to be analyzed between the wafer image to be cleaned in the historical cleaning process and the wafer image after cleaning, calculating the sum of the number of particles in the type of particle to be analyzed in the wafer image to be cleaned and the denominator adjustment factor, and calculating the ratio of the absolute value of the difference to the sum, the category removal particle index of the type of particle to be analyzed in the historical cleaning process is obtained.
[0018] Furthermore, the cleaning process parameters are process frequency and process time.
[0019] Furthermore, the method for obtaining the optimal cleaning effect parameters and the optimal cleaning process parameters of the particle cleaning category includes:
[0020] Based on a two-dimensional coordinate system, the horizontal axis of the two-dimensional coordinate system is the process time, and the vertical axis of the two-dimensional coordinate system is the process frequency; the coordinate points determined in the two-dimensional coordinate system corresponding to the process frequency and the process time of all historical cleaning processes in the particle cleaning category are taken as a data point, if the process time corresponds to one data point, the data point is taken as the target data point of the process time; if the process time corresponds to multiple data points, the mean of the multiple data points corresponding to the process time is calculated to obtain the target data point of the process time; all target data points are counted as the data to be analyzed in the particle cleaning category;
[0021] In the data to be analyzed, according to the volatility of the local process frequency of the target data point, the surrounding frequency fluctuation index of the target data point is obtained;
[0022] According to the overall distribution of the damage degree of all historical cleaning processes corresponding to the process time of the target data point, the damage measurement index of the target data point is obtained;
[0023] According to the surrounding frequency fluctuation index and the damage measurement index, a cleaning effect index of the target data point is obtained; the surrounding frequency fluctuation index and the damage measurement index are both negatively correlated with the cleaning effect index;
[0024] The best data point is selected from all target data points according to the cleaning effect index; the cleaning effect index of the best data point is used as the best cleaning effect parameter of the particle cleaning category; the cleaning process parameter of the best data point is used as the best cleaning process parameter of the particle cleaning category.
[0025] Furthermore, the method for obtaining the ambient frequency fluctuation index includes:
[0026] In the data to be analyzed, the surrounding data points of the target data point are obtained; the variance of all the surrounding data points of the target data point corresponding to the process frequency is calculated to obtain the surrounding frequency fluctuation index of the target data point.
[0027] Furthermore, the method for obtaining the overall matching degree includes:
[0028] Obtaining an overall particle removal index of the particle cleaning category according to a difference in quantity between the particles in the wafer image to be cleaned and the wafer image after cleaning in each historical cleaning process in the particle cleaning category;
[0029] According to the difference in quantity between the wafer image of the cleaning process to be processed and the particles in each wafer image to be cleaned in the particle cleaning category, obtaining an initial state similarity index of the particle cleaning category;
[0030] The overall matching degree of the particle cleaning category is obtained according to the overall particle removal index, the initial state similarity index and the best cleaning effect parameter; the overall particle removal index, the initial state similarity index and the best cleaning effect parameter are all positively correlated with the overall matching degree.
[0031] Furthermore, the method for obtaining the overall particle removal index includes:
[0032] Calculate the absolute value of the difference in the number of each particle in the wafer image to be cleaned and the wafer image after cleaning in the historical cleaning process, calculate the ratio of the absolute value of the difference in the number of each particle to the number of particles in the wafer image to be cleaned, and obtain the cleaning effect value of each particle; in the particle cleaning category, calculate the mean of the cleaning effect values of all types of particles in all the historical cleaning processes to obtain the overall particle removal index of the particle cleaning category.
[0033] Further, the method for cleaning control of the cleaning process to be processed includes:
[0034] The cleaning category corresponding to the largest overall matching degree is taken as the best matching category, and the best cleaning process parameters of the best matching category are set as the cleaning process parameters of the cleaning process to be processed.
[0035] Furthermore, the method for obtaining the damage degree includes:
[0036] A neural network is used to perform damage assessment on the damage condition of the cleaned wafer image of each historical cleaning process to obtain the damage degree of each historical cleaning process.
[0037] The present invention has the following beneficial effects:
[0038] By constructing a distance metric value between two historical cleaning processes, the difference in the particle removal effect of the historical cleaning process is reflected, and then all historical cleaning processes are clustered to obtain each particle cleaning category. The particle cleaning category can reflect the historical cleaning process with similar particle removal effect, so that the frequency and time of the cleaning process to be processed can be set according to the particle cleaning category. Considering that the particles on the surface of the wafer are affected by the joint effect of time and frequency, when the time and frequency do not meet the requirements, there is a situation where the particles are not fully cleaned. The main purpose of the present invention is to improve the rationality of setting the frequency and time of the cleaning process to be processed, thereby improving the cleaning effect of the cleaning process to be processed. The particle cleaning category reflects the historical cleaning process with similar particle removal effect, analyzes the time and frequency of the historical cleaning process in the particle cleaning category, and preliminarily reflects the best particle cleaning effect of the particle cleaning category through the best cleaning effect parameter of the particle cleaning category. The best cleaning process parameter reflects the time and frequency corresponding to the best particle cleaning effect. Then, the degree of proximity between the cleaning process to be processed and the uncleaned wafer state of the particle cleaning category and the cleaning effect of the particle cleaning category are analyzed to obtain the overall matching degree of the particle cleaning category. The higher the overall matching degree, the more suitable the cleaning process to be processed is for adopting the optimal cleaning process parameters of the cleaning category. According to the overall matching degree and the optimal cleaning process parameters of the cleaning category, the cleaning process to be processed is controlled, and the cleaning process parameters of the cleaning process to be processed are reasonably set, so that the cleaning process to be processed can better meet the wafer cleaning requirements and improve the wafer cleaning control effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0040] Figure 1 A schematic structural diagram of a wafer processing device for semiconductor automated production provided by an embodiment of the present invention;
[0041] Figure 2 A schematic structural diagram of a semiconductor automated production wafer processing equipment body provided by an embodiment of the present invention;
[0042] Figure 3 A schematic diagram of the internal structure of a cavity of a semiconductor automated production wafer processing equipment body provided by an embodiment of the present invention;
[0043] Figure 4A schematic diagram of an openable and closable lens baffle of a semiconductor automated production wafer processing equipment body provided by an embodiment of the present invention;
[0044] Figure 5 A device block diagram of a cleaning device provided by an embodiment of the present invention;
[0045] Figure 6 A flow chart of a method for obtaining a distance metric value provided by an embodiment of the present invention;
[0046] Figure 7 A flow chart of a method for obtaining optimal cleaning effect parameters and optimal cleaning process parameters provided by an embodiment of the present invention;
[0047] Figure 8 A flow chart of a method for obtaining an overall matching degree provided by an embodiment of the present invention;
[0048] Figure numerals: 1 - mounting frame, 2 - cavity, 3 - turntable, 4 - moving rocker arm, 5 - megasonic cleaning mechanism, 6 - two-fluid cleaning mechanism, 7 - air filtering mechanism, 8 - CCD camera, 9 - openable and closable lens baffle. DETAILED DESCRIPTION
[0049] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a wafer processing equipment for semiconductor automated production proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0050] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0051] A specific solution of a wafer processing device for semiconductor automated production provided by the present invention is described in detail below in conjunction with the accompanying drawings.
[0052] Figure 1 FIG. 1 shows a schematic diagram of a structure of a semiconductor automated production wafer processing device provided by an embodiment of the present invention. Figure 1 As shown, the wafer processing device 0 includes a processing device body 01 and a cleaning processing device 02 which are communicatively connected. In the embodiment of the present invention, the communication connection is performed wirelessly.
[0053] The processing device body 01 refers to the existing semiconductor automated production wafer processing equipment. The semiconductor automated production wafer processing equipment provided by the embodiment of the present invention is based on the existing semiconductor automated production wafer processing equipment and adds a cleaning processing device 02 to achieve wafer cleaning control. Figure 2 , which shows a schematic diagram of the structure of a semiconductor automated production wafer processing equipment body provided by an embodiment of the present invention. The wafer processing equipment body includes a mounting frame 1, and the mounting frame 1 has a cavity 2 inside. Figure 3 , which shows a schematic diagram of the internal structure of a cavity of a wafer processing equipment body for semiconductor automated production provided by an embodiment of the present invention.
[0054] The process of cleaning the wafer by the above-mentioned wafer processing equipment is briefly described as follows:
[0055] The wafer is placed on the turntable 3 in the chamber 2. The machine automatically controls the movable rocker 4 to move to the top of the wafer. At this time, the megasonic cleaning mechanism is started. The two-fluid cleaning mechanism 6 sprays the cleaning liquid while the megasonic cleaning mechanism 5 releases the sound waves of the set duration and frequency to clean the wafer. Among them, the CCD camera 8 is installed next to the megasonic cleaning mechanism 5. The CCD camera 8 is protected by the openable lens baffle 9 to prevent the liquid from damaging the lens. The lens baffle can be opened before and after cleaning to take images of the wafer surface. Please refer to Figure 4 , which shows a schematic diagram of an openable and closable lens baffle of a semiconductor automated production wafer processing equipment body provided by an embodiment of the present invention. During the cleaning process, impurities and pollutants in the air are removed by the air filter mechanism 7 to keep the inside of the system clean.
[0056] Since the semiconductor automated production wafer processing equipment provided by the embodiment of the present invention focuses on the cleaning processing equipment 02, please refer to Figure 5 , which shows a device block diagram of a cleaning processing device provided by an embodiment of the present invention. The device includes: a data acquisition module 101, a clustering module 102, and a cleaning control module 103.
[0057] The data acquisition module 101 is used to acquire the process data set of each historical cleaning process in the historical database and the wafer image of the cleaning process to be processed; the process data set includes: the wafer image to be cleaned, the wafer image after cleaning, the damage degree and the cleaning process parameters.
[0058] In order to achieve reasonable settings of frequency and time, it is first necessary to obtain the process data set of each historical cleaning process in the historical database and the wafer image of the cleaning process to be processed, so as to provide data reference for subsequent reasonable control of the cleaning process to be processed.
[0059] The present invention is explained by using a wafer processing device using a megasonic wave cleaning method to achieve wafer surface cleaning. It should be noted that megasonic wave cleaning is a prior art well known to those skilled in the art. Here, only a brief description of using megasonic waves to clean the surface particles of the wafer is given: the principle of megasonic wave cleaning is to remove the particles attached to the surface of the wafer by the interaction between megahertz-level high-energy sound waves and cleaning liquid, and to achieve the cleaning of the surface of the wafer by utilizing the high-energy frequency vibration effect and the chemical reaction of the cleaning liquid. Among them, the time and frequency sampled in the process of using megasonic waves to clean the wafer have a significant impact on the cleaning effect, that is, the higher the frequency, the stronger the directionality of the sound wave flow, and the more suitable it is for removing particles of smaller size. Within an appropriate cleaning time, as the cleaning time is extended, the particles on the surface of the wafer will be gradually removed, and the cleaning effect will be significantly improved. However, when the cleaning time exceeds a certain limit, as the cleaning time is extended, the cleaning effect will no longer be significantly improved.
[0060] Since the wafer processing equipment can clean the wafer through the megasonic cleaning module, the wafer is imaged through the image acquisition module. A process data set of the historical cleaning process in the historical database is obtained from the wafer processing equipment. Specifically, the present invention is explained by taking the example of using a frequency to clean a wafer surface in any historical cleaning process, and megasonic cleaning is performed on the wafer according to the frequency and time set by the equipment. Before megasonic cleaning, the wafer is imaged and the image preprocessing operation is performed on the captured image to obtain the image of the wafer to be cleaned in the historical cleaning process; at the end of megasonic cleaning, the wafer is imaged and the image preprocessing operation is performed on the captured image to obtain the image of the wafer after cleaning in the historical cleaning process; the frequency and time corresponding to the historical cleaning process are used as the process frequency and process time of the historical cleaning process, and the process frequency and process time are used as the cleaning process parameters of the historical cleaning process. It should be noted that the corresponding frequency of the historical cleaning process is the sound wave frequency set by the megasonic cleaning module in the historical cleaning process, and the corresponding time of the historical cleaning process is the duration of the historical cleaning process.
[0061] In order to reasonably set up the cleaning process for the latest wafer to be cleaned, it is necessary to capture images of it and take the cleaning process corresponding to the latest wafer to be cleaned as the cleaning process to be processed; capture images of the latest wafer to be cleaned and perform image preprocessing on the captured images to obtain wafer images of the cleaning process to be processed.
[0062] Since the analysis of the cleaning effect requires the analysis of the degree of damage to the wafer after cleaning, the degree of damage in the historical cleaning process is obtained from the wafer processing equipment. Specifically, defect detection is performed on the wafer image after cleaning to obtain the corresponding degree of damage of the wafer image after cleaning in each historical cleaning process. In one embodiment of the present invention, the method for obtaining the degree of damage includes:
[0063] The cleaned wafer image includes damage information;
[0064] According to the damage information of the wafer image after cleaning in each historical cleaning process, a neural network is used to perform damage assessment to obtain the damage degree of the historical cleaning process.
[0065] Specifically, it should be noted that CNN neural network is a technical means well known to those skilled in the art, and will not be described in detail here. Only a brief process of obtaining the damage degree of the historical cleaning process according to the damage information of the wafer image after cleaning of each historical cleaning process by using CNN neural network in one embodiment of the present invention is briefly described:
[0066] A CNN neural network is trained using a reference data set, which includes multiple images of wafers after cleaning. Since the images of wafers after cleaning include damage information of the wafers after cleaning, such as the size and shape of cracks, the overall degree of surface damage, and specific defects, the images of wafers after cleaning are scored from 0 to 1 as the degree of damage of the images of wafers after cleaning, wherein the higher the severity of damage to the wafers after cleaning, the higher the degree of damage; then the labeled data set is divided into a training set and a validation set according to a preset ratio, the reference data set is input into the CNN neural network for training, the loss function is cross entropy, and the gradient descent method is used until the loss function converges to obtain a trained CNN neural network; each image of the wafer after cleaning is input into the trained CNN neural network to obtain the degree of damage of the wafer image after cleaning, that is, to obtain the degree of damage of the historical cleaning process. In an embodiment of the present invention, the preset ratio is 7:3, and the implementer can set it according to the implementation scenario.
[0067] The specific image preprocessing operation is a technical means well known to those skilled in the art and is not limited here. In the embodiment of the present invention, the image preprocessing operation includes graying, noise reduction, and contrast enhancement. The embodiment of the present invention uses a weighted average method for graying, uses a Gaussian filter for noise reduction, and uses a histogram equalization for contrast enhancement. The implementer can set it according to the actual situation. It should be noted that in order to facilitate calculations, all indicator data involved in the calculations in the embodiment of the present invention are subjected to data preprocessing to eliminate the dimension effect. The specific means of removing the dimension effect are technical means well known to those skilled in the art and are not limited here.
[0068] The clustering module 102 is used to obtain the number of each type of particle in each wafer image; obtain the distance measurement value of any two historical cleaning processes based on the difference in the number of particles between the wafer image to be cleaned and the wafer image after cleaning in any two historical cleaning processes; cluster all historical cleaning processes based on the distance measurement value to obtain each particle cleaning category.
[0069] By constructing the distance measurement value of two historical cleaning processes, the difference in the particle removal effect of the historical cleaning processes can be reflected, and then all historical cleaning processes can be clustered to obtain each particle cleaning category. The particle cleaning category can reflect the historical cleaning processes with similar particle removal effects, so that the frequency and time of the cleaning process to be processed can be set according to the particle cleaning category.
[0070] Preferably, in one embodiment of the present invention, considering that the wafer is easily affected by tiny particles on the equipment and materials during the production process, so that particles are attached to the surface of the wafer, that is, the wafer image includes particle information, and all particles in the wafer image are identified through a neural network. Since different particle sizes have different requirements for cleaning time and cleaning frequency, all particles are divided according to the size of the particles, and the number of each type of particle in the wafer image is obtained. The method for obtaining the number of each type of particle in the wafer image includes:
[0071] The wafer image includes particle information; based on the particle information of the wafer image, the CNN neural network is used to perform image recognition on the wafer image to obtain each particle;
[0072] According to the size of the particles, all particles are divided to obtain the number of each type of particles in the wafer image. It should be noted that the CNN neural network is a technical means well known to those skilled in the art and will not be described in detail here.
[0073] Specifically, since there are particles attached to the surface of the wafer, that is, the wafer image includes particle information, the CNN neural network is used to perform image recognition on the wafer image to obtain all particles, and the area of each particle is used as the size of the particle. Particles with a size of 0-0.1 are classified as one particle, that is, the particle type is tiny particles; particles with a size of 0.1-0.2 are classified as one particle, that is, the particle type is small particles; particles with a size of 0.3-0.4 are classified as one particle, that is, the particle type is medium particles; particles with a size greater than 0.4 are classified as one particle, that is, the particle type is large particles.
[0074] See also Figure 6 , which shows a flow chart of a method for obtaining a distance metric value in one embodiment of the present invention. Preferably, in one embodiment of the present invention, preferably, in one embodiment of the present invention, the distance metric value of two historical cleaning processes is constructed to reflect the difference in the particle removal effect of the historical cleaning process. The method for obtaining the distance metric value includes:
[0075] Step S101: taking all particle types corresponding to any two historical cleaning processes as all particle types to be analyzed in the two historical cleaning processes.
[0076] In order to ensure that all possible particle size types are considered when comparing two historical cleaning processes, regardless of whether these types appear in both historical cleaning processes at the same time, the union of all particle types in the two historical cleaning processes reflects the unique and common particle types in the two historical cleaning processes, thereby providing a comprehensive perspective to reflect the possible impact of particle size on cleaning results.
[0077] In one embodiment of the present invention, the union of all particle types in the wafer images to be cleaned and the wafer images after cleaning in two historical cleaning processes is used as all particle types to be analyzed in the two historical cleaning processes.
[0078] Step S102: Obtain a category removal particle index of the type of particle to be analyzed in each of any two historical cleaning processes based on the difference in the number of particles in the corresponding type of particle to be analyzed between the wafer image to be cleaned and the wafer image after cleaning in each of any two historical cleaning processes.
[0079] The larger the category removal particle index is, the better the particle removal effect of the historical cleaning process is for the type of particles to be analyzed.
[0080] In one embodiment of the present invention, by calculating the absolute value of the difference between the number of particles in the type of particles to be analyzed in the wafer image to be cleaned in the historical cleaning process and the wafer image after cleaning, the sum of the number of particles in the type of particles to be analyzed in the wafer image to be cleaned and the denominator adjustment factor is calculated, and the ratio of the absolute value of the difference to the sum is calculated to obtain the category particle removal index of the type of particles to be analyzed in the historical cleaning process. In one embodiment of the present invention, the denominator adjustment factor is used to ensure that the denominator is not 0, and the denominator adjustment factor is a number greater than 0. This embodiment sets it to 0.1, and the implementer can set it according to the implementation scenario.
[0081] Step S103: according to the difference between the particle removal indexes of the types of the to-be-analyzed particles in any two historical cleaning processes, the removal effect difference index of the types of the to-be-analyzed particles in the two historical cleaning processes is obtained.
[0082] The larger the removal effect difference index is, the greater the difference in the particle cleaning effects of the two historical cleaning processes for the type of particle to be analyzed is.
[0083] In one embodiment of the present invention, the absolute value of the difference between the category removal particle index of each type of particle to be analyzed in two historical cleaning processes is calculated to obtain the removal effect difference index of the type of particle to be analyzed in the two historical cleaning processes, and the difference is reflected by the absolute value of the difference.
[0084] Step S104: according to the difference in the number of particles in the particle types to be analyzed corresponding to the two wafer images to be cleaned in any two historical cleaning processes, obtain the category initial condition confidence index of the particle types to be analyzed in the two historical cleaning processes.
[0085] The larger the confidence index of the category initial condition is, the closer the number of particles in the type of particles to be analyzed corresponding to the two wafer images to be cleaned is, which means that the particles in the initial type of particles to be analyzed of the two wafers are more similar, and the higher the confidence in the degree of difference in cleaning effect is.
[0086] In one embodiment of the present invention, the absolute value of the difference between the number of particles in the type of particles to be analyzed corresponding to the two wafer images to be cleaned is calculated, and the sum of the absolute value of the difference and the denominator adjustment parameter is calculated, and the reciprocal of the sum is used as the confidence index of the category initial state of the type of particles to be analyzed in the two historical cleaning processes. In one embodiment of the present invention, the denominator adjustment parameter is used to ensure that the denominator is not 0, and the denominator adjustment parameter is a number greater than 0. In this embodiment, it is set to 0.1, and the implementer can set it according to the implementation scenario.
[0087] Step S105: According to the category initial condition confidence index and the category removal effect difference index, the category distance index of the type of particle to be analyzed in two historical cleaning processes is obtained.
[0088] The category distance index more accurately reflects the degree of difference in the particle removal effects of the two historical cleaning processes on the type of particle to be analyzed.
[0089] In one embodiment of the present invention, the product of the category initial condition confidence index and the category removal effect difference index is calculated to obtain the category distance index of the types of particles to be analyzed in two historical cleaning processes.
[0090] Step S106: The category distance indicators of all the types of particles to be analyzed in the two historical cleaning processes are integrated to obtain the distance measurement values of the two historical cleaning processes.
[0091] The larger the distance metric value, the greater the difference in the overall particle removal effect of the two historical cleaning processes.
[0092] In one embodiment of the present invention, the mean of the category distance index of all the types of particles to be analyzed in two historical cleaning processes is calculated, and the mean is normalized to obtain the distance measurement value of the two historical cleaning processes. In one embodiment of the present invention, the normalization method is: normalization is performed using the norm normalization function to limit the value range to between 0 and 1.
[0093] In one embodiment of the present invention, the distance metric value acquisition formula includes:
[0094] ,
[0095] In the formula, Cleansing process for history and historical cleansing process The distance metric of ; Cleansing process for history and historical cleansing process The corresponding total number of all particle types to be analyzed; Cleansing process for history The image of the wafer to be cleaned The number of particles in the particle type to be analyzed; Cleansing process for history The image of the wafer to be cleaned is in the The number of particles in the particle type to be analyzed; is the absolute value symbol; Adjustment parameter for the denominator; Cleansing process for history and historical cleansing process In the The confidence index of the initial condition of the type of particle to be analyzed; Cleansing process for history The image of the wafer to be cleaned and the image of the wafer after cleaning The absolute value of the difference in the number of particles in the particle types to be analyzed; is the denominator adjustment factor; Cleansing process for history No. The particle removal index of the category of the particle type to be analyzed; Cleansing process for history The image of the wafer to be cleaned and the image of the wafer after cleaning The absolute value of the difference in the number of particles in the particle types to be analyzed; Cleansing process for history No. The particle removal index of the category of the particle type to be analyzed; Cleansing process for history and historical cleansing process In the The difference index of the removal effect of the particle types to be analyzed; Cleansing process for history and historical cleansing process In the The category distance index of the particle type to be analyzed; is the normalization function.
[0096] Preferably, in one embodiment of the present invention, a particle cleaning category is constructed to reflect historical cleaning processes with similar particle removal effects. The method for obtaining the particle cleaning category includes:
[0097] Using the k-means clustering algorithm, all historical cleaning processes in the historical database are clustered according to the distance measurement value between every two historical cleaning processes to obtain the cleaning categories of each particle.
[0098] It should be noted that the k-means clustering algorithm is a technical means well known to those skilled in the art, and will not be described in detail here. Only a brief description is given of a brief process of using the k-means clustering algorithm in one embodiment of the present invention to cluster all historical cleaning processes in the historical database according to the distance measurement value between every two historical cleaning processes to obtain each particle cleaning category:
[0099] Considering that the larger the distance measurement value between two historical cleaning processes, the greater the difference in the particle removal effect, the optimal k value is obtained by the elbow method, and the distance measurement value between each two historical cleaning processes is substituted into the clustering process, and all historical cleaning processes in the historical database are clustered to obtain clustering results and various particle cleaning categories. The differences in historical cleaning processes in different particle cleaning categories are made as large as possible, and the differences in historical cleaning processes in the same particle cleaning category are made as small as possible, that is, the similarity of the particle removal effect of the historical cleaning process in each particle cleaning category is as large as possible. In other embodiments of the present invention, the optimal k value of the k-means clustering algorithm can be determined by other methods such as the silhouette coefficient method, which is not limited here.
[0100] The cleaning control module 103 is used to obtain the best cleaning effect parameters and the best cleaning process parameters of the particle cleaning category according to the damage degree and cleaning process parameters of each historical cleaning process in the particle cleaning category; obtain the overall matching degree of the particle cleaning category according to the difference in the number of particles between the wafer image of the cleaning process to be processed and the wafer image to be cleaned in the particle cleaning category, the difference in the number of particles between the wafer image to be cleaned in the historical cleaning process in the particle cleaning category and the wafer image after cleaning, and the best cleaning effect parameters of the particle cleaning category; and perform cleaning control on the cleaning process to be processed according to the overall matching degree of the cleaning category and the best cleaning process parameters.
[0101] Considering that the particles on the surface of the wafer are affected by the combined effects of time and frequency, when the time and frequency do not meet the requirements, there is a situation where the particles are not fully cleaned. The main purpose of the present invention is to improve the rationality of the frequency and time settings of the cleaning process to be processed, thereby improving the cleaning effect of the cleaning process to be processed. The particle cleaning category reflects the historical cleaning process with similar particle removal effects. The time and frequency of the historical cleaning process in the particle cleaning category are analyzed. The best cleaning effect parameters of the particle cleaning category are used to preliminarily reflect the best particle cleaning effect of the particle cleaning category. The best cleaning process parameters reflect the time and frequency corresponding to the best particle cleaning effect. Then analyze the degree of proximity between the cleaning process to be processed and the uncleaned wafer state of the particle cleaning category and the cleaning effect of the particle cleaning category, and obtain the overall matching degree of the particle cleaning category. The higher the overall matching degree, the more suitable it is for the cleaning process to be processed to adopt the best cleaning process parameters of the cleaning category. According to the overall matching degree and the best cleaning process parameters of the cleaning category, the cleaning process to be processed is cleaned and controlled, and the cleaning process parameters of the cleaning process to be processed are reasonably set, so that the cleaning process to be processed better meets the wafer cleaning requirements and improves the wafer cleaning control effect.
[0102] To determine the best cleaning effect and corresponding time and frequency in the particle cleaning category, refer to Figure 7 , which shows a flow chart of a method for obtaining an optimal cleaning effect parameter and an optimal cleaning process parameter in one embodiment of the present invention. Preferably, in one embodiment of the present invention, the method for obtaining the optimal cleaning effect parameter and the optimal cleaning process parameter of the particle cleaning category includes:
[0103] Step S201: Based on a two-dimensional coordinate system, the horizontal axis of the two-dimensional coordinate system is the process time, and the vertical axis of the two-dimensional coordinate system is the process frequency; the coordinate points corresponding to the process frequency and process time of all historical cleaning processes in the particle cleaning category are determined in the two-dimensional coordinate system as a data point. If the process time corresponds to one data point, the data point is used as the target data point of the process time; if the process time corresponds to multiple data points, the mean of the multiple data points corresponding to the process time is calculated to obtain the target data point of the process time; all target data points are counted as the data to be analyzed in the particle cleaning category.
[0104] Since the particle cleaning category can reflect the historical cleaning process with similar particle removal effects, considering that the particle effects are similar, the shorter the process time, the more targeted the process frequency is for the particle cleaning effect; when the process time is short, the sound wave frequency is stable in a certain range and has strong targeting; as the process time increases, the frequency targeting becomes weaker and the frequency consistency becomes lower. If the process time corresponds to a data point, the data point is used as the target data point of the process time; considering that there are multiple process frequencies corresponding to a process time, if the process time corresponds to more than one data point, the mean of the process frequency corresponding to all data points of the process time is calculated to obtain the target data point of the process time, and all target data points are counted as the data to be analyzed for the particle cleaning category, and the relationship between the process frequency and the process time is constructed for subsequent analysis.
[0105] Step S202: in the data to be analyzed, according to the volatility of the local process frequency of the target data point, obtain the surrounding frequency fluctuation index of the target data point.
[0106] The smaller the surrounding frequency fluctuation index is, the higher the frequency consistency between the target data point and the surrounding data is. At this time, the corresponding frequency of the target data point is more targeted and more suitable for cleaning the particles corresponding to the cleaning category.
[0107] In one embodiment of the present invention, in the data to be analyzed, surrounding data points of a target data point in a preset surrounding range are obtained; the variance of the process frequency corresponding to all surrounding data points of the target data point is calculated to obtain a surrounding frequency fluctuation index of the target data point.
[0108] In order to analyze the distribution of surrounding data of the target data point and determine the preset surrounding range of the target data point, the method for obtaining the preset surrounding range includes: the preset surrounding range is an area constructed with the target data point as the center and the preset surrounding specification as the size; the preset surrounding specification is the minimum circumscribed rectangle including a preset number of target data points; all target data points in the preset surrounding range are used as surrounding data points in the preset surrounding range. In one embodiment of the present invention, the preset number is 5, and the implementer can set it according to the implementation scenario.
[0109] Step S203: Obtain a damage measurement index of the target data point according to the overall distribution of damage degrees of all historical cleaning processes corresponding to the process time of the target data point.
[0110] Considering that unreasonable process frequency and process time settings will cause damage to the wafer surface, the damage measurement index of the target data point is obtained. The larger the damage measurement index, the greater the corresponding damage degree of the target data point.
[0111] In one embodiment of the present invention, the average of the damage degrees of all historical cleaning processes corresponding to the process time of the target data point is used as the damage measurement indicator of the target data point.
[0112] Step S204: obtaining a cleaning effect index of the target data point according to the surrounding frequency fluctuation index and the damage measurement index; the surrounding frequency fluctuation index and the damage measurement index are negatively correlated with the cleaning effect index.
[0113] Considering that the larger the damage measurement index is, the greater the damage degree of the target data point is, which means the worse the cleaning effect is; the smaller the surrounding frequency fluctuation index is, the more targeted the frequency of the target data point is, and the more suitable it is for cleaning the particles corresponding to the cleaning category, which means the better the cleaning effect is, and the cleaning effect index of the target data point is obtained; the larger the cleaning effect index is, the better the cleaning effect of the target data point is.
[0114] The sum of the square of the surrounding frequency fluctuation index and the square of the damage measurement index is calculated to obtain the sum of squares; the sum of the square sum and the denominator adjustment factor is calculated, and the reciprocal of the sum is used as the cleaning effect indicator of the target data point.
[0115] Step S205: Filter out the best data point from all target data points according to the cleaning effect index; use the cleaning effect index of the best data point as the best cleaning effect parameter of the particle cleaning category; use the cleaning process parameter of the best data point as the best cleaning process parameter of the particle cleaning category.
[0116] Since the larger the cleaning effect index, the better the cleaning effect corresponding to the target data point, the target data point corresponding to the largest cleaning effect index is taken as the optimal data point. The cleaning effect index, process frequency and process time of the optimal data point are taken as the optimal cleaning effect parameters, optimal frequency and optimal time of the particle cleaning category, and the optimal frequency and optimal time are taken as the optimal cleaning process parameters. The optimal cleaning effect parameters of the particle cleaning category preliminarily reflect the optimal particle cleaning effect of the particle cleaning category, and the optimal frequency and optimal time reflect the time and frequency corresponding to the optimal particle cleaning effect.
[0117] In order to analyze the degree to which the optimal frequency and optimal time of the cleaning category are suitable for the cleaning process to be processed, it is considered that the closer the state of the cleaning process to be processed and the uncleaned wafer of the particle cleaning category is, the more suitable the cleaning process to be processed is to be classified into the particle cleaning category, that is, the more suitable it is to adopt the optimal frequency and optimal time corresponding to the particle cleaning category; in order to meet the wafer cleaning requirements corresponding to the cleaning process to be processed, the cleaning effect of the particle cleaning category must also be considered to obtain the overall matching degree of the particle cleaning category. The higher the overall matching degree, the better the cleaning effect of the cleaning category and the closer it is to the uncleaned wafer state of the cleaning process to be processed and the particle cleaning category, which means that the cleaning process to be processed is more suitable for adopting the optimal frequency and optimal time of the cleaning category.
[0118] See also Figure 8 , which shows a flow chart of a method for obtaining an overall matching degree in one embodiment of the present invention. Preferably, in one embodiment of the present invention, the method for obtaining an overall matching degree includes:
[0119] Step S301: obtaining an overall particle removal index of the particle cleaning category according to the difference in the number of particles between the wafer image to be cleaned and the wafer image after cleaning in each historical cleaning process in the particle cleaning category.
[0120] By analyzing the particle removal effects of each historical cleaning process in the particle cleaning category, the overall particle removal index of the particle cleaning category is obtained. The larger the overall particle removal index, the better the particle cleaning effect of the historical process in the particle cleaning category.
[0121] Calculate the absolute value of the difference between the number of each type of particle in the wafer image to be cleaned and the wafer image after cleaning in the historical cleaning process, calculate the ratio of the absolute value of the difference between the number of each type of particle and the number of particles in the wafer image to be cleaned, and obtain the cleaning effect value of each type of particle; in the particle cleaning category, calculate the average of the cleaning effect values of all types of particles in all historical cleaning processes to obtain the overall particle removal index of the particle cleaning category. The formula for obtaining the overall particle removal index in one embodiment of the present invention includes:
[0122] ,
[0123] In the formula, It is the overall particle removal index for the particle cleaning category; is the total number of all historical cleaning processes in the particle cleaning category; In the particle cleaning category, The total number of all particle types in the historical cleaning process; In the particle cleaning category, The image of the wafer to be cleaned and the image of the wafer after cleaning in the historical cleaning process The absolute value of the difference in the number of particles in each particle type; In the particle cleaning category, The first image of the wafer to be cleaned for the historical cleaning process The number of particles in each particle type; For the The first of a historical cleansing process The cleaning effect value for each particle type.
[0124] Step S302: obtaining an initial state similarity index of the particle cleaning category according to the difference in the number of particles between the wafer image of the to-be-processed cleaning process and each to-be-cleaned wafer image in the particle cleaning category.
[0125] The larger the initial state similarity index is, the closer the states of the cleaning process to be processed and the uncleaned wafer of the particle cleaning category are.
[0126] In one embodiment of the present invention, the types of all particles in the wafer image of the cleaning process to be processed are used as the types to be compared; the absolute value of the difference between the number of particles of the wafer image of the cleaning process to be processed and the type to be compared of the wafer image to be cleaned is used to obtain the initial type difference index of the type to be compared, the mean of the initial type difference index of all types to be compared is calculated to obtain the local state difference index of the wafer image of the cleaning process to be processed and the wafer image to be cleaned; the mean of the local state difference index of the wafer image of the cleaning process to be processed and all wafer images to be cleaned is calculated to obtain the overall difference index of the particle cleaning category, and the overall difference index is inversely normalized to obtain the initial state similarity index of the particle cleaning category. In one embodiment of the present invention, the method of inverse normalization is: the inverse of the overall difference index is used as the power of an exponential function with a natural constant as the base.
[0127] Step S303: Obtain the overall matching degree of the particle cleaning category according to the overall particle removal index, the initial state similarity index and the best cleaning effect parameter; the overall particle removal index, the initial state similarity index and the best cleaning effect parameter are all positively correlated with the overall matching degree.
[0128] The cleaning effect of the particle cleaning category is comprehensively reflected by the overall particle removal index and the optimal cleaning effect parameter. The larger the value, the better the cleaning effect. The larger the initial state similarity index, the closer the cleaning process to be processed is to the uncleaned wafer state of the particle cleaning category, that is, the more suitable it is to use the optimal frequency and optimal time corresponding to the particle cleaning category, and obtain the overall matching degree of the particle cleaning category. The higher the overall matching degree, the more suitable it is for the cleaning process to be processed to use the optimal frequency and optimal time of the cleaning category.
[0129] In one embodiment of the present invention, the product of the overall particle removal index, the initial state similarity index and the best cleaning effect parameter is calculated to obtain the overall matching degree of the particle cleaning category.
[0130] In one embodiment of the present invention, considering that the higher the overall matching degree is, the more suitable the cleaning process to be processed is for adopting the optimal cleaning process parameters of the cleaning category. The method for cleaning control of the cleaning process to be processed includes:
[0131] The cleaning category with the largest overall matching degree is regarded as the best matching category, and the best cleaning process parameters of the best matching category are set as the cleaning process parameters of the cleaning process to be processed. The cleaning process parameters of the cleaning process to be processed are input into the megasonic cleaning module to realize the control of the cleaning process to be processed. By reasonably setting the cleaning process parameters of the cleaning process to be processed, the wafer cleaning requirements can be better met and the wafer cleaning control effect can be improved.
[0132] In summary, the embodiment of the present invention provides a wafer processing equipment for semiconductor automated production. First, according to the distance metric value, all historical cleaning processes are clustered to obtain each particle cleaning category. According to the damage degree and cleaning process parameters of each historical cleaning process in the particle cleaning category, the optimal cleaning effect parameters and the optimal cleaning process parameters of the particle cleaning category are obtained; according to the difference in the number of particles between the wafer image of the cleaning process to be processed and the wafer image to be cleaned in the particle cleaning category, the difference in the number of particles between the wafer image to be cleaned in the historical cleaning process in the particle cleaning category and the wafer image after cleaning, and the optimal cleaning effect parameters of the particle cleaning category, the overall matching degree of the particle cleaning category is obtained, and the cleaning process to be processed is controlled. In the embodiment of the present invention, by reasonably setting the cleaning process parameters of the cleaning process to be processed, the cleaning process to be processed can better meet the wafer cleaning requirements and improve the wafer cleaning control effect.
[0133] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0134] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
Claims
1. A wafer processing device for semiconductor automated production, the wafer processing device comprising a processing device body, characterized in that: The wafer processing equipment also includes a cleaning processing equipment, and the cleaning processing equipment includes: A data acquisition module is used to acquire a process data set of each historical cleaning process in a historical database and a wafer image of a cleaning process to be processed; the process data set includes: an image of a wafer to be cleaned, an image of a wafer after cleaning, a degree of damage, and cleaning process parameters; A clustering module is used to obtain the number of each type of particle in each wafer image; obtain a distance measurement value between any two historical cleaning processes according to the difference in the number of particles between the wafer image to be cleaned and the wafer image after cleaning in any two of the historical cleaning processes; cluster all the historical cleaning processes according to the distance measurement value to obtain each particle cleaning category; A cleaning control module, for obtaining the best cleaning effect parameters and the best cleaning process parameters of the particle cleaning category according to the damage degree and cleaning process parameters of each historical cleaning process in the particle cleaning category; obtaining the overall matching degree of the particle cleaning category according to the difference in quantity between the wafer image of the cleaning process to be processed and the wafer image to be cleaned in the particle cleaning category, the difference in quantity between the wafer image to be cleaned and the wafer image after cleaning in the historical cleaning process in the particle cleaning category, and the best cleaning effect parameters of the particle cleaning category; and performing cleaning control on the cleaning process to be processed according to the overall matching degree of the cleaning category and the best cleaning process parameters; The method for obtaining the distance metric value includes: Taking all the particle types corresponding to any two of the historical cleaning processes as all the particle types to be analyzed in the two historical cleaning processes; Obtaining a category removal particle index of the type of particle to be analyzed in each of any two of the historical cleaning processes according to a difference in the number of particles in the type of particle to be analyzed corresponding to the wafer image to be cleaned and the wafer image after cleaning in each of any two of the historical cleaning processes; According to the difference between the particle removal indexes of the types of particles to be analyzed in any two of the historical cleaning processes, obtaining the difference index of the removal effect of the types of particles to be analyzed in the two historical cleaning processes; According to the difference in the number of particles in the type of particles to be analyzed corresponding to the two wafer images to be cleaned in any two of the historical cleaning processes, obtaining the category initial condition confidence index of the type of particles to be analyzed in the two historical cleaning processes; According to the category initial condition confidence index and the category removal effect difference index, obtaining the category distance index of the type of particulate matter to be analyzed in the two historical cleaning processes; Combining the category distance indicators of all types of particulate matter to be analyzed in the two historical cleaning processes, to obtain a distance measurement value of the two historical cleaning processes; The method for obtaining the category particle removal index includes: By calculating the absolute value of the difference between the number of particles in the type of particles to be analyzed in the wafer image to be cleaned in the historical cleaning process and the wafer image after cleaning, calculating the sum of the number of particles in the type of particles to be analyzed in the wafer image to be cleaned and the denominator adjustment factor, and calculating the ratio of the absolute value of the difference to the sum, the category removal particle index of the type of particles to be analyzed in the historical cleaning process is obtained; The method for obtaining the confidence index of the category initial condition includes: Calculate the absolute value of the difference in the number of particles in the type of particles to be analyzed corresponding to the two wafer images to be cleaned, calculate the sum of the absolute value of the difference and the denominator adjustment parameter, and use the reciprocal of the sum as the confidence index of the category initial condition of the type of particles to be analyzed in the two historical cleaning processes; wherein the denominator adjustment parameter and the denominator adjustment factor are both numbers greater than 0, and are both used to ensure that the denominator is not 0; The method for obtaining the category distance index includes: The product of the confidence index of the category initial condition and the index of the difference in the category removal effect is calculated to obtain the category distance index of the type of particles to be analyzed in the two historical cleaning processes; The method of obtaining the distance measurement value of the two historical cleaning processes by integrating the category distance indicators of all the types of particles to be analyzed in the two historical cleaning processes includes: The mean of the category distance index of all the particle types to be analyzed in the two historical cleaning processes is calculated, and the mean is normalized to obtain the distance measurement value of the two historical cleaning processes.
2. The wafer processing equipment for semiconductor automated production according to claim 1, characterized in that: The cleaning process parameters are process frequency and process time.
3. The wafer processing equipment for semiconductor automated production according to claim 2, characterized in that: The method for obtaining the optimal cleaning effect parameters and the optimal cleaning process parameters of the particle cleaning category includes: Based on a two-dimensional coordinate system, the horizontal axis of the two-dimensional coordinate system is the process time, and the vertical axis of the two-dimensional coordinate system is the process frequency; the coordinate points determined in the two-dimensional coordinate system corresponding to the process frequency and the process time of all historical cleaning processes in the particle cleaning category are taken as a data point, if the process time corresponds to one data point, the data point is taken as the target data point of the process time; if the process time corresponds to multiple data points, the mean of the multiple data points corresponding to the process time is calculated to obtain the target data point of the process time; all target data points are counted as the data to be analyzed in the particle cleaning category; In the data to be analyzed, according to the volatility of the local process frequency of the target data point, the surrounding frequency fluctuation index of the target data point is obtained; According to the overall distribution of the damage degree of all historical cleaning processes corresponding to the process time of the target data point, the damage measurement index of the target data point is obtained; According to the surrounding frequency fluctuation index and the damage measurement index, a cleaning effect index of the target data point is obtained; the surrounding frequency fluctuation index and the damage measurement index are both negatively correlated with the cleaning effect index; The best data point is selected from all target data points according to the cleaning effect index; the cleaning effect index of the best data point is used as the best cleaning effect parameter of the particle cleaning category; the cleaning process parameter of the best data point is used as the best cleaning process parameter of the particle cleaning category.
4. The wafer processing equipment for semiconductor automated production according to claim 3, characterized in that: The method for obtaining the ambient frequency fluctuation index includes: In the data to be analyzed, the surrounding data points of the target data point are obtained; the variance of all the surrounding data points of the target data point corresponding to the process frequency is calculated to obtain the surrounding frequency fluctuation index of the target data point.
5. The semiconductor automated production wafer processing equipment according to claim 1, characterized in that: The method for obtaining the overall matching degree includes: Obtaining an overall particle removal index of the particle cleaning category according to a difference in quantity between the particles in the wafer image to be cleaned and the wafer image after cleaning in each historical cleaning process in the particle cleaning category; According to the difference in quantity between the wafer image of the cleaning process to be processed and the particles in each wafer image to be cleaned in the particle cleaning category, obtaining an initial state similarity index of the particle cleaning category; The overall matching degree of the particle cleaning category is obtained according to the overall particle removal index, the initial state similarity index and the best cleaning effect parameter; the overall particle removal index, the initial state similarity index and the best cleaning effect parameter are all positively correlated with the overall matching degree.
6. The wafer processing equipment for semiconductor automated production according to claim 5, characterized in that: The method for obtaining the overall particle removal index includes: Calculate the absolute value of the difference in the number of each particle in the wafer image to be cleaned and the wafer image after cleaning in the historical cleaning process, calculate the ratio of the absolute value of the difference in the number of each particle to the number of particles in the wafer image to be cleaned, and obtain the cleaning effect value of each particle; in the particle cleaning category, calculate the mean of the cleaning effect values of all types of particles in all the historical cleaning processes to obtain the overall particle removal index of the particle cleaning category.
7. The wafer processing equipment for semiconductor automated production according to claim 1, characterized in that: Methods for cleaning control of the cleaning process to be treated include: The cleaning category corresponding to the largest overall matching degree is taken as the best matching category, and the best cleaning process parameters of the best matching category are set as the cleaning process parameters of the cleaning process to be processed.
8. The wafer processing equipment for semiconductor automated production according to claim 1, characterized in that: The method for obtaining the damage degree includes: A neural network is used to perform damage assessment on the damage of the cleaned wafer image of each historical cleaning process to obtain the damage degree of each historical cleaning process.
