Prediction method for imaging effect of mask plate defect on silicon wafer layer, product and equipment

Abstract

The invention provides a method for predicting the imaging effect of a mask defect on a silicon wafer layer, a product and equipment. The method comprises the following steps: acquiring a target mask plate graph, wherein the target mask plate graph is a layout with potential defects; generating a target aerial image of the target mask plate graph through an aerial image measurement system; analyzing and processing the target aerial image to obtain a target optical contour; and generating a target silicon wafer imaging contour corresponding to the target optical contour according to a pre-trained artificial intelligence prediction model. Through the method, the silicon wafer imaging contour of the target mask plate graph after the imaging process can be obtained according to the target aerial image, so that the physical and chemical characteristics in the imaging process are fully considered, the accurate prediction of the silicon wafer imaging contour is realized, the defect printability evaluation precision is greatly improved, and the accuracy of defect printability evaluation is improved. Therefore, the requirement of an advanced process on prediction precision is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor technology, in particular to a mask defect imaging effect prediction method, product and equipment. BACKGROUND

[0002] In the current field of semiconductor technology, lithography and etching and other patterning processes are the core link of transferring the design pattern to the silicon wafer. The surface defects of the mask as the pattern carrier are easy to form printable defects on the silicon wafer, which directly affects the chip yield and function. The aerial image measurement system is a key equipment for evaluating the printability of mask defects. By simulating the optical environment of the lithography machine to obtain the aerial image, the imaging effect of the defect can be predicted without actual exposure, which becomes the core decision basis for defect detection and repair.

[0003] However, the aerial image measurement system essentially outputs the aerial image before lithography. However, the final performance of the mask defect on the silicon wafer is not only determined by the optical image, but also affected by multiple process effects such as chemical diffusion of photoresist, development behavior in the subsequent lithography process, and material removal deviation in the etching process. The existing method usually only performs simple threshold segmentation on the aerial image to obtain the contour, and uses it as the basis for silicon wafer defect prediction. This process cannot cover the pattern deformation introduced by lithography and etching processes, resulting in insufficient prediction accuracy. SUMMARY

[0004] In view of the above problems, the present application provides a mask defect imaging effect prediction method, product and equipment to overcome the above problems or at least partially solve the above problems.

[0005] One object of the present application is to realize accurate prediction of the silicon wafer imaging contour after lithography or etching of the potential defects of the mask; Another further object of the present application is to avoid overfitting of the trained artificial intelligence prediction model.

[0006] In particular, the present application provides a mask defect imaging effect prediction method, comprising: obtaining a target mask layout, the target mask layout being a layout with potential defects; generating a target aerial image of the target mask layout by an aerial image measurement system; analyzing and processing the target aerial image to obtain a target optical contour; generating a target silicon wafer imaging contour corresponding to the target optical contour according to a pre-trained prediction model, the prediction model being used to learn the mapping relationship between the optical contour before the patterning process and the silicon wafer imaging contour after the patterning process, and output the corresponding silicon wafer imaging contour based on the input optical contour.

[0007] Optionally, the prediction model is an artificial intelligence prediction model. The training step of the artificial intelligence prediction model comprises: obtaining a pre-established patterning process model, the patterning process model being used to generate a contour of the reticle pattern after the reticle pattern undergoes a patterning process according to an input reticle pattern; generating simulation data according to the patterning process model, the simulation data being a data pair composed of an optical contour of the reticle pattern before the reticle pattern undergoes the patterning process and a silicon wafer imaging contour extracted from a scanning electron microscope image actually measured after the reticle pattern undergoes the patterning process; training the artificial intelligence prediction model according to the simulation data and actual measurement data, the actual measurement data being a data pair composed of an optical contour actually measured by an aerial image measurement system on the reticle pattern and a silicon wafer imaging contour extracted from a scanning electron microscope image actually measured after the reticle pattern undergoes the patterning process.

[0008] Optionally, the step of training the artificial intelligence prediction model according to the simulation data and the actual measurement data comprises: converting contour information recorded by both the simulation data and the actual measurement data into a signed distance field image; training the artificial intelligence prediction model according to the signed distance field image.

[0009] Optionally, the step of training the artificial intelligence prediction model according to the simulation data and the actual measurement data comprises: determining a weight value of the simulation data and the actual measurement data in the training process; inputting the simulation data and the actual measurement data into a preset neural network architecture; training the simulation data and the actual measurement data according to the weight value in the preset neural network architecture to obtain the artificial intelligence prediction model.

[0010] Optionally, the step of generating the simulation data according to the patterning process model comprises: obtaining complete layout data corresponding to a simulation target chip; grouping the complete layout data according to pattern features to obtain a plurality of pattern groups; selecting a target training pattern from the plurality of pattern groups; generating an optical contour and a silicon wafer imaging contour corresponding to the target training pattern by using the patterning process model, thereby obtaining the simulation data.

[0011] Optionally, the step of obtaining the actual measurement data comprises: obtaining a measured reticle pattern and a measured position corresponding to a measured optical contour in the actual measurement data; determining an image acquisition range of the optical proximity correction simulation; An optical profile simulation model is built based on the measured mask pattern, and the optical parameters and mask parameters in the optical profile simulation model are calibrated; A simulation optical profile is simulated in the measured mask pattern by the optical profile simulation model, with the measured position as the center and according to the image acquisition range; The center position of the simulation optical profile is replaced by the measured optical profile, and the measured optical profile and the surrounding simulation optical profile are spliced to obtain an expanded optical profile; The expanded optical profile and a silicon wafer imaging profile extracted from a scanning electron microscope image actually measured after a patterning process of the measured mask pattern are combined to form actual measurement data.

[0012] Optionally, the patterning process model includes a photoetching model and / or an etching model, the photoetching model is used to generate a photoresist profile after photoetching of the mask pattern according to the input mask pattern, and the etching model is used to generate an etching profile after etching of the mask pattern according to the input mask pattern. For the photoetching model, the silicon wafer imaging profile is the photoresist profile. For the etching model, the silicon wafer imaging profile is the etching profile after etching.

[0013] According to another aspect of the present application, a computer readable storage medium is also provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the prediction method of the imaging effect of the mask pattern defects on the silicon wafer layer.

[0014] According to another aspect of the present application, a computer program product is also provided, which includes a computer program, and the computer program is executed by a processor to implement the steps of the prediction method of the imaging effect of the mask pattern defects on the silicon wafer layer.

[0015] According to another aspect of the present application, a computer device is also provided, which includes a memory, a processor, and a machine executable program stored in the memory and running on the processor, and the processor executes the machine executable program to implement the steps of the prediction method of the imaging effect of the mask pattern defects on the silicon wafer layer.

[0016] The method for predicting the imaging effect of mask plate defects on a silicon wafer layer comprises the following steps: after obtaining a target mask plate pattern, generating a target aerial image of the target mask plate pattern by an aerial image measurement system; then analyzing and processing the target aerial image to obtain a target optical profile; then generating a target silicon wafer imaging profile corresponding to the target optical profile according to a pre-trained artificial intelligence prediction model, the artificial intelligence prediction model being used to learn the mapping relationship between the optical profile before the graphic process and the silicon wafer imaging profile after the graphic process, and output the corresponding silicon wafer imaging profile based on the input optical profile; finally, performing defect analysis on the target mask plate pattern based on the target silicon wafer imaging profile, so as to determine whether the potential defects have an impact on the graphic process result. Through this method, the silicon wafer imaging profile after the graphic process of the target mask plate pattern can be obtained according to the target aerial image, so that the physical and chemical characteristics in the graphic process are fully considered, the silicon wafer imaging profile is accurately predicted, the defect printability evaluation precision is greatly improved, and the prediction accuracy requirement of advanced processes is met.

[0017] Further, the method for predicting the imaging effect of mask plate defects on a silicon wafer layer comprises the following steps: obtaining a pre-established graphic process model, the graphic process model being used to generate a profile of the mask plate pattern after the graphic process according to the input mask plate pattern; generating simulation data according to the graphic process model, the simulation data being a data pair composed of the optical profile before the graphic process of the mask plate pattern and the silicon wafer imaging profile after the graphic process; and training the artificial intelligence prediction model according to the simulation data and actual measurement data, the actual measurement data being a data pair composed of the optical profile actually measured by the aerial image measurement system on the mask plate pattern and the silicon wafer imaging profile extracted from the scanning electron microscope image actually measured after the graphic process of the mask plate pattern. Through this method, a large amount of simulation data can be generated with the aid of the graphic process model, the shortage of actual measurement data and the limitation of covered scenarios are made up, the breadth and diversity of training data are greatly expanded, sufficient learning samples are provided for the artificial intelligence prediction model, the model is trained by combining the simulation data and the actual measurement data, the model generalization ability is ensured by the simulation data, the actual process characteristics are anchored by the real measurement data, the model overfitting is avoided, and the prediction reliability is improved.

[0018] The above and other objects, advantages and features of the present application will become more apparent from the following detailed description of specific embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0019] Some specific embodiments of the present application will be described in detail in the following with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar components or parts. Those skilled in the art should understand that the drawings are not necessarily drawn to scale. In the drawings: Figure 1 is a schematic diagram of a method for predicting the imaging effect of a mask defect on a wafer level according to an embodiment of the present application; Figure 2 is a schematic diagram of a training process of an artificial intelligence prediction model in the method for predicting the imaging effect of a mask defect on a wafer level according to an embodiment of the present application; Figure 3 is a schematic diagram of an extension process of actual measurement data in the method for predicting the imaging effect of a mask defect on a wafer level according to an embodiment of the present application; Figure 4 is a schematic diagram of a computer program product according to an embodiment of the present application; Figure 5 is a schematic diagram of a computer readable storage medium according to an embodiment of the present application; and Figure 6 is a schematic diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Those skilled in the art should understand that the embodiments described below are only a part of the embodiments of the present application, rather than all the embodiments of the present application, and are intended to explain the technical principles of the present application, rather than limit the protection scope of the present application. Based on the embodiments provided by the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should fall within the protection scope of the present application.

[0021] It should be noted that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered listing of executable instructions for implementing logical functions, and can be specifically embodied in any computer readable medium for use by an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or in conjunction with these instruction execution systems, apparatus or devices.

[0022] The most core step in semiconductor chip manufacturing is to transfer the design pattern of the chip to the silicon wafer. Among the many process steps in chip manufacturing, the processes directly related to pattern transfer are mainly lithography and etching, commonly known as (Patterning). In the current mainstream integrated circuit production process, lithography technology is to change the photoresist under special wavelength light exposure, such as 193nm or 248nm, and then develop the pattern designed on the mask into the photoresist topography on the silicon wafer. Etching process is to selectively remove unwanted material based on the help of photoresist topography, so as to finally create the required fine pattern on the silicon wafer. The starting point of these patterning processes is the mask, which is the core pattern carrier of the patterning process. The small defects (such as particles, crystal growth, electrostatic damage, etc.) on the surface of the mask can form printability defects on the silicon wafer, which can directly cause chip functional failure or yield reduction. Therefore, in industry practice, the requirement for the mask is zero defect.

[0023] In the current semiconductor technology field, aerial image measurement system has become a key device for evaluating the imaging performance and defect printability of the mask. The aerial image measurement system directly obtains the aerial image of the mask pattern by accurately simulating the optical conditions of the lithography scanner (including wavelength, numerical aperture, illumination mode and polarization state). The aerial image is optically equivalent to the potential image projected onto the photoresist layer of the silicon wafer in the lithography machine, so it can be used to predict the printing effect of the defects or patterns on the silicon wafer from the mask, and then guide the mask qualification judgment and defect repair decision.

[0024] However, the aerial image measurement system essentially outputs the aerial image before lithography. The final performance of the mask defect on the silicon wafer depends not only on the optical image, but also on the comprehensive effects of multiple processes such as chemical diffusion of photoresist, development behavior in the subsequent lithography process, and material removal bias in the etching process. The common method in the current semiconductor technology field is to perform simple threshold segmentation on the aerial image to obtain the contour, and use it as the basis for silicon wafer defect prediction. This process cannot cover the pattern distortion introduced by lithography and etching processes, resulting in insufficient prediction accuracy.

[0025] Of course, the development of multi-parameter photoresist model calibrated based on silicon wafer measurement data is relatively mature, and has been widely used in the industry practice of optical proximity correction (OPC). Ideally, the photoresist profile after lithography can be simulated based on aerial image through multi-parameter photoresist calibration model, but this requires accurate understanding of the photoresist film stack and the process used by the wafer factory, so it is difficult to obtain the corresponding photoresist calibration model. The aerial image measurement system is mainly used for mask defect evaluation and does not involve subsequent lithography process, so it is difficult to apply multi-parameter photoresist calibration model to the aerial image measurement system. Therefore, at present, the aerial image measurement system in practice is still based on simple threshold segmentation to obtain the profile based on the aerial image, and the profile is used as the basis for wafer defect prediction. This makes the process ignore the pattern deformation caused by lithography and etching, resulting in insufficient prediction accuracy.

[0026] To solve the above problems, the present application provides a method for predicting the imaging effect of mask defects on the wafer level, Figure 1 The present application is a flowchart of a method for predicting the imaging effect of mask defects on the wafer level according to an embodiment of the present application. In this embodiment, the method for predicting the imaging effect of mask defects on the wafer level includes at least the following steps S101 to S104.

[0027] Step S101, obtain the target mask pattern. The target mask pattern is the core pattern carrier of the chip patterning process. The surface of the mask may have potential defects such as particles, crystal growth, and electrostatic damage, so it is necessary to evaluate the printability of the mask pattern by the aerial image measurement system to determine whether it needs to be repaired.

[0028] Step S102, generating a target aerial image of the target mask by an aerial image measurement system. The aerial image measurement system is the core execution subject of this step, and its core advantage is that it can accurately replicate the optical imaging environment of the lithography machine. In some optional embodiments, the aerial image measurement system generally loads the key parameters of the target lithography process, including the wavelength, numerical aperture (NA), illumination mode (such as annular, quadrupole off-axis illumination), partial coherence coefficient, and polarization state (TE wave, TM wave) of the lithography machine, etc., to ensure that the optical environment of the system is completely consistent with the actual lithography process. Then the aerial image measurement system will use the same illumination system as the lithography process to illuminate the defect area of the target mask, and the light beam is modulated by the mask pattern and then diffracted, and then collected by a high-resolution objective lens to form a target aerial image that is optically equivalent to the latent image on the photoresist of the silicon wafer. Unlike the lithography machine, the aerial image measurement system projects the aerial image onto a specific camera device after magnification by the imaging lens, achieving accurate collection of the aerial image without the need for actual silicon exposure.

[0029] Step S103, analyzing and processing the target aerial image to obtain a target optical contour. The target aerial image is essentially an optical intensity distribution image, which needs to be converted into a target optical contour (Aerial Contour) that can be used as input for an artificial intelligence (Artificial Intelligence, AI for short) model through a specific analysis and processing procedure. In one optional way, the specific processing process includes: first, pre-processing the collected target aerial image, including noise removal, image normalization, etc., to improve image quality; then segmenting the pre-processed aerial image based on a pre-set threshold, which needs to be calibrated according to factors such as target lithography process parameters and photoresist sensitivity, to ensure that the segmentation result conforms to the actual optical imaging law; finally, extracting the target optical contour from the segmented image, which accurately reflects the geometric shape of the aerial image and is a key intermediate carrier linking mask defects and silicon imaging results.

[0030] Step S104, generating a target silicon wafer imaging contour corresponding to the target optical contour according to a pre-trained prediction model. The prediction model is used to learn the mapping relationship between the optical contour before the patterning process and the silicon wafer imaging contour after the patterning process, and outputs the corresponding silicon wafer imaging contour based on the input optical contour.

[0031] In some optional embodiments, the prediction model can generally be an artificial intelligence prediction model. The training step of the artificial intelligence prediction model can generally include: obtaining a pre-established patterning process model; generating simulation data according to the patterning process model, the simulation data being a data pair composed of an optical profile before the mask pattern is subjected to the patterning process and a silicon wafer imaging profile after the mask pattern is subjected to the patterning process; and training the artificial intelligence prediction model according to the simulation data and actual measurement data, the actual measurement data being a data pair composed of an optical profile actually measured by an aerial image measurement system on the mask pattern and a silicon wafer imaging profile extracted from a scanning electron microscope image actually measured after the mask pattern is subjected to the patterning process.

[0032] The patterning process model is generally used to generate a profile of the mask pattern after the mask pattern is subjected to the patterning process according to an input mask pattern. Since the patterning process generally includes photolithography and etching, the model can generally include two types of models: one is a photolithography model, which is constructed based on the core algorithm of optical proximity correction (OPC), integrates parameters related to the photolithography process, and can accurately output a three-dimensional profile of a photoresist after development according to input mask pattern data; and the other is an etching model, which is constructed based on the physical and chemical mechanism of the etching process, includes parameters related to the etching process, and can output the final pattern profile of the silicon wafer dielectric layer after etching based on the input photoresist profile data. Therefore, if the patterning process is photolithography, the aforementioned silicon wafer imaging profile is a photoresist profile; and if the patterning process is etching, the aforementioned silicon wafer imaging profile is an etching profile.

[0033] In addition, considering that the actual measurement of the aerial image measurement system is the potential defect position of the mask pattern found in the detection, if an AI model is constructed only according to the data normally simulated from a large number of mask patterns, the coverage of such potential defect data is seriously insufficient, and the prediction ability of the AI model for the measurement image of the aerial image measurement system is extremely limited. The measurement data of the aerial image and the scanning electron microscope (Critical Dimension Scanning Electron Microscope, CD-SEM) is very limited, generally only a few hundred, and the range of the measurement is also very limited, usually within 750 nm, while the range considered by ordinary photolithography or etching is much larger. Therefore, if an AI model is constructed directly based on the measured optical profile and the corresponding CD-SEM measurement profile to predict the effect of the optical profile after photolithography or etching, the process can run through, but the model is very unreliable and is prone to overfitting.

[0034] Based on the above analysis, the simulation data is generated according to the patterning process model, and the artificial intelligence prediction model is trained by combining the simulation data and the actual measurement data.

[0035] Optionally, the step of generating simulation data can generally include: obtaining complete layout data corresponding to the simulation target chip; generating an optical profile corresponding to the target training pattern and a wafer imaging profile by using the patterning process model, thereby obtaining the simulation data. The optical profile can generally be directly based on the optical part in the patterning process model, and the optical profile can be calculated from the mask pattern data and obtained under a certain threshold. The essence of the optical profile is the same as that of the aerial image. The same mask pattern data is input into the patterning process model, and a lithography model or an etching model is selected according to the requirement, and the corresponding photoresist profile or etched profile is output, that is, the wafer imaging profile can be obtained. The optical profile corresponding to the same mask pattern and the wafer imaging profile are paired, and a complete set of simulation data can be formed. By repeating the above process, a large amount of simulation data set covering multiple scenarios and multiple pattern types can be generated to meet the training requirements of the AI model.

[0036] As can be seen from the above operation, the method can select the patterning process model according to the actual requirement, so as to obtain an artificial intelligence prediction model with different functions.

[0037] For example, in the case where the patterning process model selected is the lithography model, the pre-trained artificial intelligence prediction model learns the mapping relationship between the optical profile and the photoresist profile after lithography. Therefore, the photoresist profile corresponding to the target optical profile is output, so that the influencing factors in the lithography process are integrated, and the subsequent defect analysis can further consider the influence of the lithography process. For example, a small particle defect on the mask can cause a change in the exposure sensitivity of the surrounding photoresist during lithography, resulting in line width shrinkage or edge jagged deformation after development. These details that cannot be captured by the threshold segmentation of the aerial image can be reflected by the accurate prediction of the photoresist profile.

[0038] In the case where the patterning process model selected is the etching model, the pre-trained artificial intelligence prediction model learns the mapping relationship between the optical profile and the etched profile after etching. Therefore, the etched profile corresponding to the target optical profile is output, so that the influencing factors in the lithography and etching processes are integrated, and the subsequent defect analysis can completely cover the whole-link process effects of lithography and etching. For example, a potential defect of the mask can cause a small line width deviation of the photoresist profile, and the lateral erosion and selective etching deviation in the etching process can further amplify the defect influence, and finally form a substantial pattern defect on the wafer. These etching process superposition effects that cannot be accurately captured by the target optical profile can be completely presented by the prediction of the etched profile.

[0039] The core of this step is to rely on the pre-trained artificial intelligence prediction model to realize accurate mapping from the target optical profile to the target silicon wafer imaging profile, thereby covering multiple process effects such as photoresist chemical diffusion, development behavior, and etching material removal deviation, and making up for the limitations of current semiconductor field relying only on optical profile prediction.

[0040] Further, in order to more effectively generate simulation data, the method of the present application can also introduce a pattern grouping (Pattern Grouping) technique. For an actual chip full chip layout, the pattern grouping technique is applied, and similar patterns are classified into different groups according to the geometric characteristics (such as line width, spacing, shape structure, etc.) of the patterns, so as to filter out representative typical patterns from the vast amount of patterns. Then, using the reference patterned process model, simulation profiles are generated on these representative patterns and used as input for neural network training, thereby obtaining more effective simulation data for subsequent AI model training and improving the generalization ability of the AI model. The pattern grouping technique can generally include exact grouping (Exact Grouping) and fuzzy grouping (Fuzzy Grouping). Those skilled in the art can select the method for obtaining simulation data according to the actual situation, for example, using an exact grouping algorithm or a fuzzy grouping algorithm, or a common random sampling method.

[0041] Finally, after obtaining the target silicon wafer imaging profile corresponding to the target optical profile, the target mask layout can be analyzed for defects based on the target silicon wafer imaging profile, so as to determine whether the potential defects affect the patterning process result.

[0042] Specifically, the influence of the defects on the patterning process result is judged by comparing the target silicon wafer imaging profile with the design expected profile of the chip layout. If the analysis result shows that the target silicon wafer imaging profile meets the design requirements, it is determined that the potential defects do not affect the patterning process result, and the mask does not need to be repaired; if it does not meet the design requirements, it is determined that the defects have printability, and the mask needs to be sent to the repair process. After repair, the repair effect can be confirmed again by the method.

[0043] Through this method, the limitations of the original aerial image measurement system can be broken through, and the silicon wafer imaging profile after the target mask layout graph undergoes the patterning process can be obtained according to the target aerial image, so as to fully consider the physical and chemical characteristics in the patterning process. The whole framework can be applied not only to lithography but also to etching, making it possible to predict the silicon wafer imaging profile after the aerial image of the mask potential bad point in lithography and etching, realizing accurate prediction of the silicon wafer imaging profile, greatly improving the defect printability evaluation precision, and meeting the requirements of advanced processes for prediction accuracy.

[0044] Figure 2is a schematic diagram of a training process of an artificial intelligence prediction model in a method for predicting the imaging effect of a mask plate defect on a wafer level according to an embodiment of the present application, as shown in Figure 2 The training method of the artificial intelligence prediction model includes at least the following steps S201 to S208.

[0045] Step S201, a pre-established patterning process model is obtained. The patterning process model can be a photoresist model or an etching model. The photoresist model can convert the input mask plate pattern into a photoresist contour through operation, and the etching model can convert the input photoresist contour into an etched contour through operation. Of course, the construction of these models itself needs to be based on a large amount of actual wafer data in the wafer factory, and multiple rounds of calibration and optimization are required to obtain the model.

[0046] Step S202, the complete layout data corresponding to the simulation target chip is obtained. In order to ensure that the simulation data fits the actual application scenario, a chip matching the target process node is selected as the simulation object, and the target chip covers typical device structures (such as logic gates, memory cells, interconnection lines, etc.). Then, the complete layout data corresponding to the target chip is extracted, which usually includes the geometric dimensions, layout positions, and level information of all patterns in the chip, and is the basic data source for subsequent pattern grouping and training pattern selection.

[0047] Step S203, the complete layout data is grouped according to the pattern features, and a plurality of pattern groups are obtained. In the process of generating simulation data, some patterns in certain positions may be too simple or have no representative, therefore, in order to speed up the training efficiency and ensure the generalization ability of the artificial intelligence bad point detection model, the present application further introduces a pattern grouping technology in the process of generating simulation data. Finally, similar patterns are classified into different groups according to their geometric features (such as line width, spacing, shape structure, etc.), so that representative typical patterns are selected from a large number of patterns. The grouping algorithm (i.e. Pattern Grouping technology) generally includes Exact Grouping and Fuzzy Grouping algorithms, or random sampling method can be selected, and the corresponding grouping method can be selected by the person skilled in the art according to the actual demand.

[0048] Step S204, the target training pattern is selected from the plurality of pattern groups. This step generally selects representative patterns from each pattern group as the target training pattern, for example, typical pattern features covering each group, and patterns prone to process deviation. The selected target training pattern needs to consider diversity and typicality, which ensures that the model learns comprehensive process rules, and avoids excessive increase in training cost due to the selection of redundant patterns.

[0049] Step S205, generating the optical profile corresponding to the target training pattern and the wafer imaging profile by using the patterning process model, so as to obtain simulation data. The simulation data is a data pair composed of the optical profile before the mask pattern is subjected to the patterning process and the wafer imaging profile after the mask pattern is subjected to the patterning process.

[0050] The specific execution process of the present step can generally include: first, simulating the imaging logic of the aerial image measurement system by the optical part in the patterning process model for the selected target training pattern, to calculate the optical profile consistent with the real output; then outputting the wafer imaging profile corresponding to the target training pattern by the patterning process model; finally, pairing the optical profile and the wafer imaging profile corresponding to the same target training pattern to form a set of corresponding simulation data; repeating the above process, i.e. generating a large-scale simulation data set covering multiple pattern types.

[0051] Step S206, determining the weight value of the simulation data and the actual measurement data in the training process. Although the simulation data has the advantages of wide coverage and large quantity, it deviates from the real process; the actual measurement data directly reflects the characteristics of the real process, but the cost of obtaining it is high and the sample size is small. The present step balances the training contribution of the two types of data by setting the weight value, and the determination of the weight value is based on the data reliability and the sample size ratio: if the actual measurement data sample size is sufficient and the reliability is high, a higher weight can be set; if the actual measurement data sample size is small, the weight is appropriately reduced to avoid model overfitting. Reasonable setting of the weight value can make the model learn the generalization rule of the simulation data and the real characteristics of the measured data at the same time.

[0052] Step S207, inputting the simulation data and the actual measurement data into the preset neural network architecture.

[0053] Step S208, training the simulation data and the actual measurement data in the preset neural network architecture according to the weight value to obtain an artificial intelligence prediction model.

[0054] Optionally, the step of training the artificial intelligence prediction model according to the simulation data and the actual measurement data can generally include: converting the contour information recorded by both the simulation data and the actual measurement data into a signed distance field image; and training the artificial intelligence prediction model according to the signed distance field image. The signed distance field (SDF) is a mathematical function used to represent a geometric shape. For any point in space, the signed distance function returns the distance of the point to the surface of the nearest geometric shape, and gives the distance a positive or negative sign according to the position of the point relative to the shape. Generally, when the point is inside the geometric shape, the distance is negative; when the point is outside the geometric shape, the distance is positive; and when the point is exactly on the surface of the geometric shape, the distance value is 0. If the optical contour and the corresponding photoresist contour are both converted into a signed distance field image, an AI model can be constructed to complete the mapping from the input optical contour SDF image to the output photoresist contour SDF image. During specific reasoning, the AI model first converts the input contour into an input SDF image, obtains the output SDF image after reasoning, and then converts it into an output contour, thus completing the entire Contour to Contour process. The signed distance field image can more accurately represent the geometric shape and topological structure of the contour, and compared with the original binary contour image, it can provide more rich feature information for the model, effectively improving the prediction accuracy of the model for small defect contours.

[0055] The present embodiment relies on a closed-loop design of precise data preparation-scientific grouping sampling-weighted mixed training, and systematically solves the problems of insufficient model generalization ability and low prediction accuracy caused by single data training, thereby providing high-performance model support for subsequent mask defect wafer imaging effect prediction.

[0056] In some optional embodiments, the iterative training process of the artificial intelligence prediction model can generally include: the neural network first extracts the feature information of the optical profile through the convolution layer and the pooling layer, then restores the feature dimension through the up-sampling layer, and outputs the predicted wafer imaging profile; during the training process, the loss function (such as mean square error loss) of the two types of data is calculated according to the preset weight value, and the network parameters are continuously optimized through the back propagation algorithm; when the loss function converges to the preset threshold, and the prediction accuracy of the model on the validation set meets the requirements, the training is stopped, and the final artificial intelligence prediction model is obtained. The artificial intelligence prediction model trained in this way can learn the accurate mapping rules of the optical profile and the wafer imaging profile, such as: the basic correspondence between the optical profile geometry and the wafer imaging profile; the nonlinear deformation rules caused by lithography / etching process (optical proximity effect, etching load effect, etc.); the imaging specificity rules of mask defects (association between defect type, size and wafer imaging morphology); the fusion of ideal process rules of simulation data and real process deviation rules of measured data, etc.

[0057] It should be noted that the method adopts a neural network suitable for image-image profile mapping task as the core model, and the training core framework generally includes a feature extraction module, a feature restoration / mapping module, and a feature fusion module (if needed). Each module cooperates to adapt to the core requirement of optical profile feature input-wafer imaging profile feature output, and each type of neural network with profile mapping capability (such as convolutional neural network, encoding-decoding structure network, etc.) can be trained based on this framework. The training environment selects a conventional deep learning framework and an adaptive hardware configuration; the basic parameters include the training batch size, the training round, the initial learning rate and the adjustment strategy, the loss function type, the optimizer type and the core parameters, which are all reasonably set according to the characteristics of the selected neural network and the task requirements. Among them, the loss function selects a type suitable for the profile mapping task (such as mean square error loss), which is used to measure the deviation between the predicted profile and the real profile; the optimizer selects a conventional gradient descent type optimizer to update the model parameters. Those skilled in the art can select the specific neural network architecture according to the actual situation. The training data can also use the conventional initialization method (such as normal initialization, Xavier initialization, He initialization, etc.) suitable for the selected neural network to ensure that the initial distribution of parameters is reasonable and lay a foundation for training convergence.

[0058] Optionally, the artificial intelligence prediction model training process can generally include: Forward propagation: The training set input data is fed into the model, and after collaborative processing by the model's various general functional modules, the predicted silicon wafer imaging contour data is output. The generalization components of each module typically include a feature extraction layer / module, a feature transformation layer / module, and optionally, a feature fusion layer / module and an output mapping layer / module. The processing generally proceeds as follows: First, the input optical contour standardized data is subjected to multi-scale feature extraction through a feature extraction layer (such as convolutional layers, pooling layers, etc.) to obtain contour feature maps at different levels (including basic features such as edges, textures, and geometric shapes, as well as deep semantic features). Then, the extracted feature maps are adjusted in dimension, enhanced, or restored through a feature transformation layer (such as fully connected layers, transposed convolutional layers, attention layers, etc.) to adapt to the feature representation requirements of the silicon wafer imaging contour. If the model includes a feature fusion layer (such as a stitching layer, additive layer, etc.), the feature maps at different levels are simultaneously fused to supplement feature details and improve the completeness of feature representation. Finally, the processed features are converted into silicon wafer imaging contour prediction data matching the output format through the output mapping layer.

[0059] Loss calculation: Based on the preset weights of simulation data and actual measurement data, the weighted loss is calculated to quantify the deviation between the predicted result and the actual result.

[0060] Backpropagation and parameter update: Clear the gradient cache from the previous round, calculate the gradient of the loss function with respect to the model parameters, and update the model parameters through the optimizer.

[0061] Optionally, after each training round, the model can be set to validation mode, gradient calculation can be turned off, validation set data can be loaded to perform forward propagation, and core metrics such as validation set loss and contour matching degree can be calculated to evaluate model performance.

[0062] Optionally, an early stopping strategy can be adopted to avoid model overfitting. If the loss on the validation set does not decrease effectively after several consecutive rounds, training is stopped. During training, the model with the best performance on the validation set is saved, and key information such as model parameters and optimizer status is recorded for easy loading and use later.

[0063] Finally, the optimal model is fine-tuned based on the training results, and parameters such as the learning rate are adjusted to further optimize performance. Finally, the training is confirmed to be complete according to the preset convergence criteria (such as the validation set loss stabilizing below the threshold and the contour matching degree meeting the standard), and the final artificial intelligence prediction model is output.

[0064] In some alternative embodiments, since the actual measured range of the aerial image on the mask pattern during actual operation is limited to approximately 750 nm, while the range considered by ordinary lithography or etching is much larger than this, in order to ensure the accuracy of the artificial intelligence prediction model, it is also possible to optimize the actual measurement data. Figure 3is a schematic diagram of an extended process of actual measurement data in a method for predicting the imaging effect of mask defects on the wafer level according to an embodiment of the application, as shown in Figure 3 The training method of the artificial intelligence prediction model includes at least the following steps S301 to S306.

[0065] Step S301, obtain the measured mask pattern corresponding to the measured optical profile and the measured position.

[0066] Step S302, determine the image acquisition range of the optical proximity correction simulation. The image acquisition range needs to be set in combination with the actual needs of the target patterning process, and generally needs to meet two conditions: one is to cover the area of the OPC simulation; and the other is to take the measured position obtained in S301 as the center to ensure that the measured area is at the core position of the simulation range and the peripheral extended area completely covers the process influence range.

[0067] This step can ensure that the subsequently supplemented peripheral profile contains all the key graphical information affecting the imaging of the core measured area; at the same time, the measured position is taken as the center to ensure the dominance of the core measured area.

[0068] Step S303, build an optical profile simulation model based on the measured mask pattern, and calibrate the optical parameters and mask parameters in the optical profile simulation model. The uncalibrated simulation model has parameter deviation, and the output simulation optical profile has large difference with the real process scene. If it is directly used for supplementary expansion, it will cause distortion of the expanded optical profile, and thus reduce the model training precision. Therefore, the model parameters need to be calibrated based on the real measured optical profile to ensure that the output characteristics of the simulation model are completely consistent with the aerial image measurement system, and to ensure the reliability of the supplemented profile.

[0069] Step S304, simulate the simulation optical profile in the measured mask pattern based on the optical profile simulation model and taking the measured position as the center according to the image acquisition range. This step obtains the complete optical profile in a wide range through the optical profile simulation model, which contains the simulation part of the core measured area and the simulation part of the peripheral extended area. In this way, the defect of limited measurement range of the aerial image measurement system can be made up, and since the model has been calibrated, the deviation of the simulation profile from the real process is very small, and it can be used for subsequent expansion splicing.

[0070] In step S305, the center position of the simulated optical profile is replaced by the measured optical profile, and the measured optical profile and the surrounding simulated optical profile are spliced to obtain an expanded optical profile. If the full-simulated wide-area profile is directly used, the real process characteristics of the measured data will be lost, resulting in that the expanded data loses the value of anchoring the real process. If only the measured small-range profile is retained, the problem of limited range cannot be solved. Therefore, in this step, the splicing strategy of core measurement replacement + surrounding simulation supplement is adopted, which can not only retain the real process information of the core area, but also supplement the process interaction information of the surrounding wide area. At the same time, if there is a mutation in the splicing boundary, it will lead to profile distortion, and the integrity and continuity of the profile need to be ensured through smoothing processing.

[0071] The final expanded optical profile has both authenticity and wide range: the measured profile in the core area ensures that the data is close to the real process, the surrounding simulated profile supplements the complete process interaction information, solves the core problem of insufficient measurement range of the aerial image measurement system, and the edge smoothing processing avoids splicing distortion and ensures the consistency of the expanded profile and the real wide-area optical profile, providing high-quality input data for AI model learning of complete process rules.

[0072] In step S306, the expanded optical profile and the silicon wafer imaging profile extracted from the scanning electron microscope image actually measured after the measured mask pattern is subjected to a patterning process are combined to form actual measurement data. The expanded optical profile is used as input data in the subsequent AI model training process, and the silicon wafer imaging profile is the output of the AI model training process. The size of the silicon wafer imaging profile is consistent with the CD-SEM measurement data.

[0073] Through this operation, a high-quality actual measurement data pair of large-range input and small-range output can be constructed, so as to realize the design goal of larger range of input mask to fully consider the process effect and smaller range of output to match the CD-SEM measurement. The advantages of measured data close to the real process are retained, and the defects of insufficient measurement range of the aerial image measurement system are compensated by the OPC-level large-range setting. Finally, when the data pair is used for model training, the prediction accuracy of the silicon wafer imaging profile can be significantly improved, especially for defects that are greatly affected by the surrounding process, which can significantly improve the prediction accuracy.

[0074] It should be understood that in some embodiments, each part can be realized by hardware, software, firmware or a combination thereof. In the above implementation, a plurality of steps or methods can be realized by software or firmware stored in a memory and executed by a suitable instruction execution system.

[0075] The embodiment also provides a computer program product 10, a computer readable storage medium 20, and a computer device 30. Figure 4is a schematic diagram of a computer program product 10 according to an embodiment of the application, Figure 5 is a schematic diagram of a computer readable storage medium 20 according to an embodiment of the application, Figure 6 is a schematic diagram of a computer device 30 according to an embodiment of the application. The computer program product 10 comprises a computer program 11 which, when executed by the processor 32, implements the steps of any of the above described methods for predicting the imaging effect of a reticle defect at wafer level. The computer readable storage medium 20 has stored thereon the above described computer program 11 which, when executed by the processor 32, implements the steps of any of the above described methods for predicting the imaging effect of a reticle defect at wafer level. The computer device 30 can comprise a memory 31, a processor 32 and the computer program 11 stored on the memory 31 and running on the processor 32.

[0076] The computer program 11 for performing the operations of the application can be in an assembly language, machine language, machine code, machine dependent instructions, microcode, firmware instructions, state setting data, configuration data for integrated circuits, or in source code or object code written in any combination of one or more programming languages. The computer program 11 can be executed in whole or in part on the user's computer, executed as a stand-alone software package, executed partly on the user's computer and partly on a remote computer, or executed entirely on a remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present application.

[0077] For the purposes of the description of the present embodiment, the computer program product 10 is a tangible computer program product comprising the computer program 11. For the purposes of the description of the present embodiment, the computer readable storage medium 20 is a tangible device that can retain and store the computer program 11, which can be any apparatus that can contain, store, communicate, propagate or transport the program 11 for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer readable storage medium 20 include the following: portable computer diskette, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, a floppy disk, a mechanical encoder device, and any suitable combination of the foregoing.

[0078] The computer device 30 can be, for example, a server, a desktop computer, a notebook computer, a tablet computer, or a smartphone. In some examples, the computer device 30 can be a cloud computing node. The computer device 30 can be described in the general context of computer system-executable instructions, such as program modules, being executed by a computer system. Generally, program modules can include routines, programs, objects, components, logic, data structures, and the like, that perform particular tasks or implement particular abstract data types. The computer device 30 can be practiced in distributed cloud computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed cloud computing environment, program modules can be located in local or remote computer system storage media including memory storage devices.

[0079] The computer device 30 can include a processor 32 adapted to execute instructions stored in memory 31, which provides temporary storage for operations of the instructions during execution. The processor 32 can be a single core processor, multi-core processor, computing cluster, or any number of other configurations. The memory 31 can include random access memory (RAM), read only memory, flash memory, or any other suitable memory systems.

[0080] The computer device 30 can also include a network adapter / interface and an input / output (I / O) interface. The I / O interface allows for input and output of data with external devices that can be connected to the computer device. The network adapter / interface can provide for communication between the computer device and a network, generally illustrated as communication network.

[0081] At this point, those skilled in the art will appreciate that although specific exemplary embodiments of the application have been described herein, the present application also encompasses many other variations or modifications in accordance with the principles of the application as set forth above. Accordingly, the scope of the present application should be understood to include all such variations and modifications.

Claims

1. A method for predicting the imaging effect of mask defects at the silicon wafer level, comprising: Obtain the target mask layout, wherein the target mask layout is a layout with potential defects; An aerial image of the target mask pattern is generated using an aerial image measurement system; The target's optical profile is obtained by analyzing and processing the aerial image of the target. The target silicon wafer imaging profile corresponding to the target optical profile is generated according to the pre-trained prediction model. The prediction model is used to learn the mapping relationship between the optical profile before the patterning process and the silicon wafer imaging profile after the patterning process, and outputs the corresponding silicon wafer imaging profile based on the input optical profile.

2. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 1, wherein, The prediction model is an artificial intelligence prediction model; The training steps for the artificial intelligence prediction model include: A pre-established graphical process model is obtained, which is used to generate the outline of the mask pattern after graphical processing based on the input mask pattern. Simulation data is generated based on the graphical process model. The simulation data consists of a data pair consisting of the optical contour of the mask pattern before the graphical process and the silicon wafer imaging contour after the graphical process. The artificial intelligence prediction model is trained based on the simulation data and the actual measurement data. The actual measurement data consists of a data pair composed of the optical contour obtained by the aerial image measurement system from the actual measurement of the mask pattern and the silicon wafer imaging contour extracted from the scanning electron microscope image obtained by the actual measurement of the mask pattern after the patterning process.

3. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 2, wherein, The step of training the artificial intelligence prediction model based on the simulation data and actual measurement data includes: The contour information recorded by both the simulation data and the actual measurement data is converted into a symbolic distance field image; The artificial intelligence prediction model is obtained by training the symbolic distance field image.

4. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 2, wherein, The step of training the artificial intelligence prediction model based on the simulation data and actual measurement data includes: Determine the weight values ​​of the simulation data and the actual measurement data during the training process; The simulation data and the actual measurement data are input into a preset neural network architecture; The artificial intelligence prediction model is obtained by training the simulation data and the actual measurement data according to the weight values ​​in the preset neural network architecture.

5. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 2, wherein, The step of generating simulation data based on the graphical process model includes: Obtain the complete layout data corresponding to the simulation target chip; The complete map data is grouped according to graphic features to obtain multiple graphic groups; Select the target training graph from the multiple graph groups; The optical profile and silicon wafer imaging profile corresponding to the target training pattern are generated using the graphical process model, thereby obtaining the simulation data.

6. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 2, wherein, The steps for obtaining the actual measurement data include: Obtain the measured mask pattern and measured position corresponding to the measured optical profile in the actual measurement data; Determine the image acquisition range for optical proximity correction simulation; An optical profile simulation model was built based on the measured mask pattern, and the optical parameters and mask parameters in the optical profile simulation model were calibrated. The optical profile simulation model is used to simulate the optical profile in the measured mask pattern, with the measured position as the center and according to the image acquisition range, to obtain the simulated optical profile. The measured optical profile is used to replace the center position of the simulated optical profile, and the measured optical profile is combined with the surrounding simulated optical profile to obtain the extended optical profile. The actual measurement data is composed of the silicon wafer imaging profile extracted from the scanning electron microscope image obtained by the actual measurement of the extended optical profile and the actual measured mask pattern after the patterning process.

7. The method for predicting the imaging effect of mask defects at the silicon wafer level according to claim 2, wherein, The graphical process model includes a photolithography model and / or an etching model. The photolithography model is used to generate the photoresist outline of the mask pattern after photolithography based on the input mask pattern. The etching model is used to generate the etched outline of the mask pattern after etching based on the input mask pattern. For the photolithography model, the silicon wafer imaging profile is the photoresist profile; For the etching model, the silicon wafer imaging profile is the etched profile.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that... When the computer program is executed by the processor, it implements the steps of the method for predicting the imaging effect of mask defects at the silicon wafer level as described in any one of claims 1 to 7.

9. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method for predicting the imaging effect of mask defects at the silicon wafer level as described in any one of claims 1 to 7.

10. A computer device comprising a memory, a processor, and a machine-executable program stored in the memory and running on the processor, wherein the processor, when executing the machine-executable program, implements the steps of the method for predicting the imaging effect of mask defects at the silicon wafer level according to any one of claims 1 to 7.

Images