CN-121998946-A - Line width measurement method based on CDSEM image

CN121998946ACN 121998946 ACN121998946 ACN 121998946ACN-121998946-A

Abstract

The invention discloses a line width measurement method based on a CDSEM image, which comprises the steps of obtaining an original image and preprocessing, carrying out gray value accumulation and average on a plurality of equally divided sub-ROIs (regions of interest) in the image to generate a projection curve, carrying out filtering on the projection curve to generate a smooth curve, carrying out differential calculation on the smooth curve to generate a differential curve, searching and identifying all candidate extreme points on the differential curve, calculating a dynamic threshold value based on differential values of the candidate extreme points, screening out maximum points and minimum points meeting the dynamic threshold value, mapping the maximum points and the minimum points back to the smooth curve, calculating to obtain edge feature point coordinates of sub-pixel precision, fitting a plurality of edge feature point coordinates to obtain a fitting linear equation, filtering the edge feature points, and calculating width values among the point pairs based on corresponding fitting straight lines of the edge feature point pairs to obtain the line width. The invention has the advantages of accurate and reliable line width measurement, and the like.

Inventors

SHEN YUNBO
CHEN QINGGUANG
FAN JIANGHUA
LONG HUIYUE
CHEN LONG
LI TAO
LI XUANLUN
Kuang Guoyu

Assignees

中国电子科技集团公司第四十八研究所

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. A line width measurement method based on CDSEM image, comprising the steps of: s1, acquiring an original image acquired by a CDSEM and preprocessing the original image to obtain a preprocessed image; S2, aiming at the vertical line width or the horizontal line width in the preprocessed image, carrying out gray value accumulation and averaging on a plurality of equally divided sub-ROIs in a preset region of interest (ROI), and generating a projection curve representing the gray change of the line width edge; S3, filtering the projection curve to inhibit noise and generate a smooth curve; S4, performing differential calculation on the smooth curve to generate a differential curve for representing the gradient change of the edge position; S5, searching and identifying all candidate maximum value points and minimum value points on the differential curve based on a preset neighborhood searching radius; S6, respectively calculating a maximum dynamic threshold value and a minimum dynamic threshold value based on differential values of all candidate maximum points and minimum points, and screening out the maximum points and the minimum points meeting the dynamic threshold value conditions; s7, mapping the maximum value point and the minimum value point back to the smooth curve, and calculating to obtain edge feature point coordinates with sub-pixel precision in the neighborhood of the extreme value point based on a gray threshold value and an interpolation fitting method; S8, carrying out robust straight line fitting on the edge feature point coordinates by adopting a weighted self-adaptive least square method to obtain a fitting straight line equation representing the line width edge; s9, judging inner points and outer points of edge feature points involved in fitting, and filtering the outer points; s10, calculating the width value between each point pair based on the edge characteristic point pair reserved after the outer points are filtered and the corresponding fitting straight line, and further counting to obtain the line width.
2. The line width measurement method based on CDSEM image according to claim 1, wherein in step S1, the specific process of preprocessing is: Firstly, carrying out multi-frame superposition on an original image so as to increase the electron dose and improve the brightness of the image; and secondly, sequentially performing Gaussian filtering, median filtering, gamma correction and histogram equalization on the superimposed image to generate a preprocessed image.
3. The line width measurement method based on CDSEM image according to claim 1, wherein in step S2, gray scale value accumulation and averaging are performed on the multiple equally divided sub-ROIs by using a vertical or horizontal pixel projection algorithm, specifically: Dividing a preset left rectangle ROI and a preset right rectangle ROI into N sub-ROIs along the height direction of the right rectangle ROI, and calculating coordinates of all pixel points in each sub-ROI in an image coordinate system; acquiring gray values corresponding to all pixel coordinate points; For each sub-ROI, the pixel gray values of each column or each row in the sub-ROI are accumulated and averaged along the projection direction, so as to generate a projection gray curve of the sub-ROI.
4. A line width measurement method based on CDSEM image according to claim 1, 2 or 3, wherein in step S3, the calculation formula of the smoothed curve is: Wherein the method comprises the steps of Representing the position in the image ) A smoothed gray value; Representing the position in the image ) An original gray value of the position, wherein p is a smoothing coefficient; Representing the total length of the smoothing window.
5. The line width measurement method based on CDSEM image according to claim 4, wherein the specific procedure of step S5 is: presetting a neighborhood searching radius r, and obtaining the s element from a differential curve The indexing is started and the index is started, s=r+1 r+2 in-2 p Q-r-1;q is the differential coefficient; If the s-th element satisfies Finding a candidate maximum point; If the s-th element satisfies The selected minimum point is considered found.
6. A line width measurement method based on CDSEM image according to claim 1,2 or 3, wherein in step S6, a dynamic threshold is preset Calculating maximum values in the maximum value point and the minimum value point respectively And minimum value Then by And searching and screening maximum value points and minimum value points which meet the dynamic threshold condition for the dynamic threshold.
7. A line width measurement method based on CDSEM image according to claim 1,2 or 3, wherein the specific procedure of step S7 is: mapping the screened extreme points to the corresponding smooth curves according to the index positions of the extreme points in the differential curves to obtain corresponding gray index points and gray values; searching a maximum gray value and a minimum gray value in a left-right range by taking the gray index point as a center; Calculating a target gray value according to the maximum gray value, the minimum gray value and a preset gray threshold percentage; And searching sub-pixel positions with gray values equal to target gray values on a smooth curve near the gray index points through a linear interpolation method, and taking the sub-pixel positions as final edge characteristic point coordinates.
8. A line width measurement method based on CDSEM image according to claim 1,2 or 3, wherein in step S8, each feature point is obtained by a weighted adaptive least square method Weights of (2) Adaptively calculating according to the distance d from the point to the fitting straight line: Where T is a distance threshold.
9. The line width measurement method based on the CDSEM image according to claim 1, 2 or 3, wherein the specific process of step S9 is that two points are randomly selected for straight line fitting, then distances from all points to straight lines are calculated, inner points and outer points are separated according to a distance threshold value, the random process is repeated, and the corresponding straight lines with the largest number of inner points are obtained as a final result.
10. The method for measuring line width based on CDSEM image according to claim 1, 2 or 3, wherein after step S10, the method further comprises step S11, wherein all the width values calculated in step S10 are sorted, after a part of data with the smallest value is removed, the statistical calculation in step S10 is re-executed to obtain the final optimized line width measurement result.

Description

Line width measurement method based on CDSEM image Technical Field The invention mainly relates to the technical field of semiconductor equipment, in particular to a line width measuring method based on a CDSEM image. Background With the increasing integration of semiconductor chips, the critical dimensions of devices have been rapidly reduced from micrometer scale (e.g., 125 μm) to nanometer scale (e.g., 50nm or even below 10 nm). This allows millions or more devices to be accommodated on a single chip, while placing extreme demands on the control of the process. Critical dimensions are a central indicator of the level of semiconductor manufacturing technology, and small variations can directly affect device performance, yield and reliability. Therefore, accurate, stable on-line measurements of critical dimensions during manufacturing are critical. Among many measurement techniques, a Critical Dimension Scanning Electron Microscope (CDSEM) has become a core device for real-time monitoring and line width measurement in semiconductor industrial production due to its high resolution and high accuracy. The CDSEM scans the sample surface by focusing an electron beam, detects the generated secondary electrons and other signals to generate an image, and quantitatively measures various microstructures (including line width, diameter of circular hole, gap distance, corners of various complex shapes and the like) on the wafer by using an image processing algorithm to find potential defects. The image linewidth measurement technology combines computer vision and image processing algorithm, and has been applied in quality inspection of parts in manufacturing industry to ensure the shape and size accuracy of the product. Although CDSEM measurement techniques are critical, when applied to complex pattern measurements on 8-inch and 12-inch wafers, the accuracy and stability (especially repeatability) of the measurements can be severely challenged by a series of complex image disturbance factors. These interference factors can be summarized as follows: 1. the imaging limitations inherent to electron optical systems are that the image quality of CDSEM is fundamentally limited by the resolution of the electron optical system. Specifically: The signal-to-noise ratio limitation is that the collection efficiency and the signal conversion sensitivity of the detector directly influence the signal-to-noise ratio and the definition of the image; aberration affects the spot size of the electron beam, and various aberrations such as spherical aberration, coma, astigmatism, distortion, chromatic aberration, etc., which can reduce the final resolution of the image, resulting in blurring or misregistration of the image edges. 2. Wafer self-characteristics and differences introduced by the measurement process: pattern dense and similarity interference, namely that a large number of repeated or similar patterns (such as densely arranged line widths) exist on a wafer, when a specific pattern is measured, adjacent similar patterns can generate signal interference, and the difficulty of accurate identification is increased. Positioning and pattern differences, i.e. between different measuring units of the same wafer or between the same positions of different wafers, the patterns themselves may have minor differences due to manufacturing process fluctuations. In addition, the inevitable random deviation of the CDSEM machine during positioning causes that the images shot during each measurement are not completely consistent, which brings difficulty to the measurement algorithm requiring a stable reference. Pattern complexity, namely various patterns to be measured (lines, holes, gaps, corners and the like), and the phenomena of shielding, shading and the like possibly occurring during imaging, further increases the complexity of image analysis. The interference factors and the existing relatively simple measurement algorithm form a direct causal relationship, and the existing algorithm often lacks sufficient consideration and robust design for the complex situations, so that the situation of large fluctuation of measured values and even misjudgment easily occurs when the existing algorithm faces to non-ideal image conditions. The repeatability of the measurement result, namely the consistency of the size values obtained by carrying out multiple measurements on the same graph feature, is difficult to ensure, and the severe requirement of the prior process on the measurement precision smaller than 1 nanometer cannot be met. The closest prior art solutions typically rely on simple gray threshold segmentation, edge detection or traditional model fitting of CDSEM images to extract line width information. These methods work to some extent, but their effectiveness depends heavily on the signal-to-noise ratio of the image and on the regularity of the graph. Disclosure of Invention Aiming at the technical problems existing in the prior art, the invent