Image processing method and device, computer device and computer readable storage medium

Abstract

The disclosure provides an image processing method, device, computer equipment and computer readable storage medium, and belongs to the technical field of semiconductors. The method comprises: obtaining a developed pattern and an etched pattern; processing the developed pattern and the etched pattern to obtain a developed pattern contour and an etched pattern contour, respectively; obtaining a target etching deviation between each candidate data point on the developed pattern contour and a corresponding data point on the etched pattern contour; obtaining M candidate features of each candidate data point on the developed pattern contour, M being an integer greater than 1; obtaining an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the developed pattern contour and the target etching deviation thereof; and determining M1 target features from the M candidate features according to the etching deviation correlation index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of semiconductor technology, and in particular, to an image processing method and device, a computer device, and a computer readable storage medium. BACKGROUND

[0002] In the process of integrated circuit manufacturing, etching is to remove a part of thin film layer that is not masked by a patterned mask layer, so as to obtain a pattern on the thin film layer that is completely consistent with the patterned mask layer. It is difficult to simulate the etching process physically. In the process of etching a material, whether it is a dry etching process or a wet etching process, it is coupled by various factors, such as the shape and density of the mask pattern, the diffusion of the etching substance, the flow of the etchant, the interface delamination, the chemical reaction, etc. Whether the deviation caused by etching can be accurately predicted to transfer the mask pattern to be transferred is one of the key steps to form a semiconductor device with required performance. However, in the related art, the amount of data used to train the etching deviation model is large, and there may be features with low correlation with the etching deviation in the data, thereby resulting in poor accuracy of the etching deviation model obtained by training the data. SUMMARY

[0003] The embodiment of the present disclosure provides an image processing method, which comprises: obtaining a developed pattern and an etched pattern; processing the developed pattern and the etched pattern to obtain a developed pattern contour and an etched pattern contour, respectively; obtaining a target etching deviation between each candidate data point on the developed pattern contour and a corresponding data point on the etched pattern contour; obtaining M candidate features of each candidate data point on the developed pattern contour, M being an integer greater than 1; obtaining an etching deviation correlation degree index of each candidate feature according to the candidate features of each candidate data point on the developed pattern contour and the target etching deviation thereof; and determining M1 target features from the M candidate features according to the etching deviation correlation degree index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

[0004] According to another aspect of the present disclosure, there is provided an image processing apparatus, comprising: an obtaining unit configured to obtain a post-development pattern and a post-etching pattern; a processing unit configured to process the post-development pattern and the post-etching pattern to obtain a post-development pattern contour and a post-etching pattern contour, respectively; the obtaining unit is further configured to obtain a target etching deviation between each candidate data point on the post-development pattern contour and a corresponding data point on the post-etching pattern contour; the obtaining unit is further configured to obtain M candidate features of each candidate data point on the post-development pattern contour, M being an integer greater than 1; the obtaining unit is further configured to obtain an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the post-development pattern contour and the target etching deviation thereof; and a determining unit configured to determine M1 target features from the M candidate features according to the etching deviation correlation index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

[0005] According to a further aspect of the present disclosure, there is provided a computer device, comprising one or more processors; and a memory configured to store one or more programs, which when executed by the one or more processors, cause the computer device to implement the image processing method in any of the embodiments of the present disclosure.

[0006] According to a further aspect of the present disclosure, there is provided a computer readable storage medium storing a computer program, which is adapted to be loaded and executed by a processor to cause a computer device having the processor to perform the image processing method in any of the embodiments of the present disclosure.

[0007] According to a further aspect of the present disclosure, there is provided a computer program product, which when executed by a processor, implements the image processing method in any of the embodiments of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 A flow chart of an image processing method provided by an embodiment of the present disclosure is shown.

[0009] Figure 2 A schematic diagram of a post-development pattern in an embodiment of the present disclosure is shown.

[0010] Figure 3 A schematic diagram of a post-etching pattern in an embodiment of the present disclosure is shown.

[0011] Figure 4 A schematic diagram of a post-development pattern contour and a post-etching pattern contour in an embodiment of the present disclosure is shown.

[0012] Figure 5 A schematic diagram of an etching deviation in an embodiment of the present disclosure is shown.

[0013] Figure 6 A schematic diagram showing etching bias in another embodiment of the present disclosure.

[0014] Figure 7 A schematic diagram showing target etching bias between each candidate data point on a post-development pattern profile and a corresponding data point on a post-etch pattern profile in an embodiment of the present disclosure.

[0015] Figure 8 A schematic diagram showing aperture effect in an embodiment of the present disclosure.

[0016] Figure 9 A schematic diagram showing micro-loading effect in an embodiment of the present disclosure.

[0017] Figure 10 A schematic diagram showing density kernel function in an embodiment of the present disclosure.

[0018] Figure 11 A schematic diagram showing visibility kernel function in an embodiment of the present disclosure.

[0019] Figure 12 A schematic diagram showing occlusion kernel function in an embodiment of the present disclosure.

[0020] Figure 13 A schematic diagram showing post-development pattern profile and post-etch pattern profile in another embodiment of the present disclosure.

[0021] Figure 14 A schematic diagram showing post-development pattern profile and post-etch pattern profile in yet another embodiment of the present disclosure.

[0022] Figure 15 A schematic diagram showing distribution of etching bias in an embodiment of the present disclosure.

[0023] Figure 16 A flowchart of another image processing method provided by an embodiment of the present disclosure is shown.

[0024] Figure 17 A schematic diagram of an image processing apparatus in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0025] As shown in Figure 1 The image processing method provided by an embodiment of the present disclosure can include the following steps.

[0026] In S110, a post-development pattern (After Developing Inspection, ADI) and a post-etch pattern (After Etching Inspection, AEI) are obtained.

[0027] For example, asFigure 2 Fig. 4 shows an ADI pattern obtained by SEM (scanning electron microscope). Figure 3 Fig. 5 shows an AEI pattern obtained by SEM.

[0028] Figure 3 Fig. 6 shows a schematic diagram of a post-etching pattern in an embodiment of the present disclosure. The diagram is an SEM pattern of AEI, which needs to be processed to obtain the profile of AEI.

[0029] In S120, the post-development pattern and the post-etching pattern are processed to obtain a post-development pattern profile and a post-etching pattern profile, respectively.

[0030] For example, the post-development pattern and the post-etching pattern shown in Figs. 4 and 5, respectively, are subjected to profile detection processing to obtain an ADI contour and an AEI contour. For example, as shown in Fig. 6, 41 represents the ADI contour, and 42 represents the AEI contour. Figure 2 For example, the post-development pattern and the post-etching pattern shown in Figs. 4 and 5, respectively, are subjected to profile detection processing to obtain an ADI contour and an AEI contour. For example, as shown in Fig. 6, 41 represents the ADI contour, and 42 represents the AEI contour. Figure 3 For example, the post-development pattern and the post-etching pattern shown in Figs. 4 and 5, respectively, are subjected to profile detection processing to obtain an ADI contour and an AEI contour. For example, as shown in Fig. 6, 41 represents the ADI contour, and 42 represents the AEI contour. Figure 4 For example, the post-development pattern and the post-etching pattern shown in Figs. 4 and 5, respectively, are subjected to profile detection processing to obtain an ADI contour and an AEI contour. For example, as shown in Fig. 6, 41 represents the ADI contour, and 42 represents the AEI contour.

[0031] In S130, a target etching deviation between each candidate data point on the post-development pattern profile and a corresponding data point on the post-etching pattern profile is obtained.

[0032] In an embodiment of the present disclosure, the ADI contour and the AEI contour are formed by connecting a plurality of data points in series. The candidate data points on the ADI contour can include all data points on the ADI contour, or can be part of the data points extracted from all data points on the ADI contour, such as part of the data points extracted at equal intervals or non-equal intervals. For example, data points extracted at longer intervals can be used as candidate data points for straight line portions on the ADI contour, and data points extracted at shorter intervals can be used as candidate data points for non-straight line portions (such as line ends, corners, arcs, etc.) on the ADI contour, so as to reduce the amount of data calculation and ensure that the etching deviation model trained by the training data set obtained by using the extracted candidate data points can have higher etching deviation prediction accuracy.

[0033] In an exemplary embodiment, obtaining a target etching deviation between each candidate data point on the post-development pattern profile and a corresponding data point on the post-etching pattern profile includes: obtaining a curvature of the candidate data point; drawing a straight line intersecting the post-etching pattern profile in a direction of the curvature of the candidate data point; determining a data point corresponding to the candidate data point from the intersection points on the post-etching pattern profile; and taking a distance between the candidate data point and the corresponding data point as the target etching deviation of the candidate data point.

[0034] From Figure 4 As can be seen, the ADI profile and the AEI profile are not the same, therefore, in order to accurately locate the candidate data points on the ADI profile and the corresponding data points on the AEI profile for calculating the target etching bias, the embodiments of the present disclosure calculate the curvature of each candidate data point, then draw a straight line along the curvature direction of the candidate data point, the straight line can have one intersection point or two intersection points with the AEI profile, when there are two intersection points, any one of the intersection points or the intersection point closer to the candidate data point can be selected as the corresponding data point of the candidate data point. Then the distance between the candidate data point on the ADI profile and the corresponding data point on the AEI profile is calculated as the target etching bias, i.e. the real etching bias, of the candidate data point. The embodiments of the present disclosure can accurately determine the candidate data points and the corresponding data points on the ADI profile and the AEI profile with irregular shapes (for example, with corners, circular arcs, etc.) by calculating the curvature of the candidate data points.

[0035] In the embodiments of the present disclosure, the curvature kernel function can be set as shown below:

[0036] (1)

[0037] In the above formula, k represents the curvature of a certain candidate data point on the ADI profile, (x, y) respectively represent the horizontal coordinate and the vertical coordinate of the candidate data point, and and respectively represent the first order derivative and the second order derivative of y with respect to x.

[0038] The first order derivative of y with respect to x can be calculated by the following formula:

[0039] (2)

[0040] In the above formula, (x1, y1) represents the horizontal coordinate and the vertical coordinate of a candidate data point before the candidate data point (x, y) on the ADI profile, (x2, y2) represents the horizontal coordinate and the vertical coordinate of a candidate data point after the candidate data point (x, y) on the ADI profile, i.e. the first order derivative of the candidate data point is calculated by the slope. It can be understood that the calculation method of the curvature of each candidate data point in the present disclosure is not limited to the above example.

[0041] As Figure 7 shown, the dashed line 71 represents the ADI profile, and the solid line 72 represents the AEI profile. 73 represents the target etching bias between each candidate data point on the ADI profile and the corresponding data point on the ADI profile.

[0042] In S140, M candidate features of each candidate data point on the developed pattern profile are obtained, M being an integer greater than 1. The way of obtaining M candidate features of each candidate data point on the ADI profile may be referred to the following description for example Figure 10 to Figure 12 but the disclosure is not limited to the type and way of obtaining M candidate features.

[0043] It should be noted that the execution sequence of S130 and S140 is not limited, and they can be executed in parallel, or S130 can be executed first and then S140, or S140 can be executed first and then S130.

[0044] In S150, according to the candidate features of each candidate data point on the developed pattern profile and the target etching deviation thereof, an etching deviation correlation degree index of each candidate feature is obtained.

[0045] In an exemplary embodiment, according to the candidate features of each candidate data point on the developed pattern profile and the target etching deviation thereof, an etching deviation correlation degree index of each candidate feature is obtained, comprising: processing M candidate features of each candidate data point through an etching deviation model to obtain an absolute etching deviation prediction value of each candidate data point; adding noise to each candidate feature of each candidate data point respectively, processing M candidate features of each candidate data point after adding noise through the etching deviation model to obtain a noisy etching deviation prediction value of each candidate feature of each candidate data point; and obtaining an etching deviation correlation degree index of each candidate feature according to the absolute etching deviation prediction value of each candidate data point, the noisy etching deviation prediction value of each candidate feature of each candidate data point, and the target etching deviation of each candidate data point.

[0046] In the example embodiment, the post-development pattern profile includes N candidate data points, N being an integer greater than 1. Wherein, according to the absolute etching deviation prediction value of each candidate data point, the noise-added etching deviation prediction value of each candidate feature of each candidate data point, and the target etching deviation of each candidate data point, the etching deviation correlation index of each candidate feature is obtained, including: obtaining the absolute etching deviation value of the i th candidate data point according to the absolute etching deviation prediction value of the i th candidate data point and the target etching deviation of the i th candidate data point; i is an integer greater than or equal to 1 and less than or equal to N, j is an integer greater than or equal to 1 and less than or equal to M; obtaining the noise-added etching deviation value of the j th candidate feature of the i th candidate data point according to the noise-added etching deviation prediction value of the j th candidate feature of the i th candidate data point and the target etching deviation of the i th candidate data point; obtaining the difference value of the j th candidate feature of the i th candidate data point according to the noise-added etching deviation value of the j th candidate feature of the i th candidate data point and the absolute etching deviation value of the i th candidate data point; obtaining the etching deviation correlation index of the j th candidate feature according to the difference value of the j th candidate feature of the N candidate data points.

[0047] The etching deviation model in the embodiments of the present disclosure refers to a model capable of predicting etching deviation, which can be obtained by training a machine learning model with a training data set. The trained machine learning model can be, for example, an ANN (Artificial Neural Network) model, a GBDT (Gradient Boosting Decision Tree) model, or any suitable model. The present disclosure does not limit this, and in the following embodiments, a GBDT model is used as an example. The GBDT model has better effect.

[0048] The M candidate features of each candidate data point are designed according to experience. Some candidate features can have strong correlation with etching deviation, and some candidate features can have no correlation or weak correlation with etching deviation. The method provided in the embodiments of the present disclosure filters and compresses the M candidate features, and selects M1 target features having strong correlation with etching deviation.

[0049] For example, M candidate features of each candidate data point on the ADI profile can be input into the GBDT model, the GBDT model processes the M candidate features of each candidate data point, and outputs an absolute etching deviation prediction value of each candidate data point. The absolute etching deviation prediction value of the i th candidate data point is subtracted from the target etching deviation of the i th candidate data point to obtain the absolute etching deviation value of the i th candidate data point. The absolute etching deviation values of the N candidate data points can be represented as [errOOB11, errOOB12, …, errOOB1N]. Wherein errOOB1i represents the absolute etching deviation prediction value of the i th candidate data point. The ADI profile here can be all ADI profiles of all ADI figures in the data set, and N is the number of all candidate data points on all ADI profiles of all ADI figures in the data set. N ]. Wherein errOOB1 i represents the absolute etching deviation prediction value of the i th candidate data point. The ADI profile here can be all ADI profiles of all ADI figures in the data set, and N is the number of all candidate data points on all ADI profiles of all ADI figures in the data set.

[0050] Then, the j th candidate feature F j of all samples (one sample is one candidate data point on the ADI profile) is randomly added with noise interference. For example, without noise, the i th candidate data point is represented as [F1, F2, …F j , …F M ], wherein F j represents the j th candidate feature in the i th candidate data point. After adding noise interference to the j th candidate feature of the i th candidate data point, it is represented as [F1, F2, …F j +n j , F j+1 ,…F M ], and nj represents the noise added to the j th candidate feature of the i th candidate data point. [F1, F2, …F j +n j , …F M ] is input into the GBDT model, the GBDT model processes [F1, F2, …F j +n j , …F M ] of the i th candidate data point, and outputs the noise-added etching deviation value of the j th candidate feature of the i th candidate data point, represented as errOOB2 ij , and the noise-added etching deviation values of the M candidate features of the i th candidate data point can be represented as [errOOB2 i1 , errOOB1 i2 , …, errOOB1 iM ]. In a similar manner, the noise-added etching deviation values of the M candidate features of the N candidate data points can be obtained.

[0051] For example, the etching deviation correlation index of the j th candidate feature can be calculated by the following formula:

[0052] (3)

[0053] It should be noted that the manner of obtaining the etching deviation correlation degree index of each candidate feature is not limited to the above examples, for example, it can also be obtained by the following manner:

[0054] (4)

[0055] Alternatively, it can also be obtained by the following manner:

[0056] (5)

[0057] In S160, M1 target features are determined from the M candidate features according to the etching deviation correlation degree index of each candidate feature, M1 is an integer greater than or equal to 1 and less than or equal to M.

[0058] The etching deviation correlation degree index of the M candidate features is obtained in the above manner [,, …,, …, ], which can be used to indicate the correlation degree of the candidate feature and the etching deviation, the greater the etching deviation correlation degree index, the greater the correlation degree of the corresponding candidate feature and the etching deviation, that is, the more important the candidate feature for the prediction of the etching deviation. The M etching deviation correlation degree indexes of the M candidate features can be arranged in ascending order or descending order according to the present embodiment of the disclosure, for example, taking descending order as an example, the first predetermined number of candidate features are selected as target features, and the predetermined number can be equal to M1. Alternatively, the first predetermined percentage of candidate features can also be selected as target features to obtain M1 target features, and the predetermined percentage can be set to balance accuracy and computational complexity, which is not limited by the present disclosure.

[0059] The image processing method provided by the present embodiment of the disclosure obtains the target etching deviation between each candidate data point on the ADI profile and the corresponding data point on the AEI profile, and obtains the etching deviation correlation degree index of each candidate feature according to the M candidate features of each candidate data point on the ADI profile and the target etching deviation thereof, so that M1 target features with strong correlation with the etching deviation can be selected from the M candidate features according to the etching deviation correlation degree index of the M candidate features, and the candidate features with weak correlation with the etching deviation are excluded, thereby reducing the data computational complexity. When the training and prediction of the etching deviation model are subsequently performed according to the M1 target features of each candidate data point on the ADI profile, on the one hand, the training speed and prediction speed of the etching deviation model can be improved, and the occupied computational resources and network transmission resources can be reduced; on the other hand, the accuracy of etching deviation prediction can be improved, so that the accuracy of simulation can be improved when simulating the semiconductor process, and more accurate process simulation results can be obtained.

[0060] The predicted etching bias in this disclosure can include two types: negative etch bias and positive etch bias. For example... Figure 5 As shown in (a), a photoresist layer is formed on the surface of the substrate 51. Based on a photolithography pattern (not shown), the photoresist layer is patterned, and a mask layer 52 is formed on the surface of the substrate 51, exposing a portion of the substrate 51's surface. This mask layer 52, or the boundary line of the photolithography pattern, serves as the ADI contour. Figure 5 As shown in (b), ions and radicals are implanted, and the substrate 51 is etched using the mask layer 52 as a mask, forming corresponding etch grooves within the substrate 51. The boundary line of these etch grooves on the surface of the substrate 51 serves as the AEI profile. After forming the etch grooves, the mask layer 52 can be removed to facilitate subsequent detection of the etch grooves and obtain the corresponding AEI profile. The patterned mask layer can determine the existence of preset etched and non-etched regions in the substrate 51. The preset etched regions are the areas in the substrate that are desired to be etched during the actual etching process, while the preset non-etched regions are the areas in the substrate that are not desired to be etched during the actual etching process. Figure 5 A negative etch bias 53 is formed in the etch grooves, meaning that the etch grooves include areas that are not intended to be etched. For example... Figure 6 As shown in (a), a mask layer 62 is formed on the surface of the substrate 61, exposing a portion of the surface of the substrate 61. Figure 6 As shown in (b), ions and free radicals are injected, and the substrate 61 is etched using the mask layer 62 as a mask, forming corresponding etching grooves in the substrate 61. Figure 6 A positive etch bias 63 was formed in the etched groove, meaning that the etched groove did not completely cover the area that was to be etched.

[0061] During the etching process, a loading effect exists, which refers to the decrease in etching rate or uneven distribution caused by localized consumption of etching gas exceeding supply. The loading effect can be categorized into three types: macro loading, micro loading, and aspect ratio-dependent etch (ARDE, also known as the aperture effect). In the embodiments of this disclosure, when designing candidate features for each candidate data point on the ADI profile, the etching loading effect can be comprehensively considered, for example, considering both the micro loading effect and the aperture effect.

[0062] Figure 8 A schematic diagram showing the aperture effect in the embodiments of the present disclosure. The aperture effect refers to that the width of the pitch of the litho pattern or mask layer will affect the etching depth. As shown in Figure 8 (a), it is assumed that different litho patterns 81 have the same line width W but different pitches S. From Figure 8 (b), it can be seen that the etching depth corresponding to the litho pattern with wide pitch is deeper, and the etching depth corresponding to the litho pattern with narrow pitch is shallower.

[0063] Figure 9 A schematic diagram showing the micro-loading effect in the embodiments of the present disclosure. The micro-loading effect refers to that the density of the litho pattern will affect the etching depth. As shown in Figure 9 (a), it is assumed that the left side litho patterns 91 have the same pitch and line width W2, and the right side litho patterns have the same pitch and line width, but the density of the left and right side litho patterns is different, the left side litho patterns are dense, and the right side litho patterns are sparse. From Figure 9 (b), it can be seen that the etching depth in the dense pattern area is smaller than that in the sparse pattern area.

[0064] Figure 10 A schematic diagram showing the density kernel function in the embodiments of the present disclosure. As shown in Figure 10 , a point of interest can be arbitrarily selected from the ADI profile as a candidate data point, and then a circle (the area in the circle is the interaction range of the point of interest) is drawn with the point of interest as the center and r1 as the radius, and the area ratio of the litho pattern (for example Figure 10 black rectangle in the circle) in the circle is calculated, which is the local density of the point of interest. The radius r1 can be set according to experience, and the present disclosure does not limit it. Through the density kernel function designed in the embodiments of the present disclosure, the local density of the candidate data point on the ADI profile can be obtained, which can be used as a candidate feature of the candidate data point.

[0065] Figure 11 A schematic diagram showing the visibility kernel function in the embodiments of the present disclosure. As shown in Figure 11 , a point of interest can be arbitrarily selected from the ADI profile as a candidate data point, and then a circle is drawn with the point of interest as the center and r2 as the radius, and if the litho pattern is encountered during the drawing of the circle, the drawing is stopped, thereby obtaining Figure 11The interaction range of the point of interest is shown, and the area ratio of the interaction range to the circle is the visibility of the point of interest. The radius r2 can be set according to experience, and the disclosure does not limit this. Through the visibility kernel function designed by the embodiment of the disclosure, the visibility of the candidate data point on the ADI contour can be obtained, and the visibility can be used as a candidate feature of the candidate data point.

[0066] Figure 12 A schematic diagram of the occlusion kernel function in the embodiment of the disclosure is shown. As shown in Figure 12 An arbitrary point of interest can be selected from the ADI contour as a candidate data point, and then a circle with r3 as the radius is drawn with the point of interest as the center. The overlapping area of the circle with the photolithography pattern corresponding to the ADI contour is the interaction range of the point of interest, and the area ratio of the interaction range to the circle is the occlusion degree of the point of interest. The radius r3 can be set according to experience, and the disclosure does not limit this. Through the occlusion kernel function designed by the embodiment of the disclosure, the occlusion degree of the candidate data point on the ADI contour can be obtained, and the occlusion degree can be used as a candidate feature of the candidate data point. It should be noted that Figure 10 to Figure 12 The three kernel functions shown are only used to illustrate how the candidate features of the candidate data points in the disclosure are designed, but the disclosure is not limited thereto.

[0067] Figure 13 And Figure 14 Both show the schematic diagram of the ADI contour and the AEI contour. The dashed line represents the ADI contour, and the solid line represents the AEI contour. From Figure 13 And Figure 14 And in combination with the above Figure 7 It can be learned that the shapes of the ADI contour and the AEI contour are various, depending on the actual design needs of the semiconductor device. At the same time, most of the ADI contour and the AEI contour are straight line parts or approximate straight line parts, and a small amount are line ends, corners, circular arcs and other non-straight line parts, and the line end and the corner part have a large etch bias during etching. As a result, the candidate data points of the straight line part are much more than the candidate data points of the non-straight line part, and this imbalance of data will cause the etching bias model to be sensitive to the training data during etching bias model training, and the effect of different training data is large, causing model overfitting. The prediction effect of the etching bias model trained on the etching bias of the straight line part of the ADI contour is good, while the etching bias effect on the non-straight line part of the ADI contour is poor.

[0068] In an example embodiment, the method provided by the present disclosure further includes: determining, from the candidate data points, target data points with target etching biases greater than an etching bias threshold; performing interpolation processing on the target data points to obtain interpolated data points; obtaining interpolation etching biases between the interpolated data points and corresponding data points on the etched pattern profile; and putting the target data points and their target etching biases and the interpolated data points and their interpolation etching biases into a training data set; and training an etching bias model using the training data set.

[0069] Suppose the etching bias threshold is 6 nm, and interpolation is performed on candidate data points with target etching biases greater than 6 nm or more, to ensure that candidate data points with few samples also have sufficient samples. For example, there are originally 20 candidate data points at a line end, and in order to make the small sample data more, interpolation is performed among the 20 points to obtain 30 points, and the interpolation etching biases of the corresponding interpolated data points are calculated and added to the training data set.

[0070] In an example embodiment of the present disclosure, the etching bias threshold can be set according to actual needs. For example, the mean of the actual etching biases of the candidate data points of all straight line portions on the ADI profile can be calculated, and the mean is used as the etching bias threshold. However, the present disclosure does not limit the setting method of the etching bias threshold. For example, the etching bias threshold can also be set according to experience. When the target etching bias of a candidate data point on the ADI profile is greater than the etching bias threshold, it can be determined that it is a data point of a non-straight line portion on the ADI profile. The candidate data points of the non-straight line portions are extracted as target data points, and interpolation processing is performed on the target data points, i.e., interpolation is performed on the candidate data points of the non-straight line portions such as line ends and corners to obtain interpolated data points. The target data points of the non-straight line portions and the interpolated data points are used as a training data set together. In this way, the sample data of the non-straight line portions such as line ends and corners can be expanded, and the sample data of the non-straight line portions and the sample data of the straight line portions are balanced, so that the etching bias model trained based on the training data set not only has good prediction effect on the etching bias of the straight line portions of the ADI profile, but also has good prediction effect on the etching bias of the non-straight line portions of the ADI profile. The problem that the etching bias model is sensitive to data is solved, and the long tail effect of the etching bias model during training is prevented. In addition, through interpolation processing, the sample data is increased, and more sample data is obtained. Training the etching bias model using more sample data can also solve the overfitting problem, because more samples can enable the etching bias model to learn more effective features.

[0071] In the embodiments of the present disclosure, the target data points and the interpolation data points can be put into the training data set as M candidate features or M1 target features of the target data points and the interpolation data points. When the M1 target features are selected to be put into the training data set, compared with the M candidate features, the complexity of the etching deviation model and the operation amount during the model training can be reduced, the training efficiency and speed can be improved, and the prediction accuracy of the etching deviation can be ensured.

[0072] In the example embodiments, the interpolation etching deviation between the interpolation data point and the corresponding data point on the etched pattern profile is obtained by obtaining the curvature of the interpolation data point, drawing a straight line intersecting the etched pattern profile along the direction of the curvature of the interpolation data point, determining the data point corresponding to the interpolation data point from the intersection points on the etched pattern profile, and taking the distance between the interpolation data point and the corresponding data point as the interpolation etching deviation of the interpolation data point.

[0073] In the embodiments of the present disclosure, for the interpolation data points on the non-linear part of the ADI profile, the data point corresponding to the interpolation data point on the AEI profile needs to be determined. The curvature of each interpolation data point can be calculated, a straight line intersecting the AEI profile is drawn along the direction of the curvature, and the intersection point can be one or two. When there are two intersection points, any one or the intersection point closest to the interpolation data point can be selected as the corresponding data point of the interpolation data point. The distance between the interpolation data point and the corresponding data point on the AEI profile is calculated as the interpolation etching deviation of the interpolation data point, so that the interpolation processing on the non-linear part of the AEI profile can be avoided.

[0074] Figure 15 A distribution diagram of etching deviation in an embodiment of the present disclosure is shown. Figure 15 The horizontal coordinate in the diagram is the value of the etching deviation, and the unit is nm (nanometer). The vertical coordinate is the number of candidate data points on the ADI profile with the etching deviation falling in the corresponding interval. From the diagram, it can be seen that the distribution of the candidate data points in each etching deviation value interval is not balanced, and the difference is large. Figure 15 It can be seen that the distribution of the candidate data points in each etching deviation value interval is not balanced, and the difference is large.

[0075] In an example embodiment, the method provided by the embodiments of the present disclosure can further include: performing clustering processing on the candidate data points according to the feature vectors of the candidate data points, the feature vectors of the candidate data points including M candidate features or M1 target features; performing first uniform sampling on the candidate data points in each cluster category, and putting the candidate data points and the target etching deviations thereof obtained by the first uniform sampling into a training data set; performing second uniform sampling on the remaining candidate data points after the first uniform sampling according to the target etching deviation intervals of the remaining candidate data points, and putting the candidate data points and the target etching deviations thereof obtained by the second uniform sampling into the training data set; and training an etching deviation model by using the training data set.

[0076] In the embodiments of the present disclosure, the k-means algorithm can be used to perform clustering processing on the candidate data points on the ADI profile, and k (k is a positive integer greater than 1, and k can be set according to actual needs, which is not limited in the present disclosure) initial centroids are selected as initial clusters (clusters, i.e. cluster categories); the feature vector of each candidate data point is calculated to obtain the nearest centroid, and the cluster category thereof is marked as the cluster corresponding to the centroid; the centroids corresponding to the k clusters are recalculated; and the above steps are repeatedly executed until the centroids no longer change to obtain k cluster categories. It can be understood that the algorithm used for clustering processing is not limited in the present disclosure, and the k-means is taken as an example in the following embodiments, but it is not limited thereto.

[0077] In the embodiments of the present disclosure, all candidate data points on the ADI profile are k-means clustered according to their respective feature vectors, and then the candidate data points (i.e. samples) in each cluster category are uniformly sampled. For example, assuming that there are 1000000 candidate data points in total, and there are 10 cluster categories after k-means clustering, 500 candidate data points are sampled from each of the 10 cluster categories, and the feature vectors and target etching deviations of the 5000 candidate data points obtained by the first sampling are put into a training data set as training samples. Through the first uniform sampling, on the one hand, the data amount can be reduced to speed up the training speed of the etching deviation model; on the other hand, the proportion of samples for each category can be ensured to be relatively uniform, so as to prevent the etching deviation model from being good for a certain type of pattern in the ADI profile and being poor for another type of pattern.

[0078] Then, the remaining candidate data points which are not sampled in the first sampling are secondly sampled according to the target etching bias histogram distribution (for example, the target etching bias histogram distribution of the ADI profile is divided into 10 intervals, and 500 candidate data points are sampled from each interval). Figure 15The histogram distribution is shown) is uniformly sampled, for example, 500 samples are randomly taken from the remaining candidate data points with a target etch bias value between 0.1-0.3 nanometers, 500 samples are taken from the remaining candidate data points with a target etch bias between 0.3-0.6 nm, and so on. For the samples, some samples have a large target etch bias, and some samples have a small target etch bias. The data amount of the samples with different target etch biases in the training data is different, which may cause the etching bias model obtained by training to be inclined to the sample with a large data amount. For example, 80% of the target etch biases in the training data are between 0.1-1.1 nm, and the model may be difficult to learn for the values of the target etch bias in other intervals. Therefore, the target etch bias value domain is divided, for example, 0.4 nm is taken as an interval, and each interval is sampled, for example, 0.0-0.4 nm, 0.4-0.8 nm, and 0.8-1.2 nm. This can make the sample amount in each target etching bias interval more balanced, and improve the prediction accuracy of the etching bias model.

[0079] In the embodiments of the present disclosure, for the verification data set, because it is necessary to keep consistent with the actual application scenario, no data interpolation and sampling processing is performed, so that the test result of the verification data set can truly reflect the application effect of the etching bias model.

[0080] The method provided by the embodiments of the present disclosure considers the micro-loading effect and the aperture effect of each candidate data point of the ADI contour by analyzing the ADI and the AEI, generates a series of candidate features by using empirical knowledge, selects target features that are strongly related to the etching bias by feature compression, samples and balances the generated input domain and value domain space, and finally forms a training data set and a verification data set by using the target features of the processed samples and the target etching bias. The GBDT training regression model is used as the etching bias model. The etching bias model obtained by training in the above manner has high accuracy, and the similarity between the predicted AEI contour pattern and the actual AEI contour pattern is high. The input domain here refers to ADI patterns with different characteristics, and the value domain refers to data pairs (ADI & AEI) with different etch biases. That is, the training data has a relatively large variety and a relatively uniform distribution of various etch biases. Therefore, the etching bias model obtained by training is more robust.

[0081] As Figure 16 The image processing method provided by the embodiments of the present disclosure can include the following steps.

[0082] In S201, the developed pattern contour and the etched pattern contour are obtained.

[0083] In the data preparation stage, the ADI profile and the AEI profile are obtained. The ADI profile can be obtained by directly photographing the wafer to extract the profile, or by using a photolithography model to simulate the profile. The AEI profile can be obtained by photographing the wafer to extract the contour.

[0084] In S202, the target etching deviation between each candidate data point on the developed pattern profile and the corresponding data point on the etched pattern profile is obtained.

[0085] For each candidate data point on the ADI profile, the target etching deviation with the corresponding data point on the AEI profile is calculated as the true etching deviation, which is used as a label in the etching deviation model training.

[0086] In S203, M candidate features of each candidate data point on the developed pattern profile are obtained.

[0087] In the embodiment of the present disclosure, a plurality of kernel functions such as density kernel function, visible kernel function, occlusion kernel function, and curvature kernel function can be designed by combining artificial experience with micro-loading effect and aperture effect. M candidate features of each candidate data point on the ADI profile are calculated as the feature vector of the candidate data point.

[0088] In S204, the M candidate features of each candidate data point on the developed pattern profile are compressed to screen M1 target features having strong correlation with the etching deviation.

[0089] In S205, the sample balancing processing is performed on the candidate data points on the developed pattern profile having the M1 target features to obtain a training data set.

[0090] In S206, the etching deviation model is trained using the training data set.

[0091] In S207, the M1 target features of each data point on the developed pattern profile to be predicted are input into the etching deviation model to obtain the etching deviation prediction value of each data point.

[0092] In S208, each data point on the etched pattern profile to be predicted is determined according to each data point on the developed pattern profile to be predicted and the etching deviation prediction value thereof.

[0093] In S209, the etched pattern profile to be predicted is obtained by connecting each data point on the etched pattern profile to be predicted.

[0094] In the embodiments of the present disclosure, the prepared training data set is trained using a GBDT model, taking the error between the target etch bias and the predicted bias of the samples in the training data set as the loss, and updating the GBDT model parameters through back propagation until the model converges and the error reaches a minimum value and no longer decreases. After keeping the model parameters, when the AEI profile needs to be predicted, the feature vector corresponding to each data point on the ADI profile to be predicted is input into the trained etch bias model to obtain the etch bias prediction value. The data points on the ADI profile are offset based on the etch bias prediction value, and the data points are concatenated as a profile, thereby obtaining the AEI profile to be predicted.

[0095] The method provided by the embodiments of the present disclosure filters out target features through feature compression, the target features have a strong correlation with the etch bias model, so as to reduce the operation complexity and the operation amount, make the training of the etch bias model easier, shorten the training time, and improve the prediction accuracy of the etch bias. In addition, the sampling and interpolation of the samples solve the problem of unbalanced sample data, prevent the model from overfitting, and improve the prediction accuracy of the etch bias.

[0096] The embodiments of the present disclosure adopt the feature screening method, prevent unreasonable factors existing in the feature kernel function designed by artificial experience from affecting the model effect, and consider the uniform sampling of data based on the graph features, prevent the model from overfitting due to the long tail effect, and the trained etch bias model has better robustness and is closer to the real graph in the prediction of special graphs. Based on the same inventive concept, the present disclosure also provides an image processing device, as described in the following embodiments. Since the principles of the device embodiments for solving problems are similar to those of the above-mentioned method embodiments, the implementation of the device embodiments can be referred to the implementation of the above-mentioned method embodiments, and the repeated parts will not be described here.

[0097] Figure 17 A schematic diagram of an image processing device in the embodiments of the present disclosure is shown as Figure 17As shown, the image processing apparatus 170 can include an obtaining unit 171, a processing unit 172, and a determining unit 173. The obtaining unit 171 is configured to obtain a developed pattern and an etched pattern. The processing unit 172 is configured to process the developed pattern and the etched pattern to obtain a developed pattern contour and an etched pattern contour, respectively. The obtaining unit 171 is further configured to obtain a target etching deviation between each candidate data point on the developed pattern contour and a corresponding data point on the etched pattern contour. The obtaining unit 171 is further configured to obtain M candidate features of each candidate data point on the developed pattern contour, M being an integer greater than 1. The obtaining unit 171 is further configured to obtain an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the developed pattern contour and the target etching deviation thereof. The determining unit 173 is configured to determine M1 target features from the M candidate features according to the etching deviation correlation index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

[0098] The computer device in the embodiments of the present disclosure can include one or more processors, memories, and input / output interfaces. The processor, memory, and input / output interface are connected through a bus. The memory is configured to store a computer program including program instructions, and the input / output interface is configured to receive and output data, such as for data interaction between a host and the computer device, or for data interaction between virtual machines in the host; and the processor is configured to execute the program instructions stored in the memory.

[0099] The processor can perform the following operations: obtaining a developed pattern and an etched pattern; processing the developed pattern and the etched pattern to obtain a developed pattern contour and an etched pattern contour, respectively; obtaining a target etching deviation between each candidate data point on the developed pattern contour and a corresponding data point on the etched pattern contour; obtaining M candidate features of each candidate data point on the developed pattern contour, M being an integer greater than 1; obtaining an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the developed pattern contour and the target etching deviation thereof; and determining M1 target features from the M candidate features according to the etching deviation correlation index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

[0100] The memory can include read-only memory and random access memory, and provide instructions and data to the processor and input-output interface. A part of the memory can also include non-volatile random access memory. In a specific implementation, the computer device can execute the implementation manners provided by each step in any method embodiment as described above through the built-in function modules thereof, and the specific implementation can refer to the implementation manners provided by each step in the diagrams shown in the method embodiments described above, which will not be described herein again.

[0101] The embodiments of the present disclosure provide a computer device, which comprises a processor, an input-output interface, and a memory. The computer program in the memory is acquired by the processor, and each step of the method shown in any embodiment described above is executed.

[0102] The embodiments of the present disclosure also provide a computer readable storage medium, which stores a computer program. The computer readable storage medium stores a program product capable of implementing the method of the present disclosure. The computer program is suitable for being loaded by the processor and executing the image processing method provided by each step in any embodiment described above.

[0103] The embodiments of the present disclosure also provide a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in any of the various optional manners in the embodiments described above.

Claims

1. An image processing method, characterized by, The method comprises the following steps: obtaining a post-development pattern and a post-etching pattern; processing the post-development pattern and the post-etching pattern to obtain a post-development pattern profile and a post-etching pattern profile respectively; obtaining a target etching deviation between each candidate data point on the post-development pattern profile and a corresponding data point on the post-etching pattern profile; obtaining M candidate features of each candidate data point on the post-development pattern profile, M being an integer greater than or equal to 1; obtaining an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the post-development pattern profile and the target etching deviation thereof; the etching deviation correlation index of each candidate feature comprises: processing the M candidate features of each candidate data point through an etching deviation model to obtain an absolute etching deviation prediction value of each candidate data point; adding noise to each candidate feature of each candidate data point respectively, processing the M candidate features of each candidate data point after adding noise through the etching deviation model to obtain a noise-added etching deviation prediction value of each candidate feature of each candidate data point, and obtaining the etching deviation correlation index of each candidate feature according to the absolute etching deviation prediction value of each candidate data point, the noise-added etching deviation prediction value of each candidate feature of each candidate data point, and the target etching deviation of each candidate data point; determining M1 target features from the M candidate features according to the etching deviation correlation index of each candidate feature, M1 being an integer greater than or equal to 1 and less than or equal to M.

2. The method of claim 1, wherein, the post-development pattern profile comprises N candidate data points, N being an integer greater than 1; wherein, obtaining the etching deviation correlation index of each candidate feature according to the absolute etching deviation prediction value of each candidate data point, the noise-added etching deviation prediction value of each candidate feature of each candidate data point, and the target etching deviation of each candidate data point comprises: obtaining an absolute etching deviation value of the ith candidate data point according to the absolute etching deviation prediction value of the ith candidate data point and the target etching deviation of the ith candidate data point, i being an integer greater than or equal to 1 and less than or equal to N, and j being an integer greater than or equal to 1 and less than or equal to M; obtaining a noise-added etching deviation value of the jth candidate feature of the ith candidate data point according to the noise-added etching deviation prediction value of the jth candidate feature of the ith candidate data point and the target etching deviation of the ith candidate data point; obtaining a difference value of the jth candidate feature of the ith candidate data point according to the noise-added etching deviation value of the jth candidate feature of the ith candidate data point and the absolute etching deviation value of the ith candidate data point; obtaining the etching deviation correlation index of the jth candidate feature according to the difference values of the jth candidate feature of the N candidate data points.

3. The method of claim 1, wherein obtaining the target etching deviation between each candidate data point on the post-development pattern profile and a corresponding data point on the post-etching pattern profile comprises: obtaining a curvature of the candidate data point; drawing a straight line intersecting the post-etching pattern profile in the direction of the curvature of the candidate data point; determining a data point corresponding to the candidate data point from intersection points on the post-etching pattern; taking a distance between the candidate data point and the corresponding data point as a target etching deviation of the candidate data point.

4. The method of claim 1, wherein Further comprising: determining a target data point with a target etching deviation greater than an etching deviation threshold from the candidate data points; performing interpolation processing on the target data point to obtain an interpolated data point; obtaining an interpolated etching deviation between the interpolated data point and a corresponding data point on the post-etching pattern; putting the target data point and the target etching deviation thereof and the interpolated data point and the interpolated etching deviation thereof into a training data set; and training an etching deviation model by using the training data set.

5. The method of claim 4, wherein The method further comprises: obtaining a curvature of the interpolated data point; drawing a straight line intersecting the post-etching pattern along a direction of the curvature of the interpolated data point; determining a data point corresponding to the interpolated data point from intersection points on the post-etching pattern; taking a distance between the interpolated data point and the corresponding data point as an interpolated etching deviation of the interpolated data point.

6. The method according to any one of claims 1 to 5, wherein Further comprising: performing clustering processing on the candidate data points according to feature vectors of the candidate data points, the feature vectors of the candidate data points comprising M candidate features or M1 target features; performing first uniform sampling on the candidate data points in each cluster category, and putting the first uniformly sampled candidate data points and target etching deviations thereof into a training data set; performing second uniform sampling on the remaining candidate data points after the first uniform sampling according to target etching deviations of the remaining candidate data points, and putting the second uniformly sampled candidate data points and target etching deviations thereof into the training data set; and training an etching deviation model by using the training data set.

7. An image processing apparatus characterized by comprising: The method comprises: obtaining a post-development pattern and a post-etching pattern; processing the post-development pattern and the post-etching pattern to obtain a post-development pattern contour and a post-etching pattern contour, respectively; obtaining a target etching deviation between each candidate data point on the post-development pattern contour and a corresponding data point on the post-etching pattern contour; obtaining M candidate features of each candidate data point on the post-development pattern contour, M being an integer greater than or equal to 1; obtaining an etching deviation correlation index of each candidate feature according to the candidate features of each candidate data point on the post-development pattern contour and target etching deviations thereof; and training an etching deviation model by using the training data set. The etching bias correlation degree indicator of each candidate feature comprises: processing the M candidate features of each candidate data point through an etching bias model to obtain an absolute etching bias prediction value of each candidate data point; adding noise to each candidate feature of each candidate data point respectively, processing the M candidate features of each candidate data point after adding noise through the etching bias model to obtain a noise-added etching bias prediction value of each candidate feature of each candidate data point; and obtaining the etching bias correlation degree indicator of each candidate feature according to the absolute etching bias prediction value of each candidate data point, the noise-added etching bias prediction value of each candidate feature of each candidate data point, and a target etching bias of each candidate data point. The determining unit is configured to determine M1 target features from the M candidate features according to the etching bias correlation degree indicators of the M candidate features, where M1 is an integer greater than or equal to 1 and less than or equal to M.

8. A computer device, comprising: The computer device comprises: one or more processors; a memory configured to store one or more programs, which, when executed by the one or more processors, cause the computer device to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 8. The computer program, when running on a computer, causes the computer to perform the method according to any one of claims 1 to 6.

Images