Method, device and equipment for analyzing cause of scratch on surface of wafer and storage medium

Abstract

The invention relates to the technical field of wafer detection, and discloses a method, a device, equipment and a storage medium for analyzing a cause of a scratch on the surface of a wafer, and the method comprises the following steps: acquiring a wafer with a scratch on the surface and output parameters of a chemical mechanical polishing machine corresponding to the wafer with the scratch on the surface; marking the scratch generation reason of the wafer with the scratch on the surface to obtain the label information of the wafer; constructing a wafer scratch data set based on the wafer, the output parameters of the chemical mechanical polishing machine and the label information; and a wafer scratch generation reason classification model is constructed, the wafer scratch generation reason classification model is trained based on the wafer scratch data set, and after training of the wafer scratch generation reason classification model is completed, wafer surface scratch generation reasons are analyzed based on the wafer scratch generation reason classification model. Through the trained wafer scratch generation reason classification model, the wafer scratch generation reason can be quickly and accurately obtained.

Description

Technical Field

[0001] The present invention relates to the field of wafer detection technology, and in particular to a method, a device, a equipment and a storage medium for analyzing causes of scratches on a wafer surface. Background Art

[0002] Chemical Mechanical Polish (CMP) is a key step in the semiconductor manufacturing process, used to flatten the surface of silicon wafers and reduce the height difference between different material layers to the nanometer scale. However, the introduction of scratches during the CMP process may have an adverse effect on chip quality. There are many reasons for scratches, including particles on the surface of the polishing pad, abrasive particles in the chemical liquid, diamond particles in the dresser, aging of the brush, improper process control, or the material of the wafer. The scratch analysis in related technologies mainly uses some empirical analysis, which is inefficient and cannot accurately locate the cause. Summary of the invention

[0003] In view of this, the present invention provides a method, an apparatus, a device and a storage medium for analyzing the causes of scratches on the wafer surface, so as to solve the problem of low efficiency in analyzing the causes of scratches on the wafer surface through experience.

[0004] In a first aspect, the present invention provides a method for analyzing causes of scratches on a wafer surface, the method comprising: obtaining a wafer with scratches on its surface and output parameters of a chemical mechanical polishing machine corresponding to the wafer with scratches on its surface; marking the causes of the scratches on the wafer with scratches on its surface to obtain label information of the wafer; constructing a wafer scratch data set based on the wafer, the output parameters of the chemical mechanical polishing machine and the label information; constructing a wafer scratch cause classification model, training the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model is trained, analyzing the causes of scratches on the wafer surface based on the wafer scratch cause classification model.

[0005] In an optional embodiment, the obtaining of the chemical mechanical polishing machine output parameters corresponding to the wafer with scratches on the surface includes: searching and recording the corresponding chemical mechanical polishing machine output parameters based on the wafer; the chemical mechanical polishing machine output parameters include at least one of the rotation speed of the grinding pad, the torque of the grinding pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the grinding head, the rotation speed and torque of the polishing disk, the rotation speed and torque of the polishing head, the rotation speed of the brush and the torque of the brush.

[0006] In an optional embodiment, the label information includes at least one of polishing pad particles, dresser diamond particles, polishing liquid particles, brush cleaning and wafer material.

[0007] In an optional implementation, the step of constructing a wafer scratch cause classification model and training the wafer scratch cause classification model based on the wafer scratch data set includes: calculating a loss function L by the following formula:

[0008]

[0009] Wherein, M is the number of wafers with scratches on the surface, C is the number of wafer labels, c is the serial number of the wafer label, and p m,c is the cth element in the output result obtained by the wafer scratch cause classification model for the mth wafer with scratches on the surface, y m,c The wafer with scratches on the surface is the value of the cth element; by minimizing the loss function, the weight of the characteristic value of the output parameter of the chemical mechanical polishing machine is adjusted during the training process of the classification model of the causes of wafer scratches.

[0010] In an optional implementation, the minimizing the loss function includes minimizing the loss function by the following formula:

[0011]

[0012] r(t)=β1r(t-1)+(1-β1)g(t)

[0013] v(t)=β2v(t-1)+(1-β2)g(t) 2

[0014]

[0015] Wherein, g(t) is the partial derivative of the loss function L with respect to θ, r(t) is the first-order moment estimate of the gradient in momentum form at time t, β1 is the exponential decay rate of the first-order moment estimate, r(t-1) is the first-order moment estimate of the gradient in momentum form at time t-1, v(t) is the second-order moment estimate of the gradient in momentum form at time t, v(t-1) is the second-order moment estimate of the gradient in momentum form at time t-1, β2 is the exponential decay rate of the second-order moment estimate, is the bias-corrected first-order moment estimate, is the bias-corrected second-order moment estimate, δ is the learning rate, and ε is a constant.

[0016] In an optional embodiment, the step of constructing a wafer scratch cause classification model and training the wafer scratch cause classification model based on the wafer scratch data set further includes: calculating a classification function P by the following formula:

[0017]

[0018] Among them, e h ( j ) is the wafer x with scratches on the surface, and the index value of the scratch cause being the jth category is predicted by the wafer scratch cause classification model; based on the classification function P, the index values ​​of all categories are normalized to predict the exponential normalized probability that the scratch cause of the wafer x with scratches on the surface is the jth category.

[0019] In a second aspect, the present invention provides a device for analyzing causes of scratches on a wafer surface, the device comprising: an acquisition module for acquiring a wafer with scratches on its surface and output parameters of a chemical mechanical polishing machine corresponding to the wafer with scratches on its surface; a labeling module for labeling the causes of scratches on the wafer with scratches on its surface to obtain label information of the wafer; a first construction module for constructing a wafer scratch data set based on the wafer, the output parameters of the chemical mechanical polishing machine and the label information; a second construction module for constructing a wafer scratch cause classification model, training the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model is trained, analyzing the causes of scratches on the wafer surface based on the wafer scratch cause classification model.

[0020] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the method for analyzing causes of scratches on the wafer surface according to the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for analyzing causes of scratches on a wafer surface according to the first aspect or any corresponding embodiment thereof.

[0022] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for analyzing causes of scratches on a wafer surface according to the first aspect or any corresponding embodiment thereof.

[0023] By training the classification model of the causes of wafer scratches, the output parameters of the chemical mechanical polishing machine with scratches in the history are used as the training sample data set for supervised learning. With the increase of training sample data, the learning ability of the classification model of the causes of wafer scratches is continuously enhanced, so that the classification model of the causes of wafer scratches can learn the relationship between the output parameters of the chemical mechanical polishing machine and the causes of wafer scratches. The trained classification model of the causes of wafer scratches can quickly and accurately obtain the causes of wafer scratches. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are 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.

[0025] Figure 1 A schematic flow chart of a method for analyzing causes of scratches on a wafer surface according to an embodiment of the present invention is shown;

[0026] Figure 2 A schematic flow chart of another method for analyzing causes of scratches on a wafer surface provided by an embodiment of the present invention is shown;

[0027] Figure 3 A schematic diagram of the structure of a device for analyzing causes of scratches on a wafer surface provided by an embodiment of the present invention is shown;

[0028] Figure 4 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0030] During the CMP process, a variety of factors can cause scratches on the wafer surface. If the abrasive particles are hard, they may scratch the surface, resulting in scratches; if there is a certain degree of chemical reaction or adhesion between the abrasive and the material, scratches may occur; if the lubricant is used improperly or is missing, surface friction will increase, which may introduce scratches; the surface of the silicon wafer may be contaminated by tiny impurities and particles from the environment or equipment. These tiny particles may become sharper during mechanical polishing, causing scratches; improper process control may lead to uneven distribution of mechanical force, thereby introducing scratches; the selection and stacking of different material layers may lead to differences in hardness, and harder material layers may exert greater pressure on softer layers, resulting in scratches. There are many specific ways to improve the polishing process in the future, including optimizing the selection of abrasives, improving the formulation of lubricants, or strictly controlling process parameters.

[0031] The scratch analysis in related technologies mainly relies on personal experience and judgment, which takes a long time to find the cause and has a high error rate.

[0032] Therefore, when scratches are generated on the wafer, if the cause of the scratches can be located as quickly as possible, corresponding improvements can be made as soon as possible to improve the yield rate of the wafer.

[0033] According to an embodiment of the present invention, a method embodiment for analyzing the causes of scratches on the surface of a wafer is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0034] In this embodiment, a method for analyzing the causes of scratches on a wafer surface is provided, which can be used in a terminal with an operating system, such as a mobile phone, a desktop computer, a laptop computer, a tablet computer or a server, etc. Figure 1 FIG. 1 is a flow chart showing a method for analyzing the causes of scratches on a wafer surface according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0035] Step S101, obtaining a wafer with scratches on the surface and a chemical mechanical polishing machine output parameter corresponding to the wafer with scratches on the surface.

[0036] In this step, a wafer that has undergone a chemical mechanical polishing process is selected. The wafer surface can be detected by an optical microscope, an electron microscope or other high-precision detection equipment to identify and record scratches on the wafer surface, including information such as the location, shape and size of the scratches. For wafers with scratches detected, the corresponding CMP machine output parameters are searched and recorded. The CMP machine output parameters may include but are not limited to the following parameters: the rotation speed of the grinding pad, the torque of the grinding pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the grinding head, the rotation speed and torque of the polishing disc, the rotation speed and torque of the polishing head, the rotation speed of the brush and the torque of the brush, etc.

[0037] Step S102 , marking the causes of scratches on the wafer with scratches on the surface to obtain label information of the wafer.

[0038] The generation of wafer scratches is affected by many factors, including but not limited to: grinding pad particles, dresser diamond particles, wafer material, polishing liquid particles or cleaning brushes. In the process of the aforementioned factors causing scratches on the wafer, the actual output parameters of the CMP machine will change. For example, dresser diamond particles cause wafer scratches, and the dresser torque parameters and dresser speed parameters output by the CMP machine will fluctuate; grinding pad particles cause wafer scratches, and the grinding pad speed parameters, grinding pad torque parameters and grinding head speed parameters output by the CMP machine will fluctuate; polishing liquid particles cause wafer scratches, and the polishing disk speed parameters, polishing disk torque parameters, polishing head speed parameters and polishing head torque parameters output by the CMP machine will fluctuate; brush cleaning causes wafer scratches, and the brush speed parameters and brush torque parameters output by the CMP machine will fluctuate.

[0039] Step S103, constructing a wafer scratch data set based on the wafer, chemical mechanical polishing machine output parameters and label information.

[0040] In this step, the information of each wafer with scratches can be converted into a feature vector, which contains the CMP machine output parameters as features and the label information as the target variable. A vector with n-dimensional features can be constructed based on the wafer with scratches on the surface (Xi), which can be expressed as Xi = [x1 x2 x3…xn], where i represents the i-th sampling point, that is, the i-th wafer with scratches. A matrix with m rows and n columns can be constructed based on m wafers with scratches, that is, m observation points, as shown below:

[0041]

[0042] The output value of the m observation points is Yo, which is an m-dimensional vector, Yo = [y1 y2…ym], where Yo represents the labels of the m wafer samples, that is, the cause of scratches on each wafer.

[0043] Analyze the relationship between the output parameters of the CMP machine and the causes of scratches, and select features that are useful for the wafer scratch cause classification model, such as the rotation speed of the polishing pad, the torque of the polishing pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the polishing head, the rotation speed and torque of the polishing disk, the rotation speed and torque of the polishing head, the rotation speed of the brush and the torque of the brush, etc.

[0044] The wafer scratch dataset can be divided into a training set and a test set based on a preset ratio.

[0045] Step S104, constructing a wafer scratch cause classification model, training the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model is trained, analyzing the causes of scratches on the wafer surface based on the wafer scratch cause classification model.

[0046] In this step, a classification model for the causes of wafer scratches can be constructed based on the residual convolutional neural network architecture. Specifically, the ResNet-50 network structure can be selected. The ResNet-50 network structure includes: an input layer, an initial convolution layer, a residual block, a global average pooling layer, a fully connected layer, and a normalization (softmax) layer. The input layer is used to obtain the input feature vector. The initial convolution layer is used for feature extraction and to improve the stability of training. In the residual block, the structure of each block includes a convolution layer, a normalization layer for the convolution layer, and an activation layer; the jump connection directly adds the input of the residual block to the output of the residual block to ensure that the gradient can flow in back propagation. The global average pooling layer is used to generate a global feature vector. The fully connected layer is used to map the global feature vector to the number of categories. The normalization layer is used to calculate the probability of each category and is used for the final classification decision.

[0047] A wafer scratch cause classification model is trained based on a wafer scratch data set. After the wafer scratch cause classification model is trained, the cause of scratches on the wafer surface is predicted based on the wafer scratch cause classification model.

[0048] The method for analyzing the causes of scratches on the wafer surface provided in this embodiment trains a wafer scratch cause classification model, and uses the output parameters of a chemical mechanical polishing machine with historical scratches as a training sample data set for supervised learning. As the training sample data increases, the learning ability of the wafer scratch cause classification model continues to increase, so that the wafer scratch cause classification model can learn the relationship between the output parameters of the chemical mechanical polishing machine and the causes of wafer scratches. The wafer scratch cause classification model that has been trained can quickly and accurately obtain the causes of wafer scratches. In the case that the wafer has scratches, the causes of the scratches can be located as soon as possible, and corresponding improvements can be made to improve the wafer yield.

[0049] In this embodiment, a method for analyzing the causes of scratches on a wafer surface is provided, which can be used in a terminal with an operating system, such as a mobile phone, a desktop computer, a laptop computer, a tablet computer or a server, etc. Figure 2 FIG. 1 is a flow chart of another method for analyzing the causes of scratches on a wafer surface provided by an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0050] Step S201, obtaining a wafer with scratches on the surface and a chemical mechanical polishing machine output parameter corresponding to the wafer with scratches on the surface.

[0051] Specifically, the above step S201 includes:

[0052] Step S2011, searching and recording corresponding chemical mechanical polishing machine output parameters based on the wafer.

[0053] The wafer and the corresponding CMP output parameter may be associated based on the wafer identifier or the wafer production batch number. This may be implemented based on a database query, data merging or data connection operation.

[0054] In this way, the accuracy and completeness of the data can be guaranteed, facilitating the subsequent use of the model.

[0055] In step S2012, the output parameters of the chemical mechanical polishing machine include at least one of the rotation speed of the polishing pad, the torque of the polishing pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the polishing head, the rotation speed and torque of the polishing disc, the rotation speed and torque of the polishing head, the rotation speed of the brush, and the torque of the brush.

[0056] If the rotation speed of the grinding pad is too fast or too slow, or the torque of the grinding pad is unstable, the wafer surface may receive uneven pressure, resulting in scratches. The dresser is used to maintain the surface state of the grinding pad. If the speed or torque of the dresser is not appropriate, the surface of the grinding pad may become uneven, resulting in scratches on the wafer surface. If the rotation speed of the grinding head is too fast or too slow, it may affect the grinding effect and cause damage to the wafer surface. The rotation speed and torque of the polishing disc have a direct impact on the polishing effect. If the parameters are not set properly, the wafer surface may be unevenly polished, resulting in scratches. The rotation speed and torque of the polishing head also have an important impact on the polishing effect. Improper parameter settings may cause uneven pressure distribution during polishing, resulting in scratches on the wafer. In some CMP processes, brushes are used for cleaning or auxiliary polishing. If the rotation speed or torque of the brush is too large, it may cause damage to the wafer surface.

[0057] Step S202, marking the cause of the scratches on the wafer with scratches on the surface, and obtaining the label information of the wafer. Figure 1Step S202 of the illustrated embodiment will not be described in detail here.

[0058] Step S203: construct a wafer scratch dataset based on the wafer, chemical mechanical polishing machine output parameters and label information. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0059] Step S204, construct a wafer scratch cause classification model, train the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model training is completed, analyze the causes of wafer surface scratches based on the wafer scratch cause classification model. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0060] The method for analyzing the cause of wafer surface scratches provided in this embodiment can accurately locate the cause of wafer scratches, facilitate optimization of subsequent CMP processes, reduce the generation of wafer scratches, and improve the quality of chemical mechanical polishing.

[0061] In some optional embodiments, the label information includes at least one of polishing pad particles, dresser diamond particles, polishing liquid particles, brush cleaning, and wafer material.

[0062] In this embodiment, wafer label information is marked according to the cause of wafer scratches. Wafer scratches may be caused by one factor or multiple factors.

[0063] In this way, the impact of various factors on wafer scratches can be fully considered.

[0064] In some optional implementations, a wafer scratch cause classification model is constructed, and the wafer scratch cause classification model is trained based on a wafer scratch data set, including: calculating a loss function L by the following formula:

[0065]

[0066] Where M is the number of wafers with scratches on the surface, C is the number of wafer labels, c is the number of wafer labels, and p is the number of wafer labels. m,c is the cth element in the output result of the wafer scratch cause classification model for the mth wafer with scratches on the surface, y m,c The wafer with scratches on the surface is the value of the cth element; by minimizing the loss function, the weight of the characteristic value of the output parameter of the chemical mechanical polishing machine is adjusted during the training process of the classification model of the causes of wafer scratches.

[0067] In this embodiment, y m,cThe value of label c that indicates that the wafer with scratches on the surface belongs to the label. For example, if there are three labels in total and the mth wafer belongs to the first category, then the y corresponding to the mth wafer is m,c =[1,0,0],y m,1 =1,y m,2 =0,y m,3 = 0. The loss function is used to measure the difference between the probability distribution of the wafer scratch cause classification model output and the probability distribution of the true label. By calculating the aforementioned difference, the scratch cause classification model is guided to learn the correct classification boundary. Based on the loss function, a loss value can be obtained. The loss value reflects the gap between the predicted value and the true value of the scratch cause classification model. During the training process, the scratch cause classification model continuously adjusts parameters to reduce the loss value, such as adjusting the weight of the CMP output parameter feature value, until the model converges and the scratch cause classification model training is completed.

[0068] The calculation method of the above loss function is simple, can be implemented using a standard database, and has good mathematical properties, which makes the training process of the scratch cause classification model more stable and efficient, can handle multi-category problems well, and can measure the classification error of each category, which is especially important for the multi-classification problem of the causes of wafer scratches. At the same time, the gradient of the loss function for the last layer of weights is no longer related to the derivative of the activation function, but is only proportional to the difference between the output value and the true value, which helps to accelerate the convergence speed of the scratch cause classification model.

[0069] In some optional embodiments, minimizing the loss function includes minimizing the loss function by the following formula:

[0070]

[0071] r(t)=β1r(t-1)+(1-β1)g(t)

[0072] v(t)=β2v(t-1)+(1-β2)g(t) 2

[0073]

[0074] Among them, g(t) is the partial derivative of the loss function L with respect to θ, r(t) is the first-order moment estimate of the gradient at time t in momentum form, β1 is the exponential decay rate of the first-order moment estimate, r(t-1) is the first-order moment estimate of the gradient at time t-1 in momentum form, v(t) is the second-order moment estimate of the gradient at time t in momentum form, v(t-1) is the second-order moment estimate of the gradient at time t-1 in momentum form, β2 is the exponential decay rate of the second-order moment estimate, is the bias-corrected first-order moment estimate, is the bias-corrected second-order moment estimate, δ is the learning rate, and ε is a constant.

[0075] The method for minimizing the loss function in the embodiment of the present invention is an optimization algorithm based on gradient descent. By adjusting the parameters of the scratch cause classification model to minimize the loss function, the training speed and performance of the model can be improved. The concept of momentum is introduced to make the parameter update smoother, and the accumulated information of the previous gradient is taken into account, which helps to accelerate the parameter update direction. At the same time, the concept of adaptive learning rate is introduced, that is, each parameter has a different learning rate. The size of the learning rate is adjusted according to the gradient of the parameter and the historical gradient information, so that the scaling of the amplitude of the parameter update is invariant, thereby improving the efficiency and accuracy of the scratch cause classification model training.

[0076] In some optional implementations, constructing a wafer scratch cause classification model, training the wafer scratch cause classification model based on a wafer scratch data set, and further comprising: calculating a classification function P by the following formula:

[0077]

[0078] Among them, e h(j) The wafer x with scratches on the surface is predicted by the wafer scratch cause classification model to obtain the index value of the scratch cause of the jth category, h(j) is the value predicted by the wafer scratch cause classification model for the mth wafer as the jth category, and the predicted data is exponentially normalized based on the classification function P to predict the exponential normalized probability that the scratch cause of the wafer x with scratches on the surface is the jth category. P(y=j) is the exponential normalized probability that the wafer scratch cause is j predicted by the wafer scratch cause classification model.

[0079] In this embodiment, a wafer x with scratches on the surface is given. The wafer x is predicted by the wafer scratch cause classification model to have an index value of e for the jth category of scratch causes. h(j) In order to convert this index value into probability, the index values ​​of all categories can be normalized. The denominator in the classification function P is the sum of the index values ​​of all categories, which can ensure that the sum of the probabilities of all categories is 1.

[0080] In this way, the scale differences between different features can be eliminated, so that the influence of different features on the wafer scratch cause classification model is relatively uniform, and some features can be prevented from having too much influence on the wafer scratch cause classification model while other features are ignored. This helps the model find the optimal solution faster during the training process, thereby accelerating convergence and improving the performance of the wafer scratch cause classification model.

[0081] In some optional implementations, the wafer scratch cause classification model can be verified by a training set, and an AUO_ROC curve can be used. The AUO_ROC curve is a curve of true positive rate versus false positive rate at different thresholds, reflecting the classification ability of the wafer scratch cause classification model at different threshold settings.

[0082] The AUO_ROC curve plots the true positive rate (TPR) against the false positive rate (FPR) by considering all possible thresholds, which can comprehensively evaluate the classification ability of the wafer scratch cause classification model under different threshold settings. The area under the curve (AUC) is used as a performance indicator to provide an objective quantitative indicator of the performance of the wafer scratch cause classification model. The larger the AUC, the better the performance of the wafer scratch cause classification model in distinguishing different categories.

[0083] In this embodiment, a device for analyzing the causes of scratches on the surface of a wafer is also provided. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0084] This embodiment provides a device for analyzing the causes of scratches on the surface of a wafer. Figure 3 FIG. 1 is a schematic diagram showing the structure of a device for analyzing the causes of scratches on a wafer surface provided by an embodiment of the present invention. Figure 3 As shown, including:

[0085] The acquisition module 301 is used to acquire a wafer with scratches on the surface and the output parameters of a chemical mechanical polishing machine corresponding to the wafer with scratches on the surface.

[0086] The marking module 302 is used to mark the causes of scratches on the surface of the wafer and obtain the label information of the wafer.

[0087] The first constructing module 303 is used to construct a wafer scratch data set based on the wafer, the chemical mechanical polishing machine output parameters and label information.

[0088] The second construction module 304 is used to construct a wafer scratch cause classification model, train the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model is trained, analyze the causes of scratches on the wafer surface based on the wafer scratch cause classification model.

[0089] In some optional implementations, the acquisition module 301 includes:

[0090] The first acquisition unit is used to search and record the corresponding chemical mechanical polishing machine output parameters based on the wafer.

[0091] The second acquisition unit is used for the output parameters of the chemical mechanical polishing machine, including at least one of the rotation speed of the grinding pad, the torque of the grinding pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the grinding head, the rotation speed and torque of the polishing disk, the rotation speed and torque of the polishing head, the rotation speed of the brush and the torque of the brush.

[0092] In some optional implementations, in the first building block 303, the label information includes at least one of polishing pad particles, dresser diamond particles, polishing liquid particles, brush cleaning, and wafer material.

[0093] In some optional implementations, the second building module 304 includes:

[0094] The second step is to construct the first unit, which is used to calculate the loss function L using the following formula:

[0095]

[0096] Where M is the number of wafers with scratches on the surface, C is the number of wafer labels, c is the number of wafer labels, and p is the number of wafer labels. m,c is the cth element in the output result of the wafer scratch cause classification model for the mth wafer with scratches on the surface, y m,c The wafer with scratches on the surface is the value of the cth element; by minimizing the loss function, the weight of the characteristic value of the output parameter of the chemical mechanical polishing machine is adjusted during the training process of the classification model of the causes of wafer scratches.

[0097] In some optional embodiments, the second construction first unit includes:

[0098] Second, construct the first subunit to minimize the loss function using the following formula:

[0099]

[0100] r(t)=β1r(t-1)+(1-β1)g(t)

[0101] v(t)=β2v(t-1)+(1-β2)g(t) 2

[0102]

[0103] Among them, g(t) is the partial derivative of the loss function L with respect to θ, r(t) is the first-order moment estimate of the gradient at time t in momentum form, β1 is the exponential decay rate of the first-order moment estimate, r(t-1) is the first-order moment estimate of the gradient at time t-1 in momentum form, v(t) is the second-order moment estimate of the gradient at time t in momentum form, v(t-1) is the second-order moment estimate of the gradient at time t-1 in momentum form, β2 is the exponential decay rate of the second-order moment estimate, is the bias-corrected first-order moment estimate, is the bias-corrected second-order moment estimate, δ is the learning rate, and ε is a constant.

[0104] In some optional implementations, the second building module 204 further includes:

[0105] The second unit is constructed to calculate the classification function P by the following formula:

[0106]

[0107] Among them, e h(j) The wafer x with scratches on the surface is predicted by the wafer scratch cause classification model to obtain the index value of the scratch cause of the jth category, h(j) is the value predicted by the wafer scratch cause classification model for the mth wafer as the jth category, and the predicted data is exponentially normalized based on the classification function P to predict the exponential normalized probability that the scratch cause of the wafer x with scratches on the surface is the jth category. P(y=j) is the exponential normalized probability that the wafer scratch cause is j predicted by the wafer scratch cause classification model.

[0108] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0109] The device for analyzing the causes of scratches on the wafer surface in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0110] The device for analyzing the causes of scratches on the surface of a wafer provided by an embodiment of the present invention trains a classification model for causes of wafer scratches, and uses the output parameters of a chemical mechanical polishing machine with historical scratches as a training sample data set for supervised learning. As the training sample data increases, the learning ability of the classification model for causes of wafer scratches is continuously enhanced, so that the classification model for causes of wafer scratches can learn the relationship between the output parameters of the chemical mechanical polishing machine and the causes of wafer scratches. The trained classification model for causes of wafer scratches can quickly and accurately obtain the causes of wafer scratches.

[0111] The embodiment of the present invention also provides a computer device having the above Figure 3 The device shown is used to analyze the causes of scratches on the wafer surface.

[0112] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0113] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0114] The aforementioned memory 20 stores instructions executable by at least one processor 10, so that the aforementioned at least one processor 10 executes the method shown in the above embodiment.

[0115] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0116] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0117] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0118] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0119] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0120] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0121] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for analyzing the causes of scratches on a wafer surface, characterized in that: The method comprises: Acquire a wafer with scratches on the surface and a chemical mechanical polishing machine output parameter corresponding to the wafer with scratches on the surface; Marking the causes of the scratches on the wafer with scratches on the surface to obtain label information of the wafer; Based on the wafer, the output parameters of the chemical mechanical polishing machine and the label information, construct a wafer scratch data set; A wafer scratch cause classification model is constructed, and the wafer scratch cause classification model is trained based on the wafer scratch data set. After the wafer scratch cause classification model is trained, the causes of scratches on the wafer surface are analyzed based on the wafer scratch cause classification model.

2. The method according to claim 1, characterized in that The step of obtaining the chemical mechanical polishing machine output parameters corresponding to the wafer having scratches on the surface includes: Find and record the corresponding chemical mechanical polishing machine output parameters based on the wafer; The output parameters of the chemical mechanical polishing machine include at least one of the rotation speed of the grinding pad, the torque of the grinding pad, the rotation speed of the dresser, the torque of the dresser, the rotation speed of the grinding head, the rotation speed and torque of the polishing disk, the rotation speed and torque of the polishing head, the rotation speed of the brush and the torque of the brush.

3. The method according to claim 1 or 2, characterized in that: The label information includes at least one of polishing pad particles, dresser diamond particles, polishing liquid particles, brush cleaning and wafer material.

4. The method according to claim 1, characterized in that: The step of constructing a wafer scratch cause classification model and training the wafer scratch cause classification model based on the wafer scratch data set includes: The loss function L is calculated by the following formula: Wherein, M is the number of wafers with scratches on the surface, C is the number of wafer labels, c is the serial number of the wafer label, and p m,c is the probability of the cth element in the output result obtained by the wafer scratch cause classification model for the mth wafer with scratches on the surface, y m,c The wafer with scratches on the surface is the value of the cth element; By minimizing the loss function, the weight of the characteristic value of the output parameter of the chemical mechanical polishing machine is adjusted during the training process of the wafer scratch cause classification model.

5. The method according to claim 4, characterized in that The minimizing the loss function includes minimizing the loss function by the following formula: r(t)=β1r(t-1)+(1-β1)g(t) v(t)=β2v(t-1)+(1-β2)g(t) 2 Wherein, g(t) is the partial derivative of the loss function L with respect to θ, r(t) is the first-order moment estimate of the gradient in momentum form at time t, β1 is the exponential decay rate of the first-order moment estimate, r(t-1) is the first-order moment estimate of the gradient in momentum form at time t-1, v(t) is the second-order moment estimate of the gradient in momentum form at time t, v(t-1) is the second-order moment estimate of the gradient in momentum form at time t-1, β2 is the exponential decay rate of the second-order moment estimate, is the bias-corrected first-order moment estimate, is the bias-corrected second-order moment estimate, δ is the learning rate, and ε is a constant.

6. The method according to claim 1, characterized in that The step of constructing a wafer scratch cause classification model and training the wafer scratch cause classification model based on the wafer scratch data set further includes: The classification function P is calculated by the following formula: Among them, e h(j) For the wafer x with scratches on the surface, the wafer scratch cause classification model is used to predict the index value of the scratch cause being the jth category; The index values ​​of all categories are normalized based on the classification function P, and the index normalized probability that the scratch cause of the wafer x with scratches on the surface is the jth category is predicted.

7. A device for analyzing the causes of scratches on a wafer surface, characterized in that: The device comprises: An acquisition module, used for acquiring a wafer with scratches on the surface and a chemical mechanical polishing machine output parameter corresponding to the wafer with scratches on the surface; A marking module, used for marking the causes of scratches on the wafer with scratches on the surface, and obtaining label information of the wafer; A first construction module is used to construct a wafer scratch data set based on the wafer, the chemical mechanical polishing machine output parameters and the label information; The second construction module is used to construct a wafer scratch cause classification model, train the wafer scratch cause classification model based on the wafer scratch data set, and after the wafer scratch cause classification model is trained, analyze the causes of scratches on the wafer surface based on the wafer scratch cause classification model.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for analyzing the causes of scratches on the wafer surface as described in any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for analyzing causes of scratches on a wafer surface according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to cause a computer to execute the method for analyzing causes of scratches on a wafer surface according to any one of claims 1 to 6.

Images