Optical proximity correction method and device, equipment and storage medium

Abstract

The invention relates to the field of integrated circuit production, in particular to an optical proximity correction method, device and equipment and a storage medium. Inputting the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting a random effect, and outputting a corresponding average photoresist image; reducing and improving the brightness of the average photoresist image according to preset random effect model parameters to obtain a first changed photoresist image and a second changed photoresist image; obtaining an inner ring contour edge in the first changed photoresist image according to a preset model threshold value, and obtaining an outer ring contour edge in the second changed photoresist image according to the model threshold value; and taking an area between the inner ring contour edge and the outer ring contour edge as a probability contour band after photoetching. According to the method, the hot spot prediction of the large-scale integrated circuit chip layout is realized on the premise of not obviously increasing the operation time and the calculation power occupation.

Description

Technical Field

[0001] The present invention relates to the field of integrated circuit production, and in particular to an optical proximity correction method, device, equipment, storage medium, a training method and device for an optical proximity correction model capable of predicting random effects. Background Art

[0002] An optical proximity correction (OPC) model predicts the post-lithography profile of a mask pattern based on given lithography process parameters, and is widely used in mask pattern correction and hot spot prediction of large-scale integrated circuits.

[0003] Although existing optical proximity correction models can give relatively accurate predictions of the contour lines of mask patterns after lithography, they cannot simulate the random effects in lithography. There are mainly two sources of random effects in the lithography process. First, light has wave-particle duality, and the number of photons incident on the photoresist fluctuates. This effect is also called shot noise of photons, which will cause relative fluctuations in exposure energy. Especially for extreme ultraviolet (EUV) lithography with a short light source wavelength, its photon density is very low, so the shot noise of photons increases significantly compared with deep ultraviolet (DUV) lithography. Second is chemical randomness, including factors such as uneven components in the photoresist, the migration distance and position uncertainty of photoacid during post-baking. The random effects in the lithography process will cause uneven feature sizes and rough edges of the patterns on the wafer after lithography, and even lead to serious defects such as pinching and bridging. Even if the probability of lithography random effect defects is as low as one in a million order of magnitude, for chip design layouts with millions or more repeated patterns such as memories, it is also a problem that cannot be ignored. In the prior art, to simulate lithography random effects, only the Monte Carlo method can be used for calculation, which is very time-consuming and consumes a huge amount of computing power.

[0004] Therefore, how to solve the problem in the prior art that it is impossible to predict hot spots caused by lithography random effects at a low cost and high speed in optical proximity correction has become an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide an optical proximity correction method, device, equipment, storage medium, a training method and device for an optical proximity correction model capable of predicting random effects, so as to solve the problem in the prior art that it is impossible to predict hot spots caused by lithography random effects at a low cost and high speed in optical proximity correction.

[0006] To solve the above technical problems, the present invention provides an optical proximity correction method, including:

[0007] Receive a mask pattern to be processed;

[0008] Input the mask pattern to be processed into a pre-trained optical proximity correction model that can predict random effects, and output a corresponding average photoresist image;

[0009] According to preset random effect model parameters, increase the brightness of the average photoresist image to obtain a first modified photoresist image;

[0010] According to preset random effect model parameters, decrease the brightness of the average photoresist image to obtain a second modified photoresist image;

[0011] In the first modified photoresist image, obtain an inner circle contour edge according to a preset model threshold, and in the second modified photoresist image, obtain an outer circle contour edge according to the model threshold;

[0012] Take the area between the inner circle contour edge and the outer circle contour edge as the post-lithography probability contour band.

[0013] Optionally, in the optical proximity correction method, after outputting the average photoresist image, it further includes:

[0014] In the average photoresist image, obtain an average contour edge according to a preset model threshold;

[0015] Correspondingly, after obtaining the post-lithography probability contour band, it further includes:

[0016] Add the average contour edge to the post-lithography probability contour band.

[0017] An optical proximity correction device includes:

[0018] A receiving module, configured to receive a mask pattern to be processed;

[0019] An average image module, configured to input the mask pattern to be processed into a pre-trained optical proximity correction model that can predict random effects, and output a corresponding average photoresist image;

[0020] A brightness increasing module, configured to increase the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image;

[0021] A brightness decreasing module, configured to decrease the brightness of the average photoresist image according to preset random effect model parameters to obtain a second modified photoresist image;

[0022] Inner and outer ring edge module, configured to obtain an inner ring contour edge in the first changed photoresist image according to a preset model threshold, and obtain an outer ring contour edge in the second changed photoresist image according to the model threshold;

[0023] Band division module, configured to use the area between the inner ring contour edge and the outer ring contour edge as the post-lithography probability contour band.

[0024] A training method for an optical proximity correction model capable of predicting random effects, comprising:

[0025] Obtain the target average contour, target outer ring contour and target inner ring contour of the target pattern;

[0026] Calibrate the linear coefficient and model threshold corresponding to the photoresist model term of the model to be trained according to the target average contour to obtain a calibrated model;

[0027] Input the target pattern into the calibrated model to obtain a target photoresist image;

[0028] Based on the target photoresist image, calibrate the random effect model parameters, so that in the target photoresist image with the brightness increased according to the calibrated random effect model parameters, the edge placement error between the simulated inner ring contour obtained according to the calibrated model threshold and the target inner ring contour is less than a preset tolerance value, and / or in the target photoresist image with the brightness decreased according to the calibrated random effect model parameters, the edge placement error between the simulated outer ring contour obtained according to the calibrated model threshold and the target outer ring contour is less than a preset tolerance value.

[0029] Optionally, in the training method for the optical proximity correction model capable of predicting random effects, the method for calibrating the random effect model parameters includes:

[0030] Determine that when the loss function value in the following formula is the smallest, the corresponding random effect model parameters are the calibrated random effect model parameters:

[0031] ;

[0032] where cost is the loss function value; EPE ik is the edge placement error between the simulated inner ring contour obtained according to the calibrated model threshold and the target inner ring contour in the target photoresist image with the brightness increased according to the random effect model parameters; EPE ok is the edge placement error between the simulated outer ring contour obtained according to the calibrated model threshold and the target outer ring contour in the target photoresist image with the brightness decreased according to the random effect model parameters; w k is the weight coefficient.

[0033] Optionally, in the training method of the optical proximity correction model with predictable random effects, after completing the calibration of the linear coefficient, the model threshold, and the random effect model parameters, it further includes:

[0034] Using the calibrated linear coefficient, the model threshold, and the random effect model parameters, iterate the parameters of the photoresist model term in the calibration model to minimize the loss function value corresponding to the calibration model corresponding to the parameters of the iterated photoresist model term.

[0035] Optionally, in the training method of the optical proximity correction model with predictable random effects, obtaining the target average profile, the target outer profile, and the target inner profile of the target pattern includes:

[0036] Scanning the scanned images formed by a plurality of mask patterns corresponding to the target pattern under the exposure field;

[0037] Extracting the profile of the scanned image to obtain the corresponding wafer profile;

[0038] According to all the wafer profiles, calculate the average position information of each preset point on the wafer profile to obtain the target average profile;

[0039] Calculate the position of each preset point, and the distance from each wafer profile to the average wafer profile in the normal direction as the profile distance set of the preset point;

[0040] Perform one-dimensional Gaussian function fitting on the distances from each wafer profile to the average wafer profile in the profile distance set to obtain a profile probability Gaussian function;

[0041] Determine the standard deviation of the profile probability Gaussian function, and move the corresponding preset point on the average wafer profile inward by the distance corresponding to the standard deviation in the normal direction to obtain the inner circle profile point corresponding to the preset point, and move the corresponding preset point on the average wafer profile outward by the distance corresponding to the standard deviation in the normal direction to obtain the outer circle profile point corresponding to the preset point;

[0042] Perform fitting on the inner circle profile points corresponding to all the preset points to obtain the target inner circle profile;

[0043] Perform fitting on the outer circle profile points corresponding to all the preset points to obtain the target outer circle profile.

[0044] An apparatus for training an optical proximity correction model with predictable random effects, comprising:

[0045] An acquisition module, configured to obtain a target average profile, a target outer profile, and a target inner profile of a target pattern;

[0046] A model calibration module, configured to calibrate the linear coefficient and the model threshold corresponding to the photoresist model term of the to-be-trained model according to the target average profile, so as to obtain a calibrated model;

[0047] A target image module, configured to input the target pattern into the calibrated model to obtain a target photoresist image;

[0048] A parameter calibration module, configured to calibrate the random effect model parameters based on the target photoresist image, so that in the target photoresist image with the brightness enhanced according to the calibrated random effect model parameters, the edge placement error between the simulated inner circle profile obtained according to the calibrated model threshold and the target inner circle profile is less than a preset allowable value, and / or in the target photoresist image with the brightness reduced according to the calibrated random effect model parameters, the edge placement error between the simulated outer circle profile obtained according to the calibrated model threshold and the target outer circle profile is less than a preset allowable value.

[0049] An optical proximity correction device, comprising:

[0050] A memory, configured to store a computer program;

[0051] A processor, configured to implement the steps of any one of the above optical proximity correction methods when executing the computer program.

[0052] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of any one of the above optical proximity correction methods, and / or the steps of the training method of the optical proximity correction model capable of predicting random effects as described in any one of the above.

[0053] The optical proximity correction method provided by the present invention includes receiving a mask pattern to be processed; inputting the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting random effects, and outputting a corresponding average photoresist image; enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image; reducing the brightness of the average photoresist image according to the preset random effect model parameters to obtain a second modified photoresist image; in the first modified photoresist image, obtaining an inner circle profile edge according to a preset model threshold, and in the second modified photoresist image, obtaining an outer circle profile edge according to the model threshold; taking the area between the inner circle profile edge and the outer circle profile edge as a post-lithography probability profile band.

[0054] Based on the simulated output of the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner circle contour edge and outer circle contour edge in the brightness-adjusted photoresist image through a preset model threshold. This is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs based on the speed of brightness change at each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a certain probability based on the finally output post-lithography probability contour band. Without significantly increasing the computing time and computing power occupancy, the hot spot prediction of the large-scale integrated circuit chip layout is realized. The present invention also provides an optical proximity correction device, equipment, storage medium, and a training method and device for an optical proximity correction model capable of predicting random effects with the above beneficial effects. Description of the Drawings

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0056] Figure 1 It is a schematic flowchart of a specific implementation manner of the optical proximity correction method provided by the present invention;

[0057] Figure 2 It is a brightness curve of the normal direction at a certain position in a specific implementation manner of the optical proximity correction method provided by the present invention;

[0058] Figure 3 It is a schematic structural diagram of a specific implementation manner of the optical proximity correction device provided by the present invention;

[0059] Figure 4 It is a schematic flowchart of a specific implementation manner of the training method of the optical proximity correction model capable of predicting random effects provided by the present invention;

[0060] Figure 5 It is a schematic process structure diagram of a specific implementation manner of the training method of the optical proximity correction model capable of predicting random effects provided by the present invention;

[0061] Figure 6 It is a schematic diagram of the one-dimensional Gaussian function fitting of the contour distance set at a preset point in a specific implementation manner of the training method of the optical proximity correction model capable of predicting random effects provided by the present invention;

[0062] Figure 7Schematic flowchart of a specific implementation of the training device for the optical proximity correction model with predictable random effects provided by the present invention.

[0063] Reference numerals:

[0064] 110 - Receiving module; 120 - Average image module; 130 - Brightness enhancement module; 140 - Brightness reduction module; 150 - Inner and outer circle edge module; 160 - Banding module; 210 - Acquisition module; 220 - Flat model calibration module; 230 - Target image module; 240 - Parameter calibration module. Specific implementation

[0065] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementations. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0066] The core of the present invention is to provide an optical proximity correction method. A schematic flowchart of a specific implementation thereof is as Figure 1 shown, which is called Specific Implementation 1 and includes:

[0067] S101: Receive the mask pattern to be processed.

[0068] That is, the staff needs to know what shape the pattern presented on the wafer after lithography of the mask pattern to be processed is, and correspondingly, where defect hotspots may be generated due to lithography random effects.

[0069] S102: Input the mask pattern to be processed into the pre-trained optical proximity correction model with predictable random effects, and output the corresponding average photoresist image.

[0070] The average photoresist image in the present invention is the light intensity distribution on the wafer after the mask pattern to be processed is illuminated through simulation. That is, the average photoresist image is not the shape of the pattern edge, but an image reflecting the light intensity at different positions within the range. Correspondingly, if the random effects in lithography are not considered, the edge of the post-lithography image generated by the mask pattern to be processed on the wafer can be outlined on the average photoresist image using the trained model threshold (brightness threshold).

[0071] S103: Enhance the brightness of the average photoresist image according to the preset random effect model parameters to obtain the first modified photoresist image.

[0072] S104: According to the preset random effect model parameters, reduce the brightness of the average photoresist image to obtain a second modified photoresist image.

[0073] Of course, there is no strict sequence between step S103 and step S104. It can be selected according to actual needs whether to first increase the brightness of the average photoresist image or first reduce the brightness of the average photoresist image, and the present invention does not make a limitation here.

[0074] S105: In the first modified photoresist image, obtain the inner circle contour edge according to the preset model threshold, and in the second modified photoresist image, obtain the outer circle contour edge according to the model threshold.

[0075] If the model threshold is directly applied to the average photoresist image, the edge of the mask pattern to be processed after lithography can be obtained without considering the lithography random effect. In this step, further by adjusting the brightness of the average photoresist image, the edge position corresponding to the model threshold moves inwards and outwards. At this time, in the average photoresist image, different brightness change rates are reflected in the position changes of the inner circle contour edge and the outer circle contour edge. Specifically, the faster the brightness changes, the greater the position difference between the inner circle contour edge and the outer circle contour edge relative to the model threshold before brightness adjustment, which also reflects that the area where random effects may occur corresponding to this position is larger, and vice versa. The slower the brightness changes, the smaller the position difference between the inner circle contour edge and the outer circle contour edge relative to the model threshold before brightness adjustment, which also reflects that the area where random effects may occur corresponding to this position is smaller.

[0076] S106: Take the area between the inner circle contour edge and the outer circle contour edge as the post-lithography probability contour band.

[0077] As a preferred implementation, after outputting the average photoresist image, it further includes:

[0078] A1: In the average photoresist image, obtain the average contour edge according to the preset model threshold.

[0079] Correspondingly, after obtaining the post-lithography probability contour band, it further includes:

[0080] A2: Add the average contour edge in the post-lithography probability contour band.

[0081] As described above, directly outlining according to the model threshold in the average photoresist image can obtain the average contour edge, and the average contour edge represents the edge of the pattern formed by the mask pattern to be processed on the wafer without considering the random effects in lithography.

[0082] Regarding the post-lithography probability profile band carrying the average contour edge, reference can be made to Figure 2 , Figure 2 which shows, in the form of a curve graph, the brightness curve of a certain position of the mask pattern to be processed along the corresponding normal direction. The vertical axis is the brightness, and the horizontal axis is the position in the normal direction. The three curves from top to bottom are the brightness curves in the normal direction corresponding to the first modified photoresist image, the brightness curve in the normal direction corresponding to the average photoresist image, and the brightness curve in the normal direction corresponding to the second modified photoresist image. Correspondingly, X 0 represents the position corresponding to the average contour edge, X 1 represents the position corresponding to the inner contour edge, X- 1 represents the position corresponding to the outer contour edge. The post-lithography probability profile band reflects the possible banded area where the edge of the pattern may appear considering the lithography random effect. In this preferred embodiment, the average contour edge is further added to the post-lithography probability profile band. The occurrence probability of the post-lithography edge position of the mask pattern to be processed considering the post-lithography random effect increases with the increase in the distance from the average contour edge, which is convenient for the staff to visually judge hotspots.

[0083] The optical proximity correction method provided by the present invention includes receiving a mask pattern to be processed; inputting the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting random effects to output a corresponding average photoresist image; enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image; reducing the brightness of the average photoresist image according to the preset random effect model parameters to obtain a second modified photoresist image; obtaining an inner contour edge according to a preset model threshold in the first modified photoresist image, and obtaining an outer contour edge according to the model threshold in the second modified photoresist image; and taking the area between the inner contour edge and the outer contour edge as the post-lithography probability profile band. Based on the simulation output of the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner contour edge and outer contour edge in the photoresist image after brightness adjustment through the preset model threshold, which is equivalent to inferring the potential hotspot distribution range when the lithography random effect occurs according to the speed of brightness change at each position in the average photoresist image. The staff can also predict whether there are defects such as bridging or breakpoints on the layout with a high probability by means of the finally output post-lithography probability profile band, and realizes the hotspot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupancy.

[0084] The optical proximity correction device provided by the embodiments of the present invention will be introduced below. The optical proximity correction device described below can be correspondingly referred to the optical proximity correction method described above.

[0085] Figure 3 It is a structural block diagram of the optical proximity correction device provided by the embodiments of the present invention, which is called the second specific implementation manner. Refer to Figure 3 The optical proximity correction device may include:

[0086] A receiving module 110, configured to receive a mask pattern to be processed;

[0087] An average image module 120, configured to input the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting random effects, and output a corresponding average photoresist image;

[0088] A brightness enhancement module 130, configured to enhance the brightness of the average photoresist image according to preset random effect model parameters, and obtain a first modified photoresist image;

[0089] A brightness reduction module 140, configured to reduce the brightness of the average photoresist image according to preset random effect model parameters, and obtain a second modified photoresist image;

[0090] An inner and outer circle edge module 150, configured to obtain an inner circle contour edge in the first modified photoresist image according to a preset model threshold, and obtain an outer circle contour edge in the second modified photoresist image according to the model threshold;

[0091] A banding module 160, configured to use the area between the inner circle contour edge and the outer circle contour edge as a post-lithography probability contour band.

[0092] As a preferred implementation manner, the average image module 120 further includes:

[0093] An average contour unit, configured to obtain an average contour edge in the average photoresist image according to a preset model threshold;

[0094] Correspondingly, the banding module 160 further includes:

[0095] An average contour banding unit, configured to add the average contour edge to the post-lithography probability contour band.

[0096] The optical proximity correction device provided by the present invention includes a receiving module 110 for receiving a mask pattern to be processed; an average image module 120 for inputting the mask pattern to be processed into a pre-trained optical proximity correction model that can predict random effects and output a corresponding average photoresist image; a brightness enhancement module 130 for enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image; a brightness reduction module 140 for reducing the brightness of the average photoresist image according to the preset random effect model parameters to obtain a second modified photoresist image; an inner and outer circle edge module 150 for obtaining an inner circle contour edge in the first modified photoresist image according to a preset model threshold and obtaining an outer circle contour edge in the second modified photoresist image according to the model threshold; and a banding module 160 for taking the area between the inner circle contour edge and the outer circle contour edge as a post-lithography probability contour band. On the basis of simulating and outputting the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner circle contour edge and outer circle contour edge in the photoresist image after brightness adjustment through a preset model threshold, which is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs according to the brightness change speed of each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a probability by means of the finally output post-lithography probability contour band, realizing hot spot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupation.

[0097] The optical proximity correction device in this embodiment is used to implement the foregoing optical proximity correction method. Therefore, the specific implementation manners in the optical proximity correction device can be seen in the embodiment part of the optical proximity correction method in the foregoing text. For example, the receiving module 110, the average image module 120, the brightness enhancement module 130, the brightness reduction module 140, the inner and outer circle edge module 150, and the banding module 160 are respectively used to implement steps S101, S102, S103, S104, S105, and S106 in the foregoing optical proximity correction method. Therefore, the specific implementation manners can refer to the descriptions of the corresponding individual part embodiments and will not be elaborated herein.

[0098] The present invention also provides a training method for an optical proximity correction model that can predict random effects. A schematic flowchart of a specific implementation manner is as Figure 4 shown, which is called the third specific implementation manner and includes:

[0099] S201: Obtain the target average contour, target outer circle contour, and target inner circle contour of the target pattern.

[0100] As a specific implementation manner, this step includes:

[0101] B1: Scan the scanned images formed by a plurality of mask patterns corresponding to the target pattern under the exposure field.

[0102] Of course, before performing the scanning in this step, it is necessary to first design the target pattern and manufacture the test mask (i.e., the mask pattern). The mask patterns of the same target pattern need to be repeatedly placed at multiple positions on the mask layout to facilitate subsequent data collection; then use the mask pattern to run the lithography process to obtain the wafer after lithography. This step is the scanned image obtained by scanning the wafer after lithography.

[0103] B2: Extract the contour of the scanned image to obtain the corresponding wafer contour.

[0104] Collect the scanning electron microscope images on the wafer after lithography (i.e., the scanned images) and extract the contours. For the same target pattern, collect multiple scanning electron microscope images at different positions on the mask and different exposure fields on the wafer and extract the contours. In this way, for the same target pattern, multiple wafer contours after lithography can be obtained. Align each of the wafer contours with the pattern on the mask layout. The contour extraction and alignment of the scanning electron microscope are both prior arts and will not be elaborated herein.

[0105] B3: According to all the wafer contours, calculate the average position information of each preset point on the wafer contour to obtain the target average contour.

[0106] For each target pattern, obtain the target average contour based on all the corresponding wafer contours. As Figure 5 shown, the dashed line AB in the figure is the normal line passing through a certain preset point. Specifically, preset points can be set on the target pattern at a certain step length, and then along the corresponding normal line direction, find the positions of all the wafer contours at the preset point, and then calculate the average position of all the wafer contours along the preset point along the normal line direction. Connect and fit the average positions corresponding to all the preset points to obtain the target average contour.

[0107] B4: Calculate the positions of each of the preset points, and the distances from each wafer contour to the average wafer contour in the normal line direction as the contour distance set of the preset point.

[0108] Measure the distances L1, L2,... Ln from each wafer contour to the average wafer contour along the normal line direction of each preset point. The distribution of these distances is the characterization of the lithography random effect at the preset point.

[0109] B5: Perform one-dimensional Gaussian function fitting on the distances from each wafer contour to the average wafer contour in the contour distance set to obtain the contour probability Gaussian function.

[0110] For the set of contour distances of the same preset point, the distributions of the individual distances can be approximately described by a normal distribution. Fit these distance distributions with the one-dimensional Gaussian function shown in formula (1), as Figure 6 shown.

[0111] ; (1)

[0112] where L is the distance value in the set of contour distances, and σ is the standard deviation in the one-dimensional Gaussian function.

[0113] B6: Determine the standard deviation of the contour probability Gaussian function, and move the corresponding preset point on the average wafer contour inward by a distance corresponding to the standard deviation along the normal direction to obtain the inner ring contour point corresponding to the preset point. Move the corresponding preset point on the average wafer contour outward by a distance corresponding to the standard deviation along the normal direction to obtain the outer ring contour point corresponding to the preset point.

[0114] Through Figure 6 it can be seen that the probability (68.3%) that the wafer contour falls within the standard deviation σ is greater than the probability that it falls outside the standard deviation σ. For each preset point on the average wafer contour, obtain the points that are σ away from it outward (i.e., the outer ring contour point) and inward (i.e., the inner ring contour point) respectively.

[0115] B7: Fit the inner ring contour points corresponding to all the preset points to obtain the target inner ring contour.

[0116] B8: Fit the outer ring contour points corresponding to all the preset points to obtain the target outer ring contour.

[0117] Connect the outer ring contour points and the inner ring contour points in sequence by fitting to obtain the target outer ring contour and the target inner ring contour. The area between the target outer ring contour and the target inner ring contour is the roughness band of the target pattern, and further reference can be made to Figure 5 , Figure 5 where the solid lines draw the wafer contours, and the three circles represented by the black dashed lines are, from outside to inside, the target outer ring contour, the target average contour, and the target inner ring contour. According to the normal distribution theory, there is a 68.3% probability that the wafer contour after lithography falls within this roughness band. The roughness band also provides a template for the calibration of the random effect model parameters later.

[0118] S202: Calibrate the linear coefficient and the model threshold corresponding to the photoresist model term of the to-be-trained model according to the target average contour to obtain a calibrated model.

[0119] In this step, an optical image model and a photoresist model need to be established first. The above two models are collectively referred to as the model to be trained, and the model to be trained is shown in Formula (2):

[0120] ; (2)

[0121] where I i (x, y) is the optical image of pattern i, F j is the j-th photoresist model term calculated based on the optical image, c j is the linear coefficient corresponding to F j , and the linear superposition of all photoresist model terms gives the photoresist image R i (x, y) of the target pattern. T is the model threshold of the photoresist model. The simulated post-lithography profile can be obtained through the photoresist image and the model threshold.

[0122] To calibrate the model to be trained, it is necessary to first initialize the parameters of each term in the photoresist model and calculate the image F j of each term in the photoresist model, and then optimize the linear coefficient c j of each term in the photoresist model and the model threshold T with the first objective function shown in Formula (3).

[0123] ; (3)

[0124] In Formula (3), EPE ak represents the distance from a point k on the target average profile along the normal direction to the simulated profile obtained from Formula (3), and w k is the corresponding weight.

[0125] The linear coefficient c j of each term in the photoresist model and the model threshold T corresponding to the minimum cost in Equation (3) are the linear coefficient c j of each term in the calibrated photoresist model and the model threshold T.

[0126] S203: Input the target pattern into the calibration model to obtain the target photoresist image.

[0127] Input the target pattern into the calibration model, and the obtained target photoresist image is the photoresist image that is theoretically closest to the actual production situation without considering the lithography random effect.

[0128] S204: Based on the target photoresist image, calibrate the parameters of the stochastic effect model so that in the target photoresist image with enhanced brightness according to the calibrated parameters of the stochastic effect model, the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour is less than a preset allowable value, and / or in the target photoresist image with reduced brightness according to the calibrated parameters of the stochastic effect model, the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour is less than a preset allowable value.

[0129] Please refer to Equation (4). The following Equation (4) describes the stochastic distribution law of the photoresist image in Equation (2) using a normal distribution.

[0130] ; (4)

[0131] where R 0 is the target photoresist image calculated in Equation (2), d R is the parameter of the stochastic effect model. According to Equation (4), the simulated image corresponding to the target inner circle contour is R 0 +d R (brightness enhancement), and the simulated image corresponding to the target outer circle contour is R 0 -d R (brightness reduction).

[0132] As a specific implementation, the method for calibrating the parameters of the stochastic effect model includes:

[0133] Determine that when the value of the loss function in the following Equation (5) is minimized, the corresponding parameter of the stochastic effect model is the calibrated parameter of the stochastic effect model:

[0134] ; (5)

[0135] where cost is the value of the loss function; EPE ik is the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour in the target photoresist image with enhanced brightness according to the parameter of the stochastic effect model; EPE ok is the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour in the target photoresist image with reduced brightness according to the parameter of the stochastic effect model; w k is the weight coefficient.

[0136] In this preferred embodiment, the formula (5) is used as the corresponding loss function, and the meaning expressed in the formula is to obtain the product of the sum of squares of the AND of each point and the weight coefficient. When the formula (5) takes the minimum value, it means that after changing the brightness through the parameters of a single random effect model, the difference between the simulated inner circle contour and the simulated outer circle contour and the corresponding target inner circle contour and target outer circle contour is the smallest, further improving the accuracy of the model simulation.

[0137] Furthermore, after completing the calibration of the linear coefficient, the model threshold, and the random effect model parameters, it further includes:

[0138] Using the calibrated linear coefficient, the model threshold, and the random effect model parameters, iterate the parameters of the photoresist model term in the calibration model to minimize the loss function value corresponding to the calibration model corresponding to the parameters of the iterated photoresist model term.

[0139] In the previous text, the linear coefficient and the model threshold corresponding to the photoresist model term in the formula (2) were calibrated. In this preferred embodiment, using the linear coefficient corresponding to the calibrated photoresist model term, the model threshold, and the random effect model parameters (that is, keeping the above parameters unchanged), for F in the formula (2) j The parameters inside are iterated through the optimization engine until the optimization convergence condition is met or all the iteration times are completed, thereby further improving the simulation accuracy of the model.

[0140] The training method of the optical proximity correction model capable of predicting random effects provided by the present invention includes obtaining the target average contour, target outer circle contour and target inner circle contour of the target pattern; calibrating the linear coefficient and model threshold corresponding to the photoresist model item of the model to be trained according to the target average contour to obtain a calibrated model; inputting the target pattern into the calibrated model to obtain a target photoresist image; based on the target photoresist image, calibrating the random effect model parameters, so that in the target photoresist image with increased brightness according to the calibrated random effect model parameters, the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour is less than a preset allowable value, and / or in the target photoresist image with decreased brightness according to the calibrated random effect model parameters, the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour is less than a preset allowable value. Based on the simulated output of the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner circle contour edge and outer circle contour edge in the photoresist image after brightness adjustment through a preset model threshold. It is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs according to the brightness change speed at each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a probability based on the finally output post-lithography probability contour band. Without significantly increasing the computing time and computing power occupancy, the hot spot prediction of the large-scale integrated circuit chip layout is realized.

[0141] The training device of the optical proximity correction model capable of predicting random effects provided by the embodiments of the present invention will be introduced below. The training device of the optical proximity correction model capable of predicting random effects described below can be mutually corresponding and referenced to the training method of the optical proximity correction model capable of predicting random effects described above.

[0142] Figure 7 It is a structural block diagram of the training device of the optical proximity correction model capable of predicting random effects provided by the embodiments of the present invention, which is called the fourth specific implementation manner. Refer to Figure 7 The training device of the optical proximity correction model capable of predicting random effects may include:

[0143] An acquisition module 210, configured to obtain the target average contour, target outer circle contour and target inner circle contour of the target pattern;

[0144] A model calibration module 220, configured to calibrate the linear coefficient and model threshold corresponding to the photoresist model item of the model to be trained according to the target average contour to obtain a calibrated model;

[0145] A target image module 230, configured to input the target pattern into the calibrated model to obtain a target photoresist image;

[0146] The parameter calibration module 240 is configured to calibrate the random effect model parameters based on the target photoresist image, so that in the target photoresist image with enhanced brightness according to the calibrated random effect model parameters, the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour is less than a preset allowable value, and / or in the target photoresist image with reduced brightness according to the calibrated random effect model parameters, the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour is less than a preset allowable value.

[0147] As a preferred embodiment, the parameter calibration module 240 includes:

[0148] The loss function unit is configured to determine that when the loss function value in the following formula (5) is the smallest, the corresponding random effect model parameters are the calibrated random effect model parameters:

[0149] ; (5)

[0150] where cost is the loss function value; EPE ik is the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour in the target photoresist image with enhanced brightness according to the random effect model parameters; EPE ok is the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour in the target photoresist image with reduced brightness according to the random effect model parameters; w k is the weight coefficient.

[0151] As a preferred embodiment, it further includes:

[0152] The iteration unit is configured to use the calibrated linear coefficient, the model threshold, and the random effect model parameters to iterate the parameters of the photoresist model term in the calibration model, so that the loss function value corresponding to the calibration model corresponding to the parameters of the iterated photoresist model term is the smallest.

[0153] As a preferred embodiment, the acquisition module 210 includes:

[0154] The scanning unit is configured to scan the scanning images formed by a plurality of mask patterns corresponding to the target pattern in the exposure field;

[0155] The contour extraction unit is configured to extract the contour of the scanning image to obtain the corresponding wafer contour;

[0156] An average profile calculation unit, configured to calculate the average position information of each preset point on the wafer profile according to all the wafer profiles, so as to obtain a target average profile;

[0157] A distance set unit, configured to calculate the distances from each wafer profile to the average wafer profile in the normal direction at the positions of each preset point, as the profile distance set of the preset point;

[0158] A Gaussian fitting unit, configured to perform one-dimensional Gaussian function fitting on the distances from each wafer profile to the average wafer profile in the profile distance set, so as to obtain a profile probability Gaussian function;

[0159] An inner and outer profile point unit, configured to determine the standard deviation of the profile probability Gaussian function, and move the corresponding preset point on the average wafer profile inward by a distance corresponding to the standard deviation in the normal direction to obtain the inner circle profile point corresponding to the preset point, and move the corresponding preset point on the average wafer profile outward by a distance corresponding to the standard deviation in the normal direction to obtain the outer circle profile point corresponding to the preset point;

[0160] A target inner circle unit, configured to fit the inner circle profile points corresponding to all the preset points to obtain a target inner circle profile;

[0161] A target outer circle unit, configured to fit the outer circle profile points corresponding to all the preset points to obtain a target outer circle profile.

[0162] The training device for an optical proximity correction model capable of predicting random effects provided by the present invention includes an acquisition module 210, configured to obtain the target average profile, the target outer ring profile, and the target inner ring profile of a target pattern; a model calibration module 220, configured to calibrate the linear coefficient and the model threshold corresponding to the photoresist model term of a model to be trained according to the target average profile, so as to obtain a calibrated model; a target image module 230, configured to input the target pattern into the calibrated model to obtain a target photoresist image; and a parameter calibration module 240, configured to calibrate the random effect model parameters based on the target photoresist image, so that in the target photoresist image with increased brightness according to the calibrated random effect model parameters, the edge placement error between the simulated inner ring profile obtained according to the calibrated model threshold and the target inner ring profile is less than a preset allowable value, and / or in the target photoresist image with decreased brightness according to the calibrated random effect model parameters, the edge placement error between the simulated outer ring profile obtained according to the calibrated model threshold and the target outer ring profile is less than a preset allowable value. On the basis of simulating and outputting an average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner ring profile edge and outer ring profile edge in the photoresist image after brightness adjustment through a preset model threshold, which is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs according to the brightness change speed at each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a probability by means of the finally output post-lithography probability profile band, and realize the hot spot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupancy.

[0163] The training device for an optical proximity correction model capable of predicting random effects in this embodiment is used to implement the foregoing training method for an optical proximity correction model capable of predicting random effects. Therefore, the specific implementation manners in the training device for an optical proximity correction model capable of predicting random effects can be seen in the embodiment part of the training method for an optical proximity correction model capable of predicting random effects in the foregoing text. For example, the acquisition module 210, the flat model calibration module 220, the target image module 230, and the parameter calibration module 240 are respectively used to implement steps S201, S202, S203, and S204 in the foregoing training method for an optical proximity correction model capable of predicting random effects. Therefore, the specific implementation manners can refer to the descriptions of the corresponding various part embodiments and will not be elaborated herein.

[0164] The present invention also provides an optical proximity correction device, including:

[0165] a memory, configured to store a computer program;

[0166] A processor, which is configured to implement the steps of any one of the above optical proximity correction methods when executing the computer program. The optical proximity correction method provided by the present invention includes receiving a mask pattern to be processed; inputting the mask pattern to be processed into a pre-trained optical proximity correction model that can predict random effects, and outputting a corresponding average photoresist image; enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image; reducing the brightness of the average photoresist image according to the preset random effect model parameters to obtain a second modified photoresist image; in the first modified photoresist image, obtaining an inner contour edge according to a preset model threshold, and in the second modified photoresist image, obtaining an outer contour edge according to the model threshold; taking the area between the inner contour edge and the outer contour edge as a post-lithography probability contour band. On the basis of simulating and outputting an average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner contour edge and outer contour edge in the photoresist image after brightness adjustment through a preset model threshold, which is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs according to the brightness change speed at each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a certain probability by means of the finally output post-lithography probability contour band, realizing the hot spot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupancy.

[0167] The present invention also provides a training device for an optical proximity correction model that can predict random effects, including:

[0168] A memory for storing a computer program;

[0169] A processor, which is configured to implement the steps of the training method of the optical proximity correction model with predictable random effects as described in any one of the above when executing the computer program. The optical proximity correction method provided by the present invention includes receiving a mask pattern to be processed; inputting the mask pattern to be processed into a pre-trained optical proximity correction model with predictable random effects to output a corresponding average photoresist image; enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first modified photoresist image; reducing the brightness of the average photoresist image according to the preset random effect model parameters to obtain a second modified photoresist image; obtaining an inner contour edge according to a preset model threshold in the first modified photoresist image, and obtaining an outer contour edge according to the model threshold in the second modified photoresist image; and taking the area between the inner contour edge and the outer contour edge as the post-lithography probability contour band. On the basis of simulating and outputting the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner contour edge and outer contour edge in the photoresist image after brightness adjustment through the preset model threshold, which is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs at each position in the average photoresist image according to the speed of brightness change. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a probability by means of the finally output post-lithography probability contour band, and realize the hot spot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupancy.

[0170] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, it implements the steps of any one of the above optical proximity correction methods and / or the steps of the training method of the optical proximity correction model for predictable random effects as described above. The optical proximity correction method provided by the present invention includes receiving a mask pattern to be processed; inputting the mask pattern to be processed into a pre-trained optical proximity correction model for predictable random effects to output a corresponding average photoresist image; according to preset random effect model parameters, increasing the brightness of the average photoresist image to obtain a first modified photoresist image; according to the preset random effect model parameters, decreasing the brightness of the average photoresist image to obtain a second modified photoresist image; in the first modified photoresist image, obtaining an inner contour edge according to a preset model threshold, and in the second modified photoresist image, obtaining an outer contour edge according to the model threshold; taking the area between the inner contour edge and the outer contour edge as the post-lithography probability contour band. Based on the simulated output of the average photoresist image, the present invention further adjusts the brightness of the average photoresist image through preset random effect model parameters, and then outlines the corresponding inner contour edge and outer contour edge in the photoresist image after brightness adjustment through a preset model threshold. It is equivalent to inferring the potential hot spot distribution range when the lithography random effect occurs based on the brightness change speed of each position in the average photoresist image. Workers can also predict whether there are defects such as bridging or breakpoints on the layout with a probability by means of the finally output post-lithography probability contour band, achieving hot spot prediction of the large-scale integrated circuit chip layout without significantly increasing the operation time and computing power occupancy.

[0171] The various embodiments in this specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method part.

[0172] It should be noted that in this specification, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0173] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0174] The steps of the method or algorithm described in combination with the embodiments disclosed herein can be directly implemented by hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the technical field.

[0175] The above has introduced in detail the optical proximity correction method, device, equipment, storage medium, and the training method and device of the optical proximity correction model for predictable random effects provided by the present invention. Specific examples are used herein to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principles of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. An optical proximity correction method, characterized in that: include: receiving a mask pattern to be processed; Inputting the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting random effects, and outputting a corresponding average photoresist image; According to preset random effect model parameters, the brightness of the average photoresist image is increased to obtain a first changed photoresist image; According to preset random effect model parameters, the brightness of the average photoresist image is reduced to obtain a second changed photoresist image; In the first changed photoresist image, an inner circle contour edge is obtained according to a preset model threshold, and in the second changed photoresist image, an outer circle contour edge is obtained according to the model threshold; The area between the inner circle contour edge and the outer circle contour edge is used as the post-lithography probability contour band.

2. The optical proximity correction method according to claim 1, wherein: After outputting the average photoresist image, the method further comprises: In the average photoresist image, an average contour edge is obtained according to a preset model threshold; Accordingly, after obtaining the post-lithography probability profile band, the method further includes: In the post-lithography probability contour band, the average contour edge is added.

3. An optical proximity correction device, characterized in that: include: A receiving module, used for receiving a mask pattern to be processed; An average image module, used for inputting the mask pattern to be processed into a pre-trained optical proximity correction model capable of predicting random effects, and outputting a corresponding average photoresist image; A brightness enhancement module, used for enhancing the brightness of the average photoresist image according to preset random effect model parameters to obtain a first changed photoresist image; A brightness reduction module, used for reducing the brightness of the average photoresist image according to preset random effect model parameters to obtain a second changed photoresist image; An inner and outer circle edge module, used to obtain an inner circle contour edge in the first changed photoresist image according to a preset model threshold, and to obtain an outer circle contour edge in the second changed photoresist image according to the model threshold; The scribing module is used to use the area between the inner circle contour edge and the outer circle contour edge as a post-lithography probability contour zone.

4. A training method for an optical proximity correction model capable of predicting random effects, characterized in that: include: Obtaining a target average contour, a target outer circle contour, and a target inner circle contour of a target pattern; Calibrate the linear coefficient and model threshold corresponding to the photoresist model item of the model to be trained according to the target average profile to obtain a calibration model; Inputting the target pattern into the calibration model to obtain a target photoresist image; Based on the target photoresist image, the random effect model parameters are calibrated so that in the target photoresist image after the brightness is increased according to the calibrated random effect model parameters, the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour is less than the preset allowable value, and / or in the target photoresist image after the brightness is reduced according to the calibrated random effect model parameters, the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour is less than the preset allowable value.

5. The training method of the optical proximity correction model capable of predicting random effects as claimed in claim 4, characterized in that: The method of calibrating the parameters of the random effect model comprises: When the loss function value in the following formula is determined to be the minimum, the corresponding random effect model parameters are the random effect model parameters that have completed calibration: ； Among them, cost is the loss function value; EPE ik is the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour in the target photoresist image after the brightness is enhanced according to the random effect model parameters; EPE ok is the edge placement error between the simulated outer ring contour obtained according to the calibrated model threshold and the target outer ring contour in the target photoresist image after the brightness is reduced according to the random effect model parameters; w k is the weight coefficient.

6. The training method of the optical proximity correction model capable of predicting random effects as claimed in claim 5, characterized in that: After completing the calibration of the linear coefficient, the model threshold and the random effect model parameter, the method further includes: The parameters of the photoresist model item in the calibration model are iterated using the calibrated linear coefficient, the model threshold and the random effect model parameter, so that the loss function value corresponding to the calibration model corresponding to the parameters of the iterated photoresist model item is minimized.

7. The training method of an optical proximity correction model capable of predicting random effects as claimed in claim 4, characterized in that: Obtaining the target average contour, target outer circle contour and target inner circle contour of the target pattern includes: Scanning a scanned image formed by a plurality of mask patterns corresponding to the target pattern under an exposure field; Extracting the contour of the scanned image to obtain a corresponding wafer contour; According to all wafer contours, average position information of each preset point on the wafer contour is calculated to obtain a target average contour; Calculating the position of each of the preset points and the distance from each wafer contour to the average wafer contour in the normal direction as a contour distance set of the preset points; Performing one-dimensional Gaussian function fitting on the distances from each wafer contour in the contour distance collection to the average wafer contour to obtain a contour probability Gaussian function; Determine the standard deviation of the contour probability Gaussian function, and move the corresponding preset point on the average wafer contour inwardly along the normal direction by a distance corresponding to the standard deviation to obtain an inner circle contour point corresponding to the preset point, and move the corresponding preset point on the average wafer contour outwardly along the normal direction by a distance corresponding to the standard deviation to obtain an outer circle contour point corresponding to the preset point; Fit the inner circle contour points corresponding to all preset points to obtain the target inner circle contour; Fit the outer circle contour points corresponding to all preset points to obtain the target outer circle contour.

8. A training device for an optical proximity correction model capable of predicting random effects, characterized in that: include: An acquisition module, used for obtaining a target average contour, a target outer circle contour and a target inner circle contour of a target pattern; A model calibration module, used to calibrate the linear coefficients and model thresholds corresponding to the photoresist model items of the model to be trained according to the target average profile, to obtain a calibration model; A target image module, used for inputting the target pattern into the calibration model to obtain a target photoresist image; A parameter calibration module is used to calibrate the random effect model parameters based on the target photoresist image, so that in the target photoresist image after the brightness is increased according to the calibrated random effect model parameters, the edge placement error between the simulated inner circle contour obtained according to the calibrated model threshold and the target inner circle contour is less than a preset allowable value, and / or in the target photoresist image after the brightness is reduced according to the calibrated random effect model parameters, the edge placement error between the simulated outer circle contour obtained according to the calibrated model threshold and the target outer circle contour is less than a preset allowable value.

9. An optical proximity correction device, characterized in that include: Memory for storing computer programs; A processor, configured to implement the steps of the optical proximity correction method as claimed in any one of claims 1 to 2 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the optical proximity correction method as described in any one of claims 1 to 2, and / or the steps of the training method of the optical proximity correction model capable of predicting random effects as described in any one of claims 4 to 7.

Images