Mask strip registration method

Abstract

The invention discloses a mask plate stripe registration method. The method comprises the following steps: establishing a sub-stripe registration reference point array; registering the Mask images divided on the sub-strips and the mask plate design file generation graph, and sliding the Mask images step by step according to pixels in sequence to cover local images with corresponding sizes on the mask plate design file generation graph; each pixel point of the Mask image and the corresponding local image is subjected to differencing after each time of sliding, a rough threshold value is preset, when the absolute value of the difference value after differencing is smaller than the rough threshold value, the difference value is subjected to zero assignment processing, then the difference value is subjected to accumulative summation to obtain a rough difference value sum, and the rough difference value sum is stored in a corresponding rough registration result matrix A; in the rough registration result matrix A, a mark value is assigned to the position where the sum of the difference values is 0, and when adjacent strips are registered, each sub-strip uses the registration result of the corresponding sub-strip in the previous strip registration reference point array for registration; according to the method, the registration range can be narrowed, the registration time can be shortened, and the registration efficiency is improved.

Description

Technical Field

[0001] The present invention relates to a method for registering mask strip, belonging to the technical field of semiconductor detection. Background Art

[0002] Currently, in mask defect detection, the entire scanning strip is cut into multiple sub-strips according to a fixed size according to a cutting strategy. The sub-strips are arranged in a matrix and indexed in the column direction. The mask array image has translational symmetry. When one or a group of graphics is offset by a distance in a certain direction, it can coincide with another graphic or a group of graphics. The array graphics have a periodic arrangement characteristic, that is, one or a group of graphics repeatedly appears in this image area. The most common case of the array graphics in the mask is, for example, the CNT layer. The CNT layer may be used as a conductive path in an integrated circuit or as a channel of a field effect transistor. The pattern in the mask is usually presented in the form of array graphics. Since the overall image generated by GDS (including but not limited to design files in formats such as GDS, OASIS, etc.) is larger than the Mask image obtained by the camera, when registering its pattern, because the graphics have the characteristic of array arrangement, registration reference points can be obtained at multiple positions. Avoiding the error of the reference position of the registration point of the array graphics while the found position is the local optimal registration point is the key to ensuring correct registration, detection accuracy and speed. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a method for registering mask strip, which can narrow the registration range, shorten the registration time and improve the registration efficiency.

[0004] To solve the above technical problem, the technical solution of the present invention is: a method for registering mask strip, and the steps of the method include:

[0005] S01: Establish an array of registration reference points for sub-strips and initialize the positions of the registration reference points for each sub-strip;

[0006] S02: Register the Mask image divided on the sub-strip and the image generated from the mask design file. Slide the Mask image step by step by pixel to cover the corresponding local image of the same size on the image generated from the mask design file. Among them, after each slide, the difference between each pixel point of the Mask image and the corresponding local image is calculated. A rough threshold is preset. When the absolute value of the difference after calculation is less than the rough threshold, the difference at this place is set to 0, and then the differences are accumulated and summed to obtain a rough difference sum, which is saved to the corresponding rough registration result matrix A;

[0007] S03: In the rough registration result matrix A, assign a marker value to the position where the sum of the differences is 0; when there is no marker value in the rough registration result matrix A, then use the position corresponding to the minimum rough difference sum in the rough registration result matrix A as the registration reference point, and when there is a marker value in the rough registration result matrix A, then use the position corresponding to the marker value as the registration reference point.

[0008] Further, step S02 further includes: preset a fine threshold. When the absolute value of the difference after taking the difference is less than the fine threshold, assign 0 to this difference, then accumulate and sum the differences to obtain the fine difference sum, and save it to the corresponding fine registration result matrix B.

[0009] The method further includes the following steps:

[0010] S04: Take the connected domain adjacent to the position with multiple registration reference points as a region marker bit, and count the number of region marker bits that meet the requirements in the corresponding region of the sub-strip: if there are multiple region marker bits in the region, it is considered that the position of the registration reference point marking this sub-strip position cannot be referenced by other strips; if there is only one region marker bit in the region, then the position corresponding to this region marker bit needs to be saved to the sub-strip registration reference point array.

[0011] S05: Compare the position of the registration reference point obtained based on the rough registration result matrix A with the corresponding position in the fine registration result matrix B, and find the minimum fine difference sum in the fine registration result matrix B within the region range of this corresponding position. After finding all the minimum fine difference sums, obtain the corresponding fine registration reference point positions.

[0012] S06: Refer to the fine registration reference point positions obtained in step S05 according to the reference requirements specified in step S04, and save them to the sub-strip registration reference point array.

[0013] Further, the method further includes the step:

[0014] S07: After the registration of each column of sub-strips is completed, fit the arrays that cannot be referenced in the sub-strip registration reference point arrays of the entire column with the registration reference point positions of other arrays that can be referenced to obtain the registration reference point here.

[0015] Further, the specific way to slide the Mask image step by step by pixel is as follows:

[0016] a: Slide the Mask image step by step by pixel along the row direction at the starting position of the row. After the sliding in the row direction is completed, return to the starting position of the row and then slide the Mask image one pixel down in the column direction.

[0017] b: Perform step a in the new row.

[0018] c: Perform step b successively until the sliding of the last line ends.

[0019] After adopting the above technical solution, by calculating, registration reference points are provided for subsequent sub - strips for registration reference use, the registration range is reduced and the registration time can be shortened. The method is used to judge whether the position of the current registration reference point is available for other strips to refer to. When registering, the reference registration points are used to divide and design the local area of the file generation graph for registration, which can speed up the registration speed; at the same time, it can avoid outputting the local optimal point as the registration result during the registration process. During rough registration, multiple local optimal positions can be counted as alternative positions, and then secondary registration can be carried out respectively with the help of the registration reference points to find the global optimal registration point among these alternative local optimal points. In this way, when dealing with the defects of the comparison between the design file generation graph and the Mask obtained by the camera, the defects can be better identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is the Mask image diagram of the mask strip registration method of the present invention;

[0021] Figure 2 It is the design file generation graph of the mask of the present invention;

[0022] Figure 3 It is the overlay of the Mask image and the design file generation graph of the mask of the present invention Figure 1 ;

[0023] Figure 4 It is the overlay of the Mask image and the design file generation graph of the mask of the present invention Figure 2 . DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] In order to make the content of the present invention be understood more clearly, the following further describes the present invention in detail according to specific embodiments and in combination with the drawings.

[0025] As Figures 1 to 4 shown, a mask strip registration method, the steps of the method include:

[0026] S01: Establish an array of sub - strip registration reference points, initialize the position of the registration reference point for each sub - strip, and facilitate recording the position of the registration point of each column of sub - strips of the scanned strip into the corresponding sub - strip registration reference point array;

[0027] S02: Register the Mask image divided on the sub-strip with the generated image of the mask design file. Slide the Mask image pixel by pixel in sequence to cover the local image of the corresponding size on the generated image of the mask design file. Among them, after each slide, calculate the difference between each pixel point of the Mask image and the corresponding local image. Preset a rough threshold. When the absolute value of the difference after calculation is less than the rough threshold, set the difference at this position to 0, then accumulate and sum the differences to obtain the rough difference sum, and save it to the corresponding rough registration result matrix A.

[0028] S03: Assign a marker value to the position where the difference sum is 0 in the rough registration result matrix A. When there is no marker value in the rough registration result matrix A, take the position corresponding to the minimum rough difference sum in the rough registration result matrix A as the registration reference point. When there is a marker value in the rough registration result matrix A, take the position corresponding to the marker value as the registration reference point.

[0029] In this embodiment, during design, not all regions of the generated image of the mask design file are fully used as the registration region. Instead, a local region is selected as the registration region of the Mask image (this local region is also larger than the Mask image), but the size of the entire rough registration result matrix A is designed according to all regions of the generated image of the mask design file, and all data is set to 0 during initialization.

[0030] Specifically, step S02 further includes: preset a fine threshold. When the absolute value of the difference after calculation is less than the fine threshold, set the difference at this position to 0, then accumulate and sum the differences to obtain the fine difference sum, and save it to the corresponding fine registration result matrix B.

[0031] The method further includes the following steps:

[0032] S04: Regard the connected domain adjacent to and having multiple registration reference points as a region marker bit, and count the number of region marker bits that meet the requirements in the corresponding region of the sub-strip. If there are multiple region marker bits in the region, it is considered that the position of the registration reference point marking this sub-strip position cannot be referenced by other strips. If there is only one region marker bit in the region, the position corresponding to this region marker bit needs to be saved to the sub-strip registration reference point array.

[0033] S05: Compare the position of the registration reference point obtained based on the rough registration result matrix A with the corresponding position in the fine registration result matrix B, and find the minimum fine difference sum in the region range of this corresponding position in the fine registration result matrix B. After finding all the minimum fine difference sums, obtain the corresponding fine registration reference point positions.

[0034] S06: Refer to the fine registration reference points obtained in step S05 according to the reference requirements specified in step S04, and save them to the sub-strip registration reference point array.

[0035] Specifically, the method further includes the steps:

[0036] S07: After the registration of each column of sub-strips is completed, for the arrays in the entire column of sub-strip registration reference point arrays that cannot be referred to, use the registration reference point positions of other arrays that can be referred to for fitting to obtain the registration reference points here.

[0037] In this embodiment, adjacent sub-strips are calculated according to their positions (the offset in the row addition direction) and provided for use during the registration of adjacent sub-strips. Use the positions of these registration reference points to generate the relative positions of the reference points in the figure for the mask design file, and shrink the available registration area inward, thereby reducing the calculation and improving the registration speed;

[0038] In summary, using this method, the registration position information of the sub-strips in adjacent columns can be used to provide registration position reference for adjacent sub-strips in subsequent columns, shrink the registration area, and shorten the registration time; at the same time, it can avoid directly finding the minimum value (i.e., the minimum difference sum) as the registration point when screening for the optimal registration reference point. Instead, a large-scale search is carried out through a rough threshold to find the conforming area range, and then the difference sum obtained according to the fine threshold is further calculated within the range of each registration reference point area, so as to compare and obtain the final registration result, which can effectively avoid the problem of falling into the local optimum of registration when multiple positions of the array image are matched.

[0039] Specifically, as Figure 3 、 4 shown, the Mask image is stepped and slid pixel by pixel in sequence as follows:

[0040] a: At the starting position of the row, step and slide the Mask image pixel by pixel along the row direction. After the sliding in the row direction is completed, return to the starting position of the row and then step down one pixel in the column direction to slide the Mask image;

[0041] b: Perform step a in the new row;

[0042] c: Perform step b in sequence until the sliding of the last row is completed.

[0043] In the above specific embodiments, the technical problems solved, technical solutions, and beneficial effects of the present invention are further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A mask strip registration method, characterized in that: The steps of the method include: S01: Establish a sub-strip registration reference point array and initialize the registration reference point position of each sub-strip; S02: Align the Mask image divided on the sub-strip with the generated image of the mask design file, and slide the Mask image in pixel steps to cover the local image of the corresponding size on the generated image of the mask design file; wherein, after each slide, the difference between each pixel point of the Mask image and the corresponding local image is calculated, and a rough threshold is preset. When the absolute value of the difference after the difference is less than the rough threshold, the difference at this point is assigned 0, and then the difference is accumulated and summed to obtain a rough difference sum, and saved in the corresponding rough registration result matrix A; S03: In the rough registration result matrix A, a mark value is assigned to the position where the sum of the difference values ​​is 0; when there is no mark value in the rough registration result matrix A, the minimum rough difference value and the corresponding position in the rough registration result matrix A are used as the registration reference point; when there is a mark value in the rough registration result matrix A, the position corresponding to the mark value is used as the registration reference point.

2. The mask stripe registration method according to claim 1, characterized in that: Step S02 also includes: presetting a fine threshold, when the absolute value of the difference after difference calculation is less than the fine threshold, assigning 0 to the difference at this point, then accumulating and summing the differences to obtain a fine difference sum, and saving it in the corresponding fine registration result matrix B; The method further comprises the following steps: S04: taking the adjacent connected domain with a mark value as a region mark bit, and counting the number of region mark bits that meet the requirements in the region corresponding to the sub-strip: if there are multiple region mark bits in the region, it is considered that the registration reference point position marking the position of this sub-strip cannot be referenced by other strips; if there is only one region mark bit in the region, the position corresponding to this region mark bit needs to be saved in the sub-strip registration reference point array; S05: Compare the position of the registration reference point obtained based on the rough registration result matrix A with the corresponding position in the fine registration result matrix B, and find the minimum fine difference sum in the fine registration result matrix B in the area of ​​the corresponding position, and after finding all the minimum fine difference sums, obtain the corresponding fine registration reference point position; S06: The fine registration reference points obtained in step S05 are referenced according to the reference requirements specified in step S04 and saved in the sub-strip registration reference point array.

3. The mask stripe registration method according to claim 2, characterized in that: Also includes the steps: S07: After the registration of each column of sub-strips is completed, the arrays of all sub-strip registration reference point arrays of the entire column that cannot be referenced are fitted with the registration reference point positions of other arrays that can be referenced to obtain the registration reference points here.

4. The mask stripe registration method according to claim 1, characterized in that: The specific steps of sliding the Mask image by pixel step by pixel are as follows: a: At the beginning of the row, slide the mask image one pixel at a time in the row direction. After sliding in the row direction, return to the beginning of the row and slide the mask image one pixel downward in the column direction. b: Perform step a in a new line; c: Repeat step b until the last row slides.

Images