Load effect prediction method and device, medium and product

Abstract

The invention discloses a load effect prediction method and device, a medium and a product, and relates to the technical field of semiconductor etching. The load effect prediction method comprises the steps that an optical space image corresponding to a first neighborhood of a to-be-predicted graph and a partial graph distribution density image corresponding to a second neighborhood are acquired, to-be-simulated etching input information of the to-be-predicted graph is determined, and the coverage range of the second neighborhood is larger than that of the first neighborhood; and inputting the to-be-simulated etching input information into a pre-trained etching prediction model to obtain an etching simulation result of the to-be-predicted graph under the action of the load effect. By inputting the optical space image into the etching prediction model, the etching prediction model can capture the characteristics related to the optical proximity effect affecting etching from the optical space image, then the initial condition of etching is predicted, a more accurate etching simulation result is obtained according to the predicted initial condition of etching, and the etching simulation accuracy is improved. And the accuracy of load effect prediction is improved from the side.

Description

Technical Field

[0001] This application belongs to the field of semiconductor etching technology, and in particular relates to a method, device, medium and product for predicting load effects. Background Technology

[0002] In chip manufacturing, after the pattern is transferred onto the photoresist using photoresist and a mask, the wafer can be etched. Etching removes material not protected by the photoresist, thus forming the desired pattern on the material layer.

[0003] During etching, to avoid material waste and save manpower, etching prediction models are typically built to predict the etching simulation results of the wafer. However, when the load is inconsistent in different areas of the wafer, the prediction accuracy of the etching prediction model decreases due to the load effect, resulting in reduced uniformity of the etching process and affecting chip performance and yield.

[0004] Therefore, improving the accuracy of etching prediction models in predicting load effects is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a method, device, medium, and product for predicting load effects, which can improve the accuracy of etching prediction models in predicting load effects.

[0006] A first aspect of this application provides a method for predicting load effects, comprising: Obtain a pattern distribution density image of a chip design layout, wherein the chip design layout includes multiple patterns to be predicted; For any of the aforementioned patterns to be predicted, determine the optical spatial image corresponding to the first neighborhood of the pattern to be predicted; Based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted, and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, the simulated etching input information of the pattern to be predicted is determined, wherein the coverage of the second neighborhood is greater than the coverage of the first neighborhood. The etching input information to be simulated is input into a pre-trained etching prediction model to obtain the etching simulation results of the pattern to be predicted under the load effect.

[0007] A second aspect of this application provides a load effect prediction device, comprising: The pattern distribution density image acquisition module is used to acquire the pattern distribution density image of the chip design layout, wherein the chip design layout includes multiple patterns to be predicted; An optical spatial image determination module is used to determine, for any of the predicted graphics, the optical spatial image corresponding to the first neighborhood of the predicted graphics; The module for determining the input information to be simulated etching is used to determine the input information to be simulated etching of the pattern to be predicted based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and a partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, wherein the coverage range of the second neighborhood is greater than the coverage range of the first neighborhood. The etching simulation module is used to input the etching input information to be simulated into a pre-trained etching prediction model to obtain the etching simulation results of the pattern to be predicted under the load effect.

[0008] A third aspect of the embodiments of this application provides an electronic device, the device comprising: a memory and a program or instructions stored in the memory and executable on a processor, wherein when the program or instructions are executed by the processor, they implement the load effect prediction calculation method provided in any of the embodiments of this application described above.

[0009] A fourth aspect of the embodiments of this application provides a readable storage medium on which a program or instructions are stored, and when the program or instructions are executed by a processor, they implement the load effect prediction method provided by any aspect of the embodiments of this application described above.

[0010] A fifth aspect of the embodiments of this application provides a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform the load effect prediction method provided in any of the embodiments of this application described above.

[0011] In the loading effect prediction method provided in this application, the shape of the photoresist changes due to the optical proximity effect, thereby affecting the initial etching conditions. Therefore, changes in the initial conditions during etching indirectly affect the etching simulation results. This application inputs the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and the partial pattern distribution density image corresponding to the second neighborhood into the etching prediction model. This allows the etching prediction model to capture features related to the optical proximity effect affecting etching from the optical spatial image, obtain the relationship between the etching gas consumption rate and the pattern density from the pattern distribution density image, and thus predict the initial etching conditions. Based on the predicted initial etching conditions, a more accurate etching simulation result is obtained, thereby improving the accuracy of the loading effect prediction. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1A flowchart illustrating a load effect prediction method provided in an embodiment of this application; Figure 2 A schematic diagram of a prediction point provided for an embodiment of this application; Figure 3 This application provides a schematic diagram of a model training process. Figure 4 This is a schematic diagram of another load effect prediction method provided in the embodiments of this application; Figure 5 This is a schematic diagram of the internal data flow of an etching prediction model provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a load effect prediction device provided in an embodiment of this application; Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0015] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0016] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations.

[0017] In the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0018] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0019] Etching is the process of removing material from areas of a wafer surface not covered by photoresist, under the protection of photoresist, thereby transferring the pattern defined by photolithography onto a lower thin film or silicon substrate.

[0020] Chip design layout is a design graphic data that describes the geometric layout of devices on each physical layer of a chip, including their shape, size, and location.

[0021] The loading effect refers to the phenomenon that the etching rate deviates from the ideal state as the area, density, or pattern distribution of the area to be etched changes.

[0022] Currently, when predicting etching simulation results, most methods input the density image of the chip design layout into the etching prediction model to obtain the simulation results. However, due to the influence of loading effects, the accuracy of the etching simulation results is poor. The etching simulation result refers to the actual physical structure formed on the wafer after the simulated etching process is completed. Poor accuracy of the etching simulation results indicates a significant difference between the physical structure predicted by the model and the actual physical structure formed.

[0023] In view of this, this application provides a method, apparatus, medium, and product for predicting load effects. In the load effect prediction method provided in this application, by inputting a partial pattern density image (including a larger coverage area) and an optical spatial map (including a smaller coverage area) of the pattern to be predicted into an etching prediction model, the etching prediction model can extract the relationship between the pattern density and the etching gas consumption rate from the partial pattern density image and extract features related to the optical proximity effect affecting etching from the optical spatial map. This allows for the prediction of the initial etching conditions, resulting in more accurate etching simulation results and indirectly improving the accuracy of load effect prediction. Furthermore, by acquiring a partial pattern density image (including a larger coverage area) and an optical spatial map (including a smaller coverage area), computational resource consumption can be reduced while meeting accuracy requirements, thereby improving etching simulation efficiency.

[0024] The loading effect prediction method provided in this application can be applied to the production design process of semiconductor manufacturing enterprises. It is used to predict the etching simulation results of wafers containing photoresist after photolithography, thereby reducing etching errors in actual production and avoiding waste of production resources. In practical applications, the layout file and etching parameters are transferred to a computer, which then predicts the etching simulation results for each area of ​​the layout based on the loading effect prediction method provided in this application. The etching parameters may include wafer temperature, electrode temperature, etching time, electrode spacing, type of etching gas, and flow rate.

[0025] It should be noted that the application scenarios described in the above embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will understand that with the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems. The load effect prediction method provided by the embodiments of this application can be applied to various application scenarios that require the prediction of etching simulation results.

[0026] In practical applications, the execution subject of the load effect prediction method in this application embodiment can be an electronic device, specifically a server, personal computer, etc., and this application embodiment does not limit this.

[0027] The following describes specific embodiments of the load effect prediction method, apparatus, equipment, medium, and product provided in this application. First, the load effect prediction method will be introduced.

[0028] Figure 1 This is a flowchart illustrating the load effect prediction method provided in an embodiment of this application. Figure 1 As shown, the method includes S101 to S104.

[0029] S101: Obtain the pattern distribution density image of the chip design layout, which includes multiple patterns to be predicted.

[0030] S102: For any graphic to be predicted, determine the optical spatial image corresponding to the first neighborhood of the graphic to be predicted.

[0031] S103: Based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted, and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, determine the simulated etching input information for the pattern to be predicted. Furthermore, the coverage area of ​​the second neighborhood is greater than the coverage area of ​​the first neighborhood.

[0032] S104: Input the etching input information to be simulated into the pre-trained etching prediction model to obtain the etching simulation results of the pattern to be predicted under the load effect.

[0033] In this embodiment, the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and the partial pattern distribution density image corresponding to the second neighborhood are input into the etching prediction model, thereby improving the accuracy of the etching simulation results output by the model. The coverage range of the second neighborhood is larger than that of the first neighborhood because the accuracy requirement is lower when determining the pattern distribution density image. However, this difference in coverage range also leads to the possibility that the density values ​​of each region in the determined pattern distribution density image may be the same. There may be a large number of identical density values ​​in the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted. The features obtained by the etching prediction model from this partial pattern distribution density image are relatively simple, resulting in poor accuracy of the output etching simulation results. It can also be understood that the density sampling of the pattern is large, making it difficult to obtain the different etching simulation results of different patterns under the load effect within a small range, resulting in low accuracy of the etching simulation results. By inputting higher-precision optical spatial images into the etching prediction model, the etching prediction model can capture features related to the optical proximity effect that affects etching from the optical spatial images, obtain richer features for the etching prediction model to predict, and then predict the initial conditions of etching. Based on the predicted initial conditions of etching, more accurate etching simulation results can be obtained, which indirectly improves the accuracy of loading effect prediction.

[0034] For S101, the chip design layout can be any chip design layout that requires etching simulation result prediction. The chip design layout includes multiple patterns to be predicted, which are the prediction objects of the subsequent etching prediction model. For example, the etching prediction model can output the etching simulation result of any pattern to be predicted included in the chip design layout under the action of the loading effect.

[0035] The load effect varies in different regions of the chip design layout, which is related to the different pattern distribution in each region. Therefore, determining the pattern distribution density image can quantify the pattern distribution and identify the regions that affect the strength of the load effect, so that the etching prediction model can obtain etching simulation results.

[0036] In some embodiments, a fixed-size sliding window can be slid across the chip design layout according to its side length to calculate the pattern density within each window. The combination of pattern densities within each window yields a pattern distribution density image of the chip design layout. The fixed size can be the area encompassed by the window and can be set as needed; this embodiment does not impose any limitations on this.

[0037] In some embodiments, the entire area of ​​the chip design layout can be divided into multiple sub-regions, and the pattern density of each sub-region can be determined. The pattern densities of the multiple sub-regions can be combined to obtain a pattern distribution density image of the chip design layout.

[0038] Regarding S102, since the pattern distribution density image is difficult to distinguish the differences in the loading effect of different patterns within a small area, the accuracy of the output etching simulation results can be improved by inputting etching difference information used to distinguish the different patterns within a small area under the loading effect into the etching prediction model. Specifically, the optical spatial image corresponding to the first neighborhood of each pattern to be predicted can be determined. For example, the area of ​​the first neighborhood is 1*1µm. 2 ~10*10um 2 between.

[0039] For example, when the pattern to be predicted is a transistor, the size of the first neighborhood can be 1*1um. 2 ~2*2um 2 When the image to be predicted is a memory cell array, the size of the first neighborhood can be 3*3um. 2 ~5*5um 2 .

[0040] In some embodiments, any pattern to be predicted can be used as the partitioning center to divide the chip design layout, obtaining a sub-region corresponding to the first neighborhood of the pattern to be predicted. Then, the optical spatial image of the sub-region is determined, thus obtaining the optical spatial image corresponding to the first neighborhood of the pattern to be predicted.

[0041] Specifically, during the partitioning process, the prediction point for the pattern to be predicted can be determined first based on the prediction requirements. This prediction point can be located at any position on the pattern, such as its boundary or center point. For the same pattern, different prediction requirements may lead to different prediction points. Prediction requirements characterize the location of the pattern that is of primary concern during etching. Specifically, these can include two types of prediction requirements: high-risk areas of the process and critical functional areas of the device. The former addresses situations where the pattern changes before and after etching, such as the pattern merging with other patterns or internal splitting after etching. The latter addresses situations where etching affects chip performance. For example, if the gate linewidth of a transistor changes before and after etching, the chip's switching speed will change significantly.

[0042] Figure 2 A schematic diagram of prediction points provided for embodiments of this application, such as... Figure 2 As shown.

[0043] For example, if both the pattern to be predicted 1 and the pattern to be predicted 2 are rectangles, and the prediction requirement is to avoid the two adjacent patterns connecting after simulated exposure and etching on the silicon wafer, then the prediction point for the pattern to be predicted 1 and the pattern to be predicted 2 is the midpoint representing the shortest distance between the two patterns, that is... Figure 2 A solid circle. At this point, the prediction point is located outside the shape to be predicted.

[0044] Understandably, when the forecasting demand is determined, the location of the forecasting point is fixed. Generally, for a given forecasting demand, a forecasting point may be included in a given graph.

[0045] After determining the prediction point of the pattern to be predicted, the prediction point can be used as the dividing center to divide the chip design layout, and the sub-region corresponding to the first neighborhood of the pattern to be predicted can be obtained, so as to determine the optical spatial image of the sub-region.

[0046] In some embodiments, an optical simulation model of a sub-region corresponding to a first neighborhood of a pattern to be predicted can be constructed based on preset exposure parameters, and then an optical spatial image of the sub-region corresponding to the first neighborhood can be obtained based on the optical simulation model.

[0047] Specifically, the sub-regions corresponding to the first neighborhood of the graphic to be predicted can be discretized into a numerical matrix of 0 to 1, thus converting the continuous and visualized graphic into a digital signal that can be recognized and calculated by a computer.

[0048] For example, the sub-regions corresponding to the first neighborhood can be rasterized to determine the area occupied by the image to be predicted in each grid, obtaining the value of each grid. Based on the values ​​of each grid, a numerical matrix of the sub-regions corresponding to the first neighborhood of the image to be predicted is obtained. The larger the area occupied by the image to be predicted in the grid, the closer the value of that grid is to 1. Then, preset exposure parameters and this numerical matrix can be used as input to an optical simulation model for optical simulation. This optical simulation model simulates the process of light passing through an optical system and forming a light intensity distribution on the photoresist to obtain the illumination intensity on each grid.

[0049] Obtaining the illumination intensity on each grid cell yields the optical spatial image of the sub-region corresponding to the first neighborhood of the image to be predicted. Each pixel in this optical spatial image corresponds to a light intensity value, which can be a dimensionless parameter. The parameter value can be a first preset value and a second preset value. For example, the first preset value can be 1, and the second preset value can be 0. A light intensity value of 1 indicates that the region of the pixel is transparent, and a light intensity value of 0 indicates that the region of the pixel is opaque.

[0050] For example, preset exposure parameters may include focal plane, photoresist parameters, light source parameters, mask manufacturing parameters, etc. The focal plane refers to the plane focused by the lithography machine lens. During exposure, if the wafer surface is not on the focal plane, it will cause image blurring; a clear image can be obtained by adjusting the focal length. Photoresist parameters may include photoresist thickness, refractive index, photosensitivity, absorption coefficient, etc. This application embodiment does not limit the specific photoresist parameters used. Refractive index refers to the photoresist's ability to refract light; photosensitivity refers to the photoresist's sensitivity to exposure energy; and absorption coefficient refers to the photoresist's ability to absorb exposure light waves. Light source parameters include the wavelength and shape of the exposure light source, such as a ring shape. Mask manufacturing parameters include mask type, mask material, etc., and the mask type may include a binary mask, etc.

[0051] For S103, the etching prediction model predicts any pattern to be predicted. Therefore, the etching input information to be simulated for the etching prediction model can include information related to the pattern to be predicted. As mentioned above, the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and the pattern distribution density image of the chip design layout are obtained. In order to reduce the amount of processing required by the etching prediction model, a local pattern distribution density image including the region corresponding to the pattern to be predicted can be divided from the pattern distribution density image of the chip design layout.

[0052] Specifically, a portion of the pattern distribution density image corresponding to a second neighborhood can be delineated from the pattern distribution density image. This portion of the pattern distribution density image corresponding to the second neighborhood includes the pattern to be predicted within the chip design layout. For example, the size of this second neighborhood can be 1*1mm. 2 ~10*10mm 2 Any area between the two. The size of the second neighborhood reflects the density distribution of the surrounding environment of the graphic to be predicted; therefore, the size of the second neighborhood is also related to the size of the graphic to be predicted.

[0053] For example, if the image to be predicted is a 1.2mm x 1.5mm data converter, the size of the second neighborhood can be 3mm x 3mm. Of course, different sizes of the second neighborhood can be selected based on different prediction efficiency requirements. The higher the prediction efficiency, the smaller the size of the second neighborhood should be.

[0054] For example, the optical spatial image corresponding to the first neighborhood of the pattern to be predicted, and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, can be determined as the input information for the simulated etching of the pattern to be predicted. If other information related to the pattern to be predicted is also included, this information can be used together as the input information for the simulated etching of the pattern to be predicted.

[0055] It should be noted that the coverage of the second neighborhood is larger than that of the first neighborhood because the optical proximity effect causes changes in the edge positions or other specific locations of the pattern to be predicted before and after photolithography, thus affecting the initial etching conditions. These changes are small and difficult to capture over a large area. Using an optical spatial map with a large coverage area makes it difficult to characterize these pattern changes, leading to inaccurate etching simulation results. Furthermore, ion incident angles, shading effects, and reactant transport during the etching process are all closely related to the microscopic shape of the pattern. These effects all need to be reflected in a high-precision optical spatial map; therefore, etching prediction models require high-precision optical spatial maps to capture the details of the pattern.

[0056] In terms of pattern density, the etching rate change is usually caused by a large-scale abrupt change in pattern density. For example, the consumption of etchant and the emission of products will affect the etching uniformity over a large area. This effect is manifested over a large range. Therefore, a partial pattern distribution density pattern with a large coverage area is used.

[0057] This can also be understood as the loading effect being related to the pattern density within a local area. However, the loading effect typically varies over a relatively large range. Therefore, a small partition size is not required for dividing the chip design layout; obtaining a first neighborhood with a large coverage area is sufficient for the etching prediction model to capture density variations. For optical spatial images, since these images are obtained through optical simulations, which need to capture the diffraction and interference effects of light—effects that vary on the order of wavelength, i.e., on the nanometer scale—a small partition size is needed to divide the chip design layout, obtaining a smaller first neighborhood to ensure the accuracy of the resulting optical spatial image, thereby yielding more accurate etching simulation results.

[0058] Furthermore, in terms of computational resource consumption, if a high-precision partial pattern distribution density image is used, it indicates that when executing S101, a large amount of computational resources are required to divide the chip design layout and determine the pattern density of each sub-region. This results in low efficiency, high computational resource consumption, and affects the etching simulation progress.

[0059] Alternatively, an optical image prediction model capable of outputting optical spatial images can be obtained by training a machine learning model. Then, the sub-region corresponding to the first neighborhood is input into this optical image prediction model to obtain the optical spatial image corresponding to that first neighborhood. This optical image prediction model can be trained using a supervised method.

[0060] For S104, the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted can be input into a pre-trained etching prediction model to obtain the etching simulation result of the pattern to be predicted under the load effect. The etching prediction model can be any model that can predict the etching simulation result, and the embodiments of this application do not limit the specific implementation of the etching prediction model. For example, the etching prediction model can be an analytical model based on physical mechanisms, a data-driven machine learning model, or a hybrid model of physical mechanisms and data-driven approaches.

[0061] For example, the etching prediction model can extract various features from the input information of the etching to be simulated through network layers such as convolutional neural networks, and based on these features, predict the outline of the pattern to be predicted under the load effect to obtain the etching simulation result.

[0062] In some embodiments, for S101, generally speaking, high-density regions with many patterns have denser patterns, making it difficult for etching particles to diffuse during etching, resulting in a stronger loading effect, which can manifest as a slower etching rate. Conversely, low-density regions with fewer patterns have sparser patterns, allowing etching particles to diffuse more quickly, resulting in a weaker loading effect. In regions where pattern density changes abruptly, the loading effect varies significantly, easily leading to reduced process uniformity and ultimately affecting chip performance and yield. When executing S101, the chip design layout can be divided into multiple sub-regions; Determine the ratio of the total area of ​​all graphics within each sub-region to the total area of ​​the corresponding sub-region to obtain the graphic density of the corresponding sub-region; The distribution of graphic density in each sub-region on the chip design layout constitutes a graphic distribution density image.

[0063] For example, during partitioning, the chip design layout can be divided into multiple sub-regions using any partitioning size. This partitioning size can be the area, side length, etc., of each sub-region. The partitioning size can be set as needed, and this embodiment does not impose any limitations on it.

[0064] For example, considering computational resource consumption and density accuracy requirements, the area of ​​each sub-region obtained can be as small as 1*1µm. 2 ~10*10um 2 between.

[0065] For example, if the requirement is that the sub-regions obtained by the division include various graphic arrangements, then by division, an area of ​​10*10um can be obtained. 2 Within the maximum area, a sub-region may contain multiple graphics, and thus multiple graphic arrangements.

[0066] For example, in memory areas, the graphics layout might be a dense array of uniformly arranged graphics. In areas containing analog circuits, the graphics are sparse, with individual, large-area graphics that, while having low density, exhibit significant differences in size.

[0067] It should be noted that, during the partitioning process, the chip design layout can be divided equally according to any partitioning size, resulting in multiple sub-regions with the same area.

[0068] In some embodiments, the layout can be divided according to the distribution of patterns on the chip design board. For example, a complete pattern to be predicted can be divided into a sub-region. Alternatively, the chip design board can be divided with the pattern to be predicted as the center, resulting in multiple sub-regions centered on the pattern to be predicted.

[0069] After obtaining multiple sub-regions, the graphic distribution density image of the chip design layout can be determined.

[0070] In some embodiments, after obtaining multiple sub-regions, the ratio of the total area of ​​all graphics within each sub-region to the total area of ​​the corresponding sub-region can be determined to obtain the graphic density of the corresponding sub-region.

[0071] For example, the resulting sub-regions include sub-region 1, which has an area of ​​10*10um. 2 The total area of ​​all included graphics is 2*2um. 2 If so, the graphic density of subregion 1 is 0.04.

[0072] In some embodiments, different density weights can be assigned to different patterns. The density weight is multiplied by the ratio of the total area of ​​all patterns in the corresponding sub-region to the total area of ​​the corresponding sub-region to obtain the pattern density of the corresponding sub-region. It should be noted that, generally speaking, the same pattern may have different load effects in different locations on the chip design layout. Therefore, different density weights can be assigned to each pattern.

[0073] For example, if the sub-region includes a single graphic, the density weight is multiplied by the ratio of the area of ​​all graphics within the corresponding sub-region to the total area of ​​the corresponding sub-region to obtain the graphic density of the corresponding sub-region. If the sub-region includes multiple graphics, the density weights of each graphic are summed to obtain a summed result. This summed result is then multiplied by the ratio of the area of ​​all graphics within the corresponding sub-region to the total area of ​​the corresponding sub-region to obtain the graphic density of the corresponding sub-region. This graphic density represents the graphic density of all pixels within the same sub-region. It is understood that the density values ​​of all pixels within the same sub-region are the same.

[0074] After obtaining the pattern density of each sub-region, the pattern densities of each sub-region are stitched together according to their positions on the chip design layout to obtain the pattern distribution density image of the chip design layout.

[0075] It should be noted that the density values ​​of each pixel in the obtained graphic distribution density image are dimensionless parameters. This is because the density value is obtained by comparing two areas. The original dimension is area, and after obtaining the ratio, a dimensionless ratio is obtained, which is the density value.

[0076] In this embodiment of the application, by dividing the chip design layout into multiple sub-regions, the graphic density of each sub-region is determined, thereby obtaining a graphic distribution density image of the chip design layout. This method can quantify the distribution of each graphic in the chip design layout without complex calculations, resulting in high efficiency.

[0077] In some embodiments, to further improve the accuracy of the determined etching simulation results, for S103, the positional relationship information of the pattern to be predicted can be combined to determine the etching input information to be simulated. By providing richer input information, when executing S104, the etching prediction model can consider the influence of etching particle distribution, obtain more accurate etching simulation results, and avoid overfitting.

[0078] Specifically, the positional relationship information of the first and second neighborhoods of the pattern to be predicted can be determined first. Then, based on the optical spatial image and positional relationship information corresponding to the first neighborhood, and the partial pattern distribution density image and positional relationship information corresponding to the second neighborhood, the input information for the simulated etching of the pattern to be predicted can be determined.

[0079] Since the distribution of etched particles is a distribution that gradually decreases with distance, this positional relationship information represents the spatial distribution of etched gas during the etching process. It can also be understood as the influence of other pixels on the predicted point during the etching process, and also represents the probability of etched gas diffusing from other locations to this location.

[0080] In some embodiments, the positional relationship information includes coordinate information, such as the coordinates of the predicted point and the coordinates of other pixels in its corresponding neighborhood. This coordinate information can be determined by establishing a coordinate system for the chip design layout. In some embodiments, the positional relationship information may include the positional deviation between the predicted point of the pattern to be predicted and other pixels in its corresponding neighborhood. The positional deviation is used to measure the distance between other pixels and the predicted point. In some embodiments, a graph structure can be constructed between the predicted point and other pixels as the positional relationship information of the predicted point. This graph structure includes the edges and edge weights between the predicted point and other pixels, where the edge weights characterize the distance between the predicted point and other pixels.

[0081] In some embodiments, taking the determination of positional relationship information corresponding to the first neighborhood as an example, the positional relationship information of the predicted points of the image to be predicted within the first neighborhood is determined based on the following steps: Determine the distance between the predicted point of the image to be predicted and other pixels in the first neighborhood; Normalize each distance to obtain the positional deviation of each other pixel in the first neighborhood from the predicted point.

[0082] For example, normalizing each distance can be achieved by summing the distances. Each distance is then divided by the sum to obtain the normalized distance, which represents the positional deviation between the other pixel and the predicted point.

[0083] In this embodiment, the positional deviation determined based on the distance between the predicted point and other pixels characterizes the degree of influence of the etching particle distribution on the etching prediction of the pattern to be predicted. The larger the positional deviation, the smaller the influence of the etching gas on the predicted point. Conversely, the smaller the positional deviation, the greater the influence of the etching gas on the predicted point.

[0084] Because the spatial non-uniformity of the etching gas distribution during etching is taken into account, the influence of different positional relationships on the prediction point varies. The distribution characteristic of the etching gas is that the parameter value gradually decreases with increasing distance from the prediction point. This allows the etching prediction model to learn the importance of each location to the prediction point, capture the diffusion characteristics of the etching gas, and thus obtain more accurate etching simulation results.

[0085] Similarly, the positional relationship information of the predicted points in the second neighborhood of the image to be predicted is determined based on the following steps: Determine the distance between the predicted point of the graphic to be predicted and other pixels in the second neighborhood; Normalize each distance to obtain the positional deviation of each other pixel in the second neighborhood from the predicted point.

[0086] The method for determining the positional relationship information of the predicted points in the second neighborhood is the same as that for determining the positional relationship information of the predicted points in the first neighborhood, and the details will not be repeated here.

[0087] Furthermore, regarding S103, for example, the optical spatial image and positional relationship information corresponding to the first neighborhood, and the partial pattern distribution density image and positional relationship information corresponding to the second neighborhood can be directly used as the input information for the simulated etching of the pattern to be predicted. Alternatively, the aforementioned information can be fused to obtain the input information for the simulated etching of the pattern to be predicted. Furthermore, sine and cosine encoding can be performed on the positional relationship information corresponding to the first and second neighborhoods to convert each positional relationship information into a high-dimensional vector, so that the etching prediction model can extract positional relationship information features and predict the etching simulation results.

[0088] In some embodiments, in order to enable the etching prediction model to extract more accurate feature information of different dimensions from the simulated input information to be etched, and to avoid establishing false associations between density information and positional relationship information and between optical spatial information and positional relationship information, the positional relationship information can be fused with the density information and optical spatial information of the corresponding neighborhood, respectively.

[0089] During information fusion, the positional relationship information of the predicted points of the image to be predicted in the first neighborhood can be fused with the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information. The positional relationship information of the predicted points of the graphic to be predicted in the second neighborhood is fused with the feature of the partial graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information. Density location fusion information and light intensity location fusion information are used as the input information for the etching to be simulated.

[0090] This application does not restrict how feature fusion is performed. For example, the positional relationship information at the same location can be directly added to or multiplied by the density value and light intensity value to obtain density-positional fusion information and light intensity-positional fusion information. This information fusion method can also be understood as the way positional relationship information is introduced in deep learning models.

[0091] In load effect prediction, even if the density and light intensity distributions of two regions are similar, the etching simulation results at different locations can differ due to local differences in the distribution of etching gas. Therefore, by introducing positional relationship information, the etching prediction model can output different prediction results for different regions with similar density and light intensity distributions, thereby improving the accuracy of the output etching simulation results.

[0092] In this embodiment, during the etching process, densely patterned areas consume more etching gas, leading to a decrease in local gas concentration and thus a reduction in the etching rate. Conversely, sparsely patterned areas experience the opposite effect. Therefore, the loading effect is essentially a spatially related phenomenon. Density sub-images allow the etching prediction model to acquire the density of the pattern, while optical spatial images enable the model to more accurately determine the initial state of the etching process. Furthermore, the loading effect depends on the gas consumption in the surrounding area. By adding additional positional information, the etching prediction model can learn the gas characteristic relationship that locations closer to the prediction point have a greater impact on the gas supply, while locations farther away have a smaller impact, thereby improving the accuracy of the loading effect prediction.

[0093] In some embodiments, information fusion may include the aforementioned methods. To prevent the fused information from being mixed with different physical information, such as density information and positional relationship information, which would make it difficult for the etching prediction model to determine the correlation between the etching prediction result and each piece of physical information, information fusion can be performed by adding the positional deviation of each other pixel in the first neighborhood from the prediction point and the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood, thus obtaining light intensity position fusion information. The positional deviation of each other pixel in the second neighborhood from the predicted point is added to the density value of the corresponding pixel in the partial graphic distribution density image of the second neighborhood to obtain the density position fusion information.

[0094] It should be noted that the density value, light intensity value, and positional deviation are all dimensionless parameters.

[0095] When adding, the positional deviation at the same location can be added to the density value and light intensity value at that location, respectively.

[0096] For example, the positional relationship information of a pixel with coordinates (1,1) and the predicted point is 2. The density value of the corresponding pixel in the partial graphic distribution density image corresponding to the second neighborhood is 1, and the light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood is 0. Then, the density position fusion information of this pixel is 3, and the light intensity position fusion information is 2.

[0097] In this embodiment, by adding the positional deviation to the density and light intensity values ​​respectively, the information on the etching gas distribution at the same location is prevented from being mixed with the density and optical information. This maintains the independence of different physical information, enabling the etching prediction model to better understand and utilize this information, accurately establish the correlation between different physical information, and learn the correct correspondence between physical information and etching phenomena, thereby improving the accuracy of the predicted etching simulation results. Furthermore, for each prediction point, the etching gas distribution information near the prediction point is emphasized through positional relationship information, while information at a distance is weakened. This ensures that the input information of each pixel in each neighborhood obtained by the etching prediction model is locally weighted, thereby learning the influence of local etching gas on the prediction point to obtain more accurate etching simulation results, indirectly improving the accuracy of the prediction of the loading effect.

[0098] In some embodiments, the positional deviation of each other pixel in the first neighborhood from the predicted point and the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood can be feature-stitched together to obtain light intensity position fusion information. The positional deviation of each other pixel in the second neighborhood from the predicted point is combined with the density value of the corresponding pixel in the partial graphic distribution density image of the second neighborhood to obtain density position fusion information.

[0099] Feature stitching refers to the process of directly stitching the positional relationship information of the predicted points in the first neighborhood of each image to be predicted with the optical spatial image in the first neighborhood to form a multi-channel feature map, and stitching the positional relationship information of the predicted points in the second neighborhood with the image distribution density image in the second neighborhood to form a multi-channel feature map.

[0100] Specifically, the positional deviation at the same location is spliced ​​together with the light intensity value, and the positional deviation at the same location is spliced ​​together with the density value.

[0101] In the embodiments of this application, feature stitching can also ensure the independence of each feature, so that the etching prediction model can better understand and utilize this information to obtain more accurate etching simulation results.

[0102] Understandably, when performing feature fusion, using the positional deviation of different neighborhoods and the corresponding light intensity and density values ​​for feature fusion can achieve multi-scale information complementarity. Specifically, the etching prediction model can capture macroscopic trends through density-positional fusion information and microscopic details through light intensity-positional fusion information, thus predicting the etching size of the pattern to be predicted. This allows the etching prediction model to possess both macroscopic and microscopic information, thereby improving the accuracy of the output etching simulation results. It avoids the problem of difficulty in distinguishing small differences between a pattern and other patterns in the same region when covering a large area of ​​density information, thus preventing overfitting of the etching simulation results due to pattern differences.

[0103] This application also provides a model training method for training an etching prediction model, so as to obtain the etching simulation results of the pattern to be predicted through the trained etching prediction model.

[0104] Figure 3 This application provides a schematic diagram of a model training process, as shown in the embodiment of the present application. Figure 3 As shown, the etching prediction model is trained based on the following steps: S301: Obtain the graphic distribution density image of the sample design layout and several optical spatial images corresponding to the first sample neighborhood, including the sample graphic.

[0105] In this embodiment, the sample design layout can be any sample design layout that can be actually etched. The determination of the pattern distribution sample density image and several optical spatial images corresponding to the first sample neighborhood including the sample pattern are similar to the aforementioned determination of the pattern distribution density image and the optical spatial image corresponding to the first neighborhood of the chip design layout, and will not be described again in this embodiment.

[0106] S302: Obtain the etching results of each sample image under the load effect and use them as labels.

[0107] The labels can be the etching results of each sample pattern obtained after etching the chip design layout during the actual production process.

[0108] S303: For each sample pattern, based on the partial pattern distribution sample density image corresponding to the second sample neighborhood of the sample pattern and the optical spatial image corresponding to the first sample neighborhood of the sample pattern, determine the etching input information.

[0109] It should be noted that the coverage area of ​​the second sample neighborhood is larger than that of the first sample neighborhood.

[0110] S304: Input the optical spatial image of the etching input information into the etching prediction model to obtain the etching simulation results of the sample image.

[0111] S305: Train the etching prediction model based on the etching simulation results and labels of each sample image.

[0112] In some embodiments, various loss functions, such as mean squared error, can be used to determine the difference between the etching simulation results of the sample image and the label. The etching prediction model is then trained with the goal of reducing this difference until a preset termination condition is reached. This preset termination condition can be set as needed, such as reaching a preset number of iterations or a preset accuracy level. Methods for determining model accuracy are well-known to those skilled in the art and will not be elaborated upon in this application.

[0113] In this embodiment of the application, by jointly training the optical spatial image and the pattern distribution density image, the etching prediction model can acquire richer input features, thereby improving the accuracy of the predicted etching simulation results.

[0114] Adaptively, for S303, if the input information includes positional relationship information when actually applying the trained etching prediction model, then the positional relationship information of each sub-region can also be obtained as input during the training of the etching prediction model. Here, a sub-region refers to any region divided from the sample design layout.

[0115] In some embodiments, the partial pattern distribution sample density image corresponding to the second sample neighborhood of the sample pattern and the optical spatial image corresponding to the first sample neighborhood of the sample pattern can be determined as etching input information. Then, the positional relationship information of the first and second sample neighborhoods is obtained, and the etching input information is determined based on the positional relationship information of the first and second sample neighborhoods, the partial pattern distribution sample density image corresponding to the second sample neighborhood, and the optical spatial image corresponding to the first sample neighborhood of the sample pattern.

[0116] Similarly, the positional relationship information of the first and second sample neighborhoods can be fused with the partial graphic distribution sample density image of the corresponding neighborhoods to obtain sample density fusion information and sample light intensity fusion information, thereby obtaining etching input information. The specific fusion process is similar to the aforementioned fusion process and will not be repeated here.

[0117] The optional implementation methods of this application will be described below.

[0118] Figure 4 This is a schematic diagram of another load effect prediction method provided in an embodiment of this application, as shown below. Figure 4 As shown.

[0119] The chip design layout that needs to be etched can be obtained, and the chip design layout can be divided into multiple sub-regions according to a certain division size. The area of ​​each sub-region can be 1*1um.2 ~10*10um 2 Then, the ratio of the total area of ​​the graphic in each sub-region to the area of ​​the corresponding sub-region is calculated to obtain the graphic density of each sub-region. The graphic densities of each sub-region are then stitched together according to their positional relationship in the chip design layout to obtain the graphic distribution density image of the chip design layout.

[0120] The first neighborhood of the pattern to be predicted is delineated from the chip design layout. Based on the various process parameters used in the exposure process of the chip design layout, such as focal plane, photoresist parameters, light source parameters, and mask manufacturing parameters, an optical spatial image corresponding to the first neighborhood of the pattern to be predicted is obtained through an optical simulation model. The size of the optical spatial image can be as small as 1*1µm. 2 ~10*10um 2 between.

[0121] From the pattern density distribution image of the chip design layout, a partial pattern density distribution image corresponding to the second neighborhood of the pattern to be predicted is extracted. The size of the partial pattern density distribution image can be 1*1mm. 2 ~10*10mm 2 between.

[0122] It can also obtain positional relationship information of the first neighborhood and the second neighborhood. Specifically, it can be the positional deviation obtained by normalizing the distance between the predicted point of the image to be predicted and other pixels in the first neighborhood, and the positional deviation obtained by normalizing the distance between the predicted point of the image to be predicted and other pixels in the second neighborhood.

[0123] Subsequently, referring to the method of introducing positional relationship information in the large language model, the positional relationship information of the first neighborhood is added to the optical spatial image of the first neighborhood, and the positional relationship information of the second neighborhood is added to the partial graphic distribution density image of the second neighborhood to obtain the simulated etching input information of the graphic to be predicted.

[0124] The etching input information of the pattern to be predicted is input into the convolutional neural network to obtain the etching simulation result of the pattern to be predicted under the load effect.

[0125] Figure 5 This is a schematic diagram of the internal data flow of an etching prediction model provided in an embodiment of this application, such as... Figure 5 As shown.

[0126] The fused light intensity and density location information from multiple input channels are convolved with convolutional kernels to obtain convolutional features. Specifically, this can be understood as extracting local spatial features from the light intensity and density location fusion information. Subsequent multi-layer convolution and pooling operations can be performed. Multi-layer convolution allows the convolutional layers to extract more abstract spatial features, while pooling layers reduce the dimensionality of the extracted spatial feature maps, preserving key features while reducing computational load. Finally, the spatial features obtained from multi-layer convolution and pooling are input into a fully connected layer to obtain the etching simulation results of the pattern to be predicted under load effects.

[0127] Based on the load effect prediction method, this application also provides specific embodiments of the load effect prediction device.

[0128] like Figure 6 As shown, the load effect prediction device 600 provided in this application embodiment includes the following modules: The pattern distribution density image acquisition module 601 is used to acquire the pattern distribution density image of the chip design layout, wherein the chip design layout includes multiple patterns to be predicted; The optical spatial image determination module 602 is used to determine the optical spatial image corresponding to the first neighborhood of any of the predicted graphics. The simulated etching input information determination module 603 is used to determine the simulated etching input information of the pattern to be predicted based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, wherein the coverage range of the second neighborhood is greater than the coverage range of the first neighborhood. The etching simulation module 604 is used to input the etching input information to be simulated into a pre-trained etching prediction model to obtain the etching simulation results of the pattern to be predicted under the load effect.

[0129] As an optional embodiment, the input information determination module 603 to be simulated etching is specifically used to perform feature fusion of the positional relationship information of the predicted point of the pattern to be predicted in the first neighborhood with the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information. The positional relationship information of the predicted point in the second neighborhood of the graphic to be predicted is fused with the feature of the graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information; The density location fusion information and the light intensity location fusion information are used as the input information for the etching to be simulated.

[0130] As an optional embodiment, the positional relationship information of the predicted points of the image to be predicted within the first neighborhood is determined based on the following steps: Determine the distance between the predicted point of the graphic to be predicted and other pixels in the first neighborhood; The distances are normalized to obtain the positional deviation between each of the other pixels in the first neighborhood and the predicted point.

[0131] As an optional embodiment, the input information determination module 603 to be simulated etching is specifically used to add the position deviation degree of each of the other pixels in the first neighborhood from the prediction point and the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information; The simulated etching input information determination module 603 is specifically used to add the positional deviation of each of the other pixels in the second neighborhood from the predicted point and the density value of the corresponding pixel in the part of the graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information.

[0132] As an optional embodiment, the input information determination module 603 to be simulated etching is specifically used to perform feature splicing on the position deviation degree between each of the other pixels in the first neighborhood and the prediction point and the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood, so as to obtain light intensity position fusion information. The simulated etching input information determination module 603 is specifically used to perform feature splicing on the positional deviation of each of the other pixels in the second neighborhood from the predicted point and the density value of the corresponding pixel in the part of the graphic distribution density image corresponding to the second neighborhood, to obtain density position fusion information.

[0133] As an optional embodiment, the apparatus further includes: A region density determination module is used to divide the chip design layout into multiple sub-regions; The ratio of the total area of ​​all graphics within each sub-region to the total area of ​​the corresponding sub-region is determined to obtain the graphic density of the corresponding sub-region; The distribution of the graphic density of each sub-region in the chip design layout constitutes the graphic distribution density image.

[0134] As an optional embodiment, the etching prediction model is trained based on the following steps: Obtain the graphic distribution sample density image of the sample design layout and several optical spatial images corresponding to the first sample neighborhood, including the sample graphic. Obtain the etching simulation results of each sample pattern under the load effect, and use them as labels; For each sample pattern, based on the partial pattern distribution sample density image corresponding to the second sample neighborhood of the sample pattern and the optical spatial image corresponding to the first sample neighborhood of the sample pattern, etching input information is determined, wherein the coverage of the second sample neighborhood is greater than the coverage of the first sample neighborhood. The etching input information is input into the etching prediction model to obtain the etching result of the sample pattern; The etching prediction model is trained based on the etching simulation results of each sample pattern and the labels.

[0135] Based on the load effect prediction method, this application also provides specific embodiments of the load effect prediction device.

[0136] Figure 7 This is a schematic diagram of the hardware structure of a load effect prediction device provided in an embodiment of this application.

[0137] The load effect prediction device may include a processor 701 and a memory 702 storing computer program instructions.

[0138] Specifically, the processor 701 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0139] Memory 702 may include mass storage for data or instructions. For example, and not limitingly, memory 702 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 702 may include removable or non-removable (or fixed) media. Where appropriate, memory 702 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 702 is non-volatile solid-state memory.

[0140] The processor 701 reads and executes computer program instructions stored in the memory 702 to implement any of the load effect prediction methods in the above embodiments.

[0141] In one example, the electronic device may also include a communication interface 703 and a bus 704. Wherein, as... Figure 7 As shown, the processor 701, memory 702, and communication interface 703 are connected through bus 704 and complete communication with each other.

[0142] The communication interface 703 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0143] Bus 704 includes hardware, software, or both, that couples the components of the electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 704 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0144] Furthermore, in conjunction with the load effect prediction methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the load effect prediction methods in the above embodiments.

[0145] In addition, in conjunction with the load effect prediction method in the above embodiments, this application embodiment can provide a computer program product to implement it. When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the load effect prediction method provided by any aspect of the above embodiments of this application.

[0146] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0147] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0148] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0149] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0150] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for predicting load effects, characterized in that, include: Obtain a pattern distribution density image of a chip design layout, wherein the chip design layout includes multiple patterns to be predicted; For any of the aforementioned patterns to be predicted, determine the optical spatial image corresponding to the first neighborhood of the pattern to be predicted; Based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted, and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, the simulated etching input information of the pattern to be predicted is determined, wherein the coverage of the second neighborhood is greater than the coverage of the first neighborhood. The etching input information to be simulated is input into a pre-trained etching prediction model to obtain the etching simulation results of the pattern to be predicted under the load effect.

2. The method as described in claim 1, characterized in that, Based on the optical spatial image corresponding to the first neighborhood of the pattern to be predicted, and the partial pattern distribution density image corresponding to the second neighborhood of the pattern to be predicted, the simulated etching input information of the pattern to be predicted is determined, including: The positional relationship information of the predicted points of the image to be predicted in the first neighborhood is fused with the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information. The positional relationship information of the predicted point in the second neighborhood of the graphic to be predicted is fused with the feature of the graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information; The density location fusion information and the light intensity location fusion information are used as the input information for the etching to be simulated.

3. The method as described in claim 2, characterized in that, The positional relationship information of the predicted points of the image to be predicted within the first neighborhood is determined based on the following steps: Determine the distance between the predicted point of the graphic to be predicted and other pixels in the first neighborhood; The distances are normalized to obtain the positional deviation between each of the other pixels in the first neighborhood and the predicted point.

4. The method as described in claim 3, characterized in that, The positional relationship information of the predicted points of the image to be predicted within the first neighborhood is fused with the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information, including: The positional deviation of each other pixel in the first neighborhood from the predicted point is added to the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood to obtain the light intensity position fusion information. The positional relationship information of the predicted points of the graphic to be predicted within the second neighborhood is fused with the feature of the partial graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information, including: The positional deviation of each of the other pixels in the second neighborhood from the predicted point and the density value of the corresponding pixel in the part of the graphic distribution density image corresponding to the second neighborhood are added together to obtain density position fusion information.

5. The method as described in claim 3, characterized in that, The positional relationship information of the predicted points of the image to be predicted within the first neighborhood is fused with the optical spatial image corresponding to the first neighborhood to obtain light intensity position fusion information, including: The positional deviation of each other pixel in the first neighborhood from the predicted point and the dimensionless light intensity value of the corresponding pixel in the optical spatial image corresponding to the first neighborhood are concatenated to obtain light intensity position fusion information. The positional relationship information of the predicted points of the graphic to be predicted within the second neighborhood is fused with the feature of the partial graphic distribution density image corresponding to the second neighborhood to obtain density position fusion information, including: The positional deviation of each other pixel in the second neighborhood from the predicted point and the density value of the corresponding pixel in the partial graphic distribution density image corresponding to the second neighborhood are concatenated to obtain density position fusion information.

6. The method as described in claim 1, characterized in that, Also includes: The chip design layout is divided into multiple sub-regions; The ratio of the total area of ​​all graphics within each sub-region to the total area of ​​the corresponding sub-region is determined to obtain the graphic density of the corresponding sub-region; The distribution of the graphic density of each sub-region in the chip design layout constitutes the graphic distribution density image.

7. The method as described in claim 1, characterized in that, The etching prediction model is trained based on the following steps: Obtain the graphic distribution sample density image of the sample design layout and several optical spatial images corresponding to the first sample neighborhood, including the sample graphic. Obtain the etching simulation results of each sample pattern under the load effect, and use them as labels; For each sample pattern, based on the partial pattern distribution sample density image corresponding to the second sample neighborhood of the sample pattern and the optical spatial image corresponding to the first sample neighborhood of the sample pattern, etching input information is determined, wherein the coverage of the second sample neighborhood is greater than the coverage of the first sample neighborhood. The etching input information is input into the etching prediction model to obtain the etching result of the sample pattern; The etching prediction model is trained based on the etching simulation results of each sample pattern and the labels.

8. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the load effect prediction method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the load effect prediction method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs the load effect prediction method as described in any one of claims 1-7.

Images