Graphics correction method

Abstract

A method for pattern correction includes: providing an underlying structure, the underlying structure including a plurality of conductive layers arranged in parallel; providing a layout to be corrected, the layout to be corrected having a plurality of patterns to be corrected, the layout to be corrected being used to form a patterned layer on the underlying structure; obtaining first information through the pattern to be corrected; obtaining second information through the conductive layer; obtaining etching deviation according to the first information and the second information; and correcting the pattern to be corrected according to the etching deviation to obtain a target pattern. The target pattern obtained by the method has a high pattern accuracy.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor manufacturing technology, and in particular to a pattern correction method. Background Art

[0002] In order to transfer the pattern from the mask to the surface of the silicon wafer, it is usually necessary to go through an exposure step, a development step after the exposure step, and an etching step after the development step. In the exposure step, light is irradiated onto the silicon wafer coated with photoresist through the light-transmitting area of ​​the mask, and the photoresist undergoes a chemical reaction under the irradiation of light; in the development step, the photolithography pattern is formed by utilizing the different degrees of solubility of the photosensitive and non-photosensitive photoresists in the developer, and the mask pattern is transferred to the photoresist; in the etching step, the silicon wafer is etched based on the photolithography pattern formed by the photoresist layer, and the mask pattern is further transferred to the silicon wafer.

[0003] In semiconductor manufacturing, as the design size continues to shrink, the design size is getting closer and closer to the limit of the lithography imaging system, and the diffraction effect of light becomes more and more obvious, resulting in optical image degradation of the design pattern. The actual lithography pattern is seriously distorted relative to the pattern on the mask. Finally, the actual pattern formed by lithography on the silicon wafer is different from the designed pattern. This phenomenon is called the optical proximity effect (OPE: Optical Proximity Effect).

[0004] In order to correct the optical proximity effect, optical proximity correction (OPC) was created. The core idea of ​​optical proximity correction is to establish an optical proximity correction model based on the consideration of offsetting the optical proximity effect, and design the photomask pattern according to the optical proximity correction model. In this way, although the optical proximity effect occurs in the photolithography pattern corresponding to the photomask pattern after photolithography, since the offset of this phenomenon has been considered when designing the photomask pattern according to the optical proximity correction model, the photolithography pattern after photolithography is close to the target pattern that the user actually wants to obtain. Etching deviation is one of the parameters of optical proximity correction.

[0005] However, the etching deviation obtained by the existing method for obtaining the etching deviation has a large size error between the photolithography pattern obtained by exposing after correcting the mask pattern and the target pattern. Summary of the invention

[0006] The technical problem solved by the present invention is to provide a pattern correction method to reduce the size error between the photolithography pattern and the target pattern.

[0007] In order to solve the above technical problems, the technical solution of the present invention provides a graphic correction method, including: providing an underlying structure, the underlying structure including a plurality of conductive layers arranged in parallel; providing a layout to be corrected, the layout to be corrected having a plurality of graphics to be corrected, the layout to be corrected being used to form a graphic layer on the underlying structure; obtaining first information through the graphics to be corrected; obtaining second information through the conductive layer; obtaining an etching deviation based on the first information and the second information; and correcting the graphics to be corrected based on the etching deviation to obtain a target graphic.

[0008] Optionally, the method for obtaining an etching deviation based on the first information and the second information includes: obtaining a convolutional neural network model; and using the convolutional neural network model to process the first information and the second information to obtain an etching deviation.

[0009] Optionally, the method of processing the first information and the second information to obtain an etching deviation includes: obtaining an input vector v according to the first information and the second information; and using a convolutional neural network model to process the input vector v to obtain the etching deviation.

[0010] Optionally, the method for obtaining an input vector v according to the first information and the second information includes: obtaining a first vector v0 according to the first information; obtaining a second vector v according to the second information g ; The first vector v0 and the second vector v g Concatenate and get the input vector v=v0+v g .

[0011] Optionally, the figure to be corrected includes: a first short side LE1 and a second short side LE2 that are opposite to each other; and a second side that is perpendicular to the first short side LE1 and the second short side LE2, and two ends of the second side are respectively connected to the first short side LE1 and the second short side LE2.

[0012] Optionally, the method for obtaining the first information of the figure to be corrected includes: obtaining a first size CD, the first size CD is the length of the first short side LE1 and the second short side LE2; obtaining a second size L, the second size L is the length of the second side, and the second size L is greater than the first size CD; obtaining a first spacing space, the first spacing space is the spacing between adjacent parallel figures to be corrected.

[0013] Optionally, the method for obtaining the first vector v0 includes: obtaining the first vector v0 according to a first size CD, a second size L and a first spacing space; the first vector v0 = [CD, space, L].

[0014] Optionally, the method for obtaining the first information of the figure to be corrected also includes: dividing the second side into several line segments; obtaining a second spacing dist1, the second spacing dist1 is the distance between the midpoint of any of the line segments and the first short side LE1; obtaining a third spacing dist2, the third spacing dist2 is the distance between the midpoint of any of the line segments and the second short side LE2; the line segment type type includes: first-class line segments, second-class line segments or third-class line segments, the first-class line segments are line segments among several line segments that are connected to the first short side LE1 or the second short side LE2, the second-class line segments are line segments among several line segments that are not connected to the first short side LE1 and the second short side LE2, and the third-class line segments include the first short side LE1 and the second short side LE2.

[0015] Optionally, the method for obtaining the first vector v0 also includes: obtaining the first vector v0 according to the first size CD, the second size L, the first spacing space, the line segment type type, the second spacing dist1 and the third spacing dist2, wherein the first vector v0 = [CD, space, L, type, dist1, dist2].

[0016] Optionally, the method for obtaining the second information of the conductive layer includes: obtaining a density function P(x, y) of the conductive layer in the underlying structure; providing a first Gaussian function G S1 , the second Gaussian function G S2 , the third Gaussian function G S3 and the fourth Gaussian function G S4 , the first Gaussian function G S1 With a first variance S1, the second Gaussian function G S2 With a second variance S2, the third Gaussian function G S3 With the third difference S3, the fourth Gaussian function G S4 having a fourth variance S4, wherein the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4 are different from each other; performing a first Gaussian convolution process on the density function P(x, y) to obtain a first Gaussian convolution function Q1(x, y), Perform a second Gaussian convolution process on the density function P(x, y) to obtain a second Gaussian convolution function Q2(x, y), Performing a third Gaussian convolution process on the density function P(x, y) to obtain a third Gaussian convolution function Q3(x, y), Performing a fourth Gaussian convolution process on the density function P(x, y) to obtain a fourth Gaussian convolution function Q4(x, y), Obtain the midpoint coordinates (x', y') of any line segment; obtain second information according to the midpoint coordinates (x', y'), wherein the second information includes the function value of the first Gaussian convolution function Q1(x, y) at the midpoint (x', y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment

[0017] Optionally, the method for obtaining the density function P(x, y) of the conductive layer includes: establishing a two-dimensional coordinate system (X, Y) in the plane of the surface of the underlying structure; obtaining a coordinate range set of the contour of the conductive layer in the two-dimensional coordinate system; and obtaining the density function P(x, y) of the conductive layer based on the coordinate range set of the conductive layer.

[0018] Optionally, P(x,y)=0 or P(x,y)=1.

[0019] Optionally, when any coordinate (x, y) in the two-dimensional coordinate system is within the coordinate range of the conductive layer, P(x, y) = 1; when any coordinate (x, y) in the two-dimensional coordinate system is outside the coordinate range of the conductive layer, P(x, y) = 0.

[0020] Optionally, get the second vector v g The method includes: according to the function value of the first Gaussian convolution function Q1(x,y) at the midpoint (x',y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment Get the second vector v g , the second vector

[0021] Optionally, the method for obtaining the convolutional neural network model includes: providing an initial convolutional neural network model; providing several groups of training samples, the training samples including information of several graphics to be corrected and corresponding target etching deviations; obtaining an input vector based on the information of several graphics to be corrected; inputting the input vector into the initial convolutional neural network model for iterative training to obtain the convolutional neural network model.

[0022] Optionally, the method of inputting the input vector into the initial convolutional neural network model for iterative training includes: inputting the input vector into the initial convolutional neural network model, and outputting a predicted etching deviation; obtaining the difference between the predicted etching deviation and the target etching deviation; determining whether the difference is within a preset range; if the difference exceeds the preset range, continuing to use the training samples to iteratively train the initial convolutional neural network model until the range of the difference is within the preset range; if the difference is within the preset range, obtaining the convolutional neural network model.

[0023] Optionally, the number of training samples is greater than 10,000 groups.

[0024] Optionally, the convolutional neural network model includes: an input layer, used to vectorize the first information and the second information input to obtain an input vector; a hidden layer, used to classify the input vector; and an output layer, used to output the classification processing result.

[0025] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0026] By obtaining the first information of the graphic to be corrected and the second information of the underlying structure, and then obtaining the etching deviation based on the first information and the second information, the accuracy of the obtained etching deviation is improved, and the graphic to be corrected is corrected using the etching deviation, so that the correction effect of the target graphic is better.

[0027] Furthermore, a convolutional neural network model is obtained, and based on the first information and the second information, the etching deviation is obtained from the convolutional neural network model, so that the accuracy of the etching deviation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 and Figure 2 is a schematic cross-sectional view and a top view of a semiconductor structure in an embodiment of the present invention;

[0029] Figure 3 and Figure 4 is a schematic diagram of a layout to be modified in an embodiment of the present invention;

[0030] Figure 5 A flowchart of the steps of obtaining the second information in an embodiment of the present invention;

[0031] Figure 6 Schematic diagram of the steps for obtaining a convolutional neural network model in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] As described in the background art, the etching deviation obtained by the existing method for obtaining the etching deviation results in a large size error between the photolithography pattern obtained by exposing after correcting the mask pattern and the target pattern.

[0033] Specifically, in the back-end process of semiconductor structures, when using etching deviation to correct the optical proximity effect of the etching pattern of the metal layer, only the influence of the line width of the metal layer and the spacing between adjacent metal layers on the etching deviation is considered. The OPC engineer will establish an etching deviation table, and the etching deviation of a specific segment is a function of the relevant line width and spacing. This method only considers very limited information about the current metal layer and completely ignores the shape of the underlying layer. In fact, there are several metal connection layers on the underlying layer, and the metal layer is partially or completely located on the metal connection layer. In the actual etching process, the etching process is affected by the underlying material, and the etching rates of the partial metal layers located on the metal connection layer and the partial metal layers not located on the metal connection layer are different, thereby affecting the size of the formed metal layer. Therefore, the etching deviation calculated based on the above deviation table may not be accurately corrected.

[0034] In order to make the above-mentioned objects, features and beneficial effects of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0035] Figure 1 and Figure 2 1 is a schematic diagram of the cross-sectional structure and a top view of the semiconductor structure in an embodiment of the present invention.

[0036] Please refer to Figure 1 and Figure 2 , Figure 1 for Figure 2 Schematic diagram of the cross-sectional structure along the AA' direction, Figure 2 for Figure 1 A top view of the substrate 100 is provided, wherein the substrate 100 includes a plurality of conductive layers 103 arranged in parallel.

[0037] In a first direction X and a second direction Y parallel to the surface of the underlying structure, a two-dimensional coordinate system (X, Y) is established in the plane of the surface of the underlying structure, and the outline of the conductive layer 103 has a coordinate range set U in the two-dimensional coordinate system (X, Y), and the first direction X and the second direction Y are perpendicular to each other.

[0038] In this embodiment, the underlying structure also includes: a substrate 100; a device layer 101 located on the substrate 100, the device layer 101 including an isolation structure (not shown) and a device structure located within the isolation structure (not shown), the device structure including a transistor, a diode, a triode, a capacitor, an inductor or a conductive structure, etc.; a dielectric layer 102 located on the device layer 101; a conductive layer 103 located within the dielectric layer 102, the conductive layer 103 being electrically connected to the device structure.

[0039] In this embodiment, the material of the substrate 100 is silicon.

[0040] In other embodiments, the material of the substrate includes silicon carbide, silicon germanium, a multinary semiconductor material composed of group III-V elements, silicon on insulator (SOI) or germanium on insulator (GOI). Among them, the multinary semiconductor material composed of group III-V elements includes InP, GaAs, GaP, InAs, InSb, InGaAs or InGaAsP.

[0041] The material of the isolation structure includes a dielectric material, and the dielectric material includes a combination of one or more of silicon oxide, silicon nitride, silicon oxynitride, aluminum oxide, aluminum nitride, silicon carbide nitride, and silicon carbon nitride. The material of the dielectric layer 102 includes a dielectric material, and the dielectric material includes a combination of one or more of silicon oxide, silicon nitride, silicon oxynitride, aluminum oxide, aluminum nitride, silicon carbide nitride, and silicon carbon nitride. The material of the conductive layer 103 includes a metal, and the metal includes a combination of one or more of copper, aluminum, tungsten, cobalt, and titanium nitride.

[0042] In this embodiment, the material of the isolation structure includes silicon oxide; the material of the dielectric layer 102 includes silicon oxide.

[0043] Figure 3 and Figure 4 Schematic diagram of a layout to be modified in an embodiment of the present invention.

[0044] Please refer to Figure 3 , providing a layout 200 to be corrected, wherein the layout 200 to be corrected has a plurality of graphics 201 to be corrected, and the layout 200 to be corrected is used to form a patterned layer on the underlying structure.

[0045] A plurality of the to-be-corrected patterns 201 are arranged in parallel along a first direction X parallel to the to-be-corrected layout, and a plurality of the to-be-corrected patterns 201 are arranged in parallel along a second direction Y parallel to the to-be-corrected layout.

[0046] There is a space between adjacent parallel patterns 201 to be corrected.

[0047] In this embodiment, the layout to be modified 200 and the underlying structure share the same coordinate system.

[0048] Please refer to Figure 4 , Figure 4 for Figure 3 An enlarged schematic diagram of any to-be-corrected graphic 201.

[0049] The figure to be corrected 201 includes: a first short side LE1 and a second short side LE2 opposite to each other; a second side (not marked) perpendicular to the first short side LE1 and the second short side LE2, and two ends of the second side are respectively connected to the first short side LE1 and the second short side LE2.

[0050] The side lengths of the first short side LE1 and the second short side LE2 are a first dimension CD, and the side length of the second side is a second dimension L, and the second dimension L is greater than the first dimension CD.

[0051] Next, first information is obtained according to the graphic to be corrected 201 .

[0052] Please continue to refer to Figure 3 and Figure 4 , obtaining first information through the graphic to be corrected 201.

[0053] The method for obtaining the first information of the to-be-corrected figure 201 includes: obtaining a first size CD, the first size CD being the length of the first short side LE1 and the second short side LE2; obtaining a second size L, the second size L being the length of the second side, and the second size L being greater than the first size CD; and obtaining a first spacing space, the first spacing space being the spacing between adjacent parallel to-be-corrected figures.

[0054] The first information includes a first size CD, a second size L, and a first spacing space.

[0055] Next, a first vector v0 is obtained according to the first information.

[0056] The method for obtaining a first vector v0 includes: obtaining the first vector v0 according to a first size CD, a second size L and a first spacing space; the first vector v0 = [CD, space, L].

[0057] Please continue to refer to Figure 3 and Figure 4 In this embodiment, the method further includes: dividing the second side into a plurality of line segments.

[0058] The line segment types type include: first-type line segments N1, second-type line segments N2 or third-type line segments, the first-type line segments are line segments among several line segments that are connected to the first short side LE1 or the second short side LE2, the second-type line segments are line segments among several line segments that are not connected to the first short side LE1 and the second short side LE2, and the third-type line segments include the first short side LE1 and the second short side LE2.

[0059] The first type of line segment N1 has a first midpoint C1, and the coordinates of the first midpoint C1 are (x'1, y'1); the second type of line segment N2 has a second midpoint C2, and the coordinates of the second midpoint C2 are (x'2, y'2); the first short side LE1 or the second short side LE2 has a third midpoint C3, and the coordinates of the third midpoint C3 are (x'3, y'3).

[0060] The method for obtaining the first information of the figure to be corrected 201 also includes: obtaining the second distance dist1, the second distance dist1 is the distance between the midpoint C1, C2 or C3 of any of the line segments and the first short side LE1; obtaining the third distance dist2, the third distance dist2 is the distance between the midpoint C1, C2 or C3 of any of the line segments and the second short side LE2; obtaining the line segment type type, the line segment type type includes the first type of line segment N1, the second type of line segment N2 or the third type of line segment, and the third type of line segment includes the first short side LE1 and the second short side LE2.

[0061] In this embodiment, the first information includes a first size CD, a second size L, a first spacing space, a second spacing dist1, a third spacing dist2, and a line segment type type.

[0062] In addition to the first size CD, the second size L, and the first spacing space, the first information also includes the second spacing dist1, the third spacing dist2, and the line segment type type, so that the content of the first information is richer, and when the etching deviation is obtained through the first information and the second information, the etching deviation is more accurate.

[0063] In other embodiments, the first information may not include the second distance dist1, the third distance dist2, and the line segment type type.

[0064] Next, second information is obtained according to the conductive layer 103 .

[0065] Figure 5 This is a flow chart of the steps of obtaining the second information in an embodiment of the present invention.

[0066] Please refer to Figure 5 , obtaining the second information through the conductive layer 103, the steps of obtaining the second information of the conductive layer 103 are as follows:

[0067] S10: Obtaining a density function P(x, y) of the conductive layer in the underlying structure;

[0068] S20: Provide the first Gaussian function G S1 , the second Gaussian function G S2 , the third Gaussian function GS3 and the fourth Gaussian function G S4 , the first Gaussian function G S1 With a first variance S1, the second Gaussian function G S2 With a second variance S2, the third Gaussian function G S3 With the third difference S3, the fourth Gaussian function G S4 having a fourth variance S4, wherein the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4 are different from each other;

[0069] S30: Perform a first Gaussian convolution process on the density function P(x, y) to obtain a first Gaussian convolution function Q1(x, y).

[0070] S40: performing a second Gaussian convolution process on the density function P(x, y) to obtain a second Gaussian convolution function Q2(x, y),

[0071] S50: performing a third Gaussian convolution process on the density function P(x, y) to obtain a third Gaussian convolution function Q3(x, y).

[0072] S60: performing a fourth Gaussian convolution process on the density function P(x, y) to obtain a fourth Gaussian convolution function Q4(x, y),

[0073] S70: Obtain the midpoint coordinates (x', y') of any line segment;

[0074] S80: Acquire second information according to the midpoint coordinates (x', y'), the second information including the function value of the first Gaussian convolution function Q1(x, y) at the midpoint (x', y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment

[0075] Please refer to step S10: obtaining the density function P(x, y) of the conductive layer in the underlying structure.

[0076] The method for obtaining the density function P(x, y) of the conductive layer includes: obtaining the density function P(x, y) of the conductive layer 103 according to a coordinate range set U of the contour of the conductive layer 103 .

[0077] In this embodiment, P(x, y)=0 or P(x, y)=1.

[0078] When any coordinate (x, y) in the two-dimensional coordinate system (X, Y) is within the coordinate range set U of the outline of the conductive layer 103, P(x, y) = 1; when any coordinate (x, y) in the two-dimensional coordinate system (X, Y) is outside the coordinate range set U of the outline of the conductive layer 103, P(x, y) = 0.

[0079] Please refer to step S20: providing a first Gaussian function G S1 , the second Gaussian function G S2 , the third Gaussian function G S3 and the fourth Gaussian function G S4 , the first Gaussian function G S1 With a first variance S1, the second Gaussian function G S2 With a second variance S2, the third Gaussian function G S3 With the third difference S3, the fourth Gaussian function G S4 There is a fourth variance S4, and the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4 are different from each other.

[0080] The first Gaussian function G S1 , the second Gaussian function G S2 , the third Gaussian function G S3 and the fourth Gaussian function G S4 Having different variances, that is, the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4 can be used as different diffusion radii to perform Gaussian convolution processing on the density function P(x, y) to blur the density function P(x, y) to simulate the influence of the underlying structure on the etching of the upper layer graphics under different physical states of ion bombardment in the actual etching process.

[0081] The larger the values ​​of the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4, the larger the diffusion radius of the convolution processing of the density function P(x, y), and the higher the degree of fuzzification processing of the density function P(x, y), thereby simulating the influence of the underlying structure on the etching of the upper layer graphics under different physical states of ion bombardment.

[0082] Please refer to steps S30, S40, S50 and S60: perform a first Gaussian convolution process on the density function P(x, y) to obtain a first Gaussian convolution function Q1(x, y), Perform a second Gaussian convolution process on the density function P(x, y) to obtain a second Gaussian convolution function Q2(x, y), Performing a third Gaussian convolution process on the density function P(x, y) to obtain a third Gaussian convolution function Q3(x, y), Performing a fourth Gaussian convolution process on the density function P(x, y) to obtain a fourth Gaussian convolution function Q4(x, y),

[0083] Perform a first Gaussian convolution on the density function P(x, y) Get the first Gaussian convolution function Q1(x,y), That is, based on the underlying structure, the density function P(x, y) is fuzzy processed with the first variance S1 as the diffusion radius to simulate the influence of the underlying structure on the etching of the upper pattern in the actual etching process.

[0084] Perform a second Gaussian convolution on the density function P(x, y) Get the second Gaussian convolution function Q2(x,y), That is, based on the underlying structure, the density function P(x, y) is fuzzy processed with the second variance S2 as the diffusion radius to simulate the influence of the underlying structure on the etching of the upper pattern in the actual etching process.

[0085] Perform the third Gaussian convolution on the density function P(x,y) Get the third Gaussian convolution function Q3(x,y), That is, based on the underlying structure, the density function P(x, y) is fuzzy processed with the third difference S3 as the diffusion radius to simulate the influence of the underlying structure on the etching of the upper pattern in the actual etching process.

[0086] Perform the fourth Gaussian convolution process on the density function P(x, y) Get the fourth Gaussian convolution function Q4(x,y), That is, based on the underlying structure, the density function P(x, y) is fuzzy processed with the fourth variance S4 as the diffusion radius to simulate the influence of the underlying structure on the etching of the upper pattern in the actual etching process.

[0087] Please refer to step S70: obtaining the midpoint coordinates (x', y') of any line segment.

[0088] Any line segment includes a first-type line segment N1, a second-type line segment N2 or a third-type line segment. The first-type line segment is a line segment among several line segments that is connected to the first short side LE1 or the second short side LE2. The second-type line segment is a line segment among several line segments that is not connected to the first short side LE1 and the second short side LE2. The third-type line segment includes the first short side LE1 and the second short side LE2.

[0089] The midpoint coordinates of any line segment include: the coordinates (x'1, y'1) of the first midpoint C1 of the first type line segment N1; the coordinates (x'2, y'2) of the second midpoint C2 of the second type line segment N2; and the coordinates (x'3, y'3) of the third midpoint C3 of the first short side LE1 or the second short side LE2.

[0090] Please refer to step S80: obtaining second information according to the midpoint coordinates (x', y'), the second information including the function value of the first Gaussian convolution function Q1 (x, y) at the midpoint (x', y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment

[0091] In this embodiment, the midpoint coordinates of the line segment include (x'1, y'1), (x'2, y'2) or (x'3, y'3).

[0092] After acquiring the first information and the second information, an etching deviation is acquired according to the first information and the second information.

[0093] By acquiring the first information of the pattern to be corrected and the second information of the underlying structure, and then acquiring the etching deviation according to the first information and the second information, the accuracy of the acquired etching deviation is improved.

[0094] The method for obtaining etching deviation according to the first information and the second information includes: obtaining a convolutional neural network model; using the convolutional neural network model, processing the first information and the second information to obtain the etching deviation.

[0095] The method for processing the first information and the second information to obtain an etching deviation includes: obtaining an input vector v according to the first information and the second information; and using a convolutional neural network model to process the input vector v to obtain the etching deviation.

[0096] The method for obtaining an input vector v according to first information and second information includes: obtaining a first vector v0 according to the first information; obtaining a second vector v according to the second information g ; The first vector v0 and the second vector v g Concatenate and get the input vector v=v0+v g .

[0097] In this embodiment, the first information includes a first size CD, a second size L, a first spacing space, a second spacing dist1, a third spacing dist2, and a line segment type type.

[0098] The method for obtaining the first vector v0 includes: obtaining the first vector v0 according to the first size CD, the second size L, the first spacing space, the line segment type type, the second spacing dist1 and the third spacing dist2, wherein the first vector v0 = [CD, space, L, type, dist1, dist2].

[0099] The second information includes the function value of the first Gaussian convolution function Q1(x,y) at the midpoint (x',y') of the line segment. The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment

[0100] Get the second vector v g The method includes: according to the function value of the first Gaussian convolution function Q1(x,y) at the midpoint (x',y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment Get the second vector v g , the second vector

[0101] Get the second vector v g Afterwards, the first vector v0 and the second vector v g Concatenate and get the input vector v=v0+v g .

[0102] That is, the input vector

[0103] A convolutional neural network model is used to process the first information and the second information to obtain the etching deviation, thereby improving the accuracy of the etching deviation.

[0104] Figure 6Schematic diagram of the steps for obtaining a convolutional neural network model in an embodiment of the present invention.

[0105] Please refer to Figure 6 , the steps of obtaining the convolutional neural network model include:

[0106] S100: Provides an initial convolutional neural network model;

[0107] S200: providing a plurality of groups of training samples, wherein the training samples include information of a plurality of patterns to be corrected and corresponding target etching deviations;

[0108] S300: Obtaining an input vector according to information of a number of graphics to be corrected;

[0109] S400: Inputting the input vector into an initial convolutional neural network model for iterative training to obtain the convolutional neural network model.

[0110] The method of inputting the input vector into the initial convolutional neural network model for iterative training includes: inputting the input vector into the initial convolutional neural network model, and outputting a predicted etching deviation; obtaining the difference between the predicted etching deviation and the target etching deviation; determining whether the difference is within a preset range; if the difference exceeds the preset range, continuing to use the training sample to iteratively train the initial convolutional neural network model until the range of the difference is within the preset range; if the difference is within the preset range, obtaining the convolutional neural network model.

[0111] In this embodiment, the number of the training samples is greater than 10,000 groups.

[0112] In this embodiment, the convolutional neural network model includes: an input layer, which is used to vectorize the first information and the second information input to obtain an input vector; a hidden layer, which is used to classify the input vector; and an output layer, which is used to output the classification processing result.

[0113] After the etching deviation is obtained, the pattern to be corrected 201 is corrected according to the etching deviation to obtain a target pattern.

[0114] The process of correcting the to-be-corrected pattern 201 to obtain the target pattern is a common process of correcting the optical proximity effect, which will not be described in detail here.

[0115] In summary, by obtaining the first information of the graphic to be corrected and the second information of the underlying structure, and then obtaining the etching deviation based on the first information and the second information, the accuracy of the obtained etching deviation is improved, and the etching deviation is used to correct the graphic to be corrected, so as to obtain a better correction effect of the target graphic.

[0116] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A graphics correction method, characterized in that: include: Providing an underlying structure, the underlying structure comprising a plurality of conductive layers arranged in parallel; Providing a layout to be corrected, the layout to be corrected having a plurality of patterns to be corrected, the layout to be corrected being used to form a patterned layer on the underlying structure, the patterns to be corrected comprising: a first short side and a second short side opposite to each other, a second side perpendicular to the first short side and the second short side, and two ends of the second side respectively connected to the first short side and the second short side; Acquire first information through the to-be-corrected figure, the first information including a first size, a second size, a first spacing, a second spacing, a third spacing and a line segment type, the first size being the length of the first short side and the second short side, the second size being the length of the second side, the second size being greater than the first size, the first spacing being the spacing between adjacent parallel to-be-corrected figures, the second spacing being the distance between the midpoint of any line segment after the second side is divided into a plurality of line segments and the first short side, the third spacing being the distance between the midpoint of any line segment after the second side is divided into a plurality of line segments and the second short side, the line segment type including: first-category line segments, second-category line segments or third-category line segments, the first-category line segments being line segments connected to the first short side or the second short side among the plurality of line segments, the second-category line segments being line segments not connected to the first short side and the second short side among the plurality of line segments, and the third-category line segments including the first short side and the second short side; acquiring second information through the conductive layer; Acquire an etching deviation according to the first information and the second information; The pattern to be corrected is corrected according to the etching deviation to obtain the target pattern.

2. The image correction method according to claim 1, characterized in that: The method for obtaining etching deviation according to the first information and the second information includes: obtaining a convolutional neural network model; using the convolutional neural network model, processing the first information and the second information to obtain the etching deviation.

3. The image correction method according to claim 2, characterized in that: The method for processing the first information and the second information to obtain an etching deviation includes: obtaining an input vector v according to the first information and the second information; and using a convolutional neural network model to process the input vector v to obtain the etching deviation.

4. The image correction method according to claim 3, characterized in that: The method for obtaining an input vector v according to first information and second information includes: obtaining a first vector v0 according to the first information; obtaining a second vector v according to the second information g ; The first vector v0 and the second vector v g Concatenate to get the input vector.

5. The image correction method according to claim 4, characterized in that: The method for obtaining a first vector v0 includes: obtaining the first vector v0 according to a first size, a second size, a first spacing, a line segment type, a second spacing and a third spacing, wherein the first vector v0 = [CD, space, L, type, dist1, dist2], wherein CD is the first size, L is the second size, space is the first spacing, type is the line segment type, dist1 is the second spacing, and dist2 is the third spacing.

6. The image correction method according to claim 5, characterized in that: The method for obtaining the second information of the conductive layer comprises: obtaining the density function P(x, y) of the conductive layer in the underlying structure; providing a first Gaussian function G S1 , the second Gaussian function G S2 , the third Gaussian function G S3 and the fourth Gaussian function G S4 , the first Gaussian function G S1 With a first variance S1, the second Gaussian function G S2 With a second variance S2, the third Gaussian function G S3 With the third difference S3, the fourth Gaussian function G S4 having a fourth variance S4, wherein the first variance S1, the second variance S2, the third variance S3 and the fourth variance S4 are different from each other; performing a first Gaussian convolution process on the density function P(x, y) to obtain a first Gaussian convolution function Q1(x, y), Perform a second Gaussian convolution process on the density function P(x, y) to obtain a second Gaussian convolution function Q2(x, y), Performing a third Gaussian convolution process on the density function P(x, y) to obtain a third Gaussian convolution function Q3(x, y), Performing a fourth Gaussian convolution process on the density function P(x, y) to obtain a fourth Gaussian convolution function Q4(x, y), Obtain the midpoint coordinates (x', y') of any line segment; obtain second information according to the midpoint coordinates (x', y'), wherein the second information includes the function value of the first Gaussian convolution function Q1(x, y) at the midpoint (x', y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment 7. The image correction method according to claim 6, characterized in that: The method for obtaining the density function P(x, y) of the conductive layer includes: establishing a two-dimensional coordinate system (X, Y) in the plane of the surface of the underlying structure; obtaining a coordinate range set of the contour of the conductive layer in the two-dimensional coordinate system; and obtaining the density function P(x, y) of the conductive layer according to the coordinate range set of the contour of the conductive layer.

8. The image correction method according to claim 7, characterized in that: The P(x,y)=0 or P(x,y)=1.

9. The image correction method according to claim 8, characterized in that: When any coordinate (x, y) in the two-dimensional coordinate system is within the coordinate range of the conductive layer, P(x, y) = 1; when any coordinate (x, y) in the two-dimensional coordinate system is outside the coordinate range of the conductive layer, P(x, y) = 0.

10. The image correction method according to claim 6, characterized in that: Get the second vector v g The method includes: according to the function value of the first Gaussian convolution function Q1(x,y) at the midpoint (x',y') of the line segment The function value of the second Gaussian convolution function Q2(x,y) at the midpoint (x',y') of the line segment The function value of the third Gaussian convolution function Q3(x,y) at the midpoint (x',y') of the line segment And the function value of the fourth Gaussian convolution function Q4(x,y) at the midpoint (x',y') of the line segment Get the second vector v g , the second vector 11. The image correction method according to claim 2, characterized in that: The method for obtaining the convolutional neural network model includes: Provide an initial convolutional neural network model; Providing a plurality of sets of training samples, wherein the training samples include information of a plurality of graphics to be corrected and corresponding target etching deviations; Obtaining an input vector according to information of a number of graphics to be corrected; The input vector is input into an initial convolutional neural network model for iterative training to obtain the convolutional neural network model.

12. The image correction method according to claim 11, characterized in that: The method of inputting the input vector into the initial convolutional neural network model for iterative training includes: inputting the input vector into the initial convolutional neural network model, and outputting a predicted etching deviation; obtaining the difference between the predicted etching deviation and the target etching deviation; determining whether the difference is within a preset range; if the difference exceeds the preset range, continuing to use the training sample to iteratively train the initial convolutional neural network model until the range of the difference is within the preset range; if the difference is within the preset range, obtaining the convolutional neural network model.

13. The image correction method according to claim 11, characterized in that: The number of the training samples is greater than 10,000 groups.

14. The image correction method according to claim 2, characterized in that: The convolutional neural network model includes: an input layer, used to vectorize the input first information and second information to obtain an input vector; A hidden layer, used for classifying the input vector; The output layer is used to output the classification processing result.

Images