A large-size precision measurement method and system based on CD-SEM

Abstract

This invention provides a large-size precision measurement method and system based on CD-SEM, applied in the field of precision measurement technology. During continuous platform movement, it triggers line-by-line scanning in real time and performs dynamic compensation to obtain high-precision large-size images. Line-by-line sampling is triggered by displacement pulses, while platform displacement is monitored in real time. The positional deviation is calculated as a compensation amount and superimposed on the deflection drive signal to achieve dynamic correction of the scanned points, ensuring the geometric consistency and accuracy of the image. During calibration, the pixel distance of the standard graphic is compared with the actual platform displacement to calculate the corresponding pixel equivalent, which is used for calculating the physical size of the image. This invention effectively solves the accuracy degradation problem caused by displacement errors, vibration, or uneven sampling in traditional CD-SEM multi-frame stitching. It has the advantages of high efficiency, wide range, and measurability, and is suitable for online inspection and quality control in the semiconductor and precision machining fields.

Description

Technical Field

[0001] This invention belongs to the field of precision measurement technology, specifically relating to a large-size precision measurement method and system based on CD-SEM. Background Technology

[0002] Scanning electron microscopy (SEM) is characterized by high resolution and strong imaging contrast, and is widely used in semiconductor manufacturing and precision machining for critical dimension (CD) inspection, micro / nano structure defect observation, and morphology analysis. In particular, CD-SEM equipment can inspect geometric parameters such as linewidth, spacing, aperture, and edge contours at the sub-micron and even nanoscale, making it an important tool for quality control in advanced process technologies. However, existing CD-SEMs typically use a fixed field of view (FOV) scanning method for dimensional measurements. The imaging range of a single frame is limited by factors such as electron beam scanning amplitude, imaging distortion control, and sampling bandwidth, resulting in a relatively small directly measurable area. When measuring a larger area of ​​structure (e.g., long distances across multiple single-frame fields of view, the overall size of an array structure, cross-regional alignment deviations, or the contour morphology of large-sized graphics), traditional CD-SEMs often require multi-point repetitive positioning, multi-frame stitching, or integration with optical measurement equipment.

[0003] In practical applications, existing technologies typically suffer from the following problems: First, multi-frame discrete sampling requires frequent stop-and-go positioning and refocusing, reducing measurement throughput and easily introducing repetitive positioning errors; Second, multi-frame image stitching often relies on nominal platform displacement or image feature registration. If the platform motion has servo following errors, speed fluctuations, micro-vibrations, or instantaneous drift, it can easily lead to distortion of the geometric relationship between adjacent frames or adjacent scan lines, causing problems such as inconsistent stitching scales, edge position jitter, and local stretching and compression, ultimately making the stitched large-size image difficult to use as a reliable metrological basis; Third, if only the feedback from the platform's grating ruler / encoder is used to generate... Displacement pulses are used to trigger scanning and sampling. This displacement feedback is usually closer to the nominal displacement in the control sense, which is difficult to fully reflect the transient error and vibration components of the platform at the nanoscale. This leads to a discrepancy between the "real physical position corresponding to the trigger moment" and the nominal position, resulting in the offset of the scanning point and the non-strict equidistant spacing of the mapping lines. Fourth, in the existing technology, the mapping between the pixel scale and the physical scale (pixel equivalent) is often decoupled from the scanning control, motion error compensation, and large-size mapping process. This results in insufficient stability of the calibration results under different speeds, different paths, or different motion states, and the repeatability and traceability of the measurement results are difficult to meet the requirements of high-precision measurement.

[0004] Therefore, in response to the need to achieve continuous precision measurement of a large size range under the small field of view of CD-SEM, the existing technology urgently needs a measurement system and method that can simultaneously complete line-by-line sampling during the continuous movement of the motion platform, dynamically compensate for the instantaneous position deviation introduced by the platform movement, and establish a stable and reliable mapping relationship between pixels and physical scale, so as to maintain nanometer-level geometric consistency and dimensional measurement accuracy while expanding the measurable range. Summary of the Invention

[0005] In view of the above-mentioned problems in the prior art, the purpose of this invention is to provide a large-size precision measurement method based on CD-SEM, which takes into account measurement accuracy, stability and measurement efficiency, and is suitable for application in large-size high-precision measurement scenarios in semiconductor manufacturing and precision processing.

[0006] A large-size precision measurement method based on CD-SEM includes the following steps: System calibration: Place the standard block or standard wafer on the motion platform, control the motion platform to move at a constant speed with a preset angle and step distance, and use a scanning electron microscope to scan and sample the calibration pattern area of ​​the standard. Based on the pixel distance of the predetermined features in the calibration pattern in the image and the actual physical length corresponding to the position feedback of the motion platform and the displacement feedback of the laser interferometer, calculate the pixel equivalent including at least the X-axis direction and the Y-axis direction. Image acquisition and construction: The wafer under test is fixed and the motion platform is controlled to move along a preset scanning path at a constant speed. The grating ruler module outputs an image acquisition trigger signal based on the displacement pulse of the motion platform to trigger the scanning electron microscope electron beam to deflect and sample line by line. A one-dimensional image line sequence arranged in displacement order is obtained by acquiring line by line. At the same time, the real-time position offset signal output by the laser interferometer is acquired and the electron beam scanning point is dynamically compensated based on the real-time position offset signal to correct the influence of vibration and position deviation caused by the motion platform on the scanning point. Then, the one-dimensional image line sequence is constructed into a two-dimensional grayscale image according to the acquisition order. The X-axis direction of the two-dimensional grayscale image corresponds to the physical displacement of the motion platform, and the Y-axis direction of the two-dimensional grayscale image corresponds to the row scanning pixel coordinates of the scanning electron microscope. Image processing and edge extraction measurement: edge extraction is performed on the two-dimensional grayscale image to obtain edge point cloud data of the workpiece contour, and the edge point cloud data is converted into physical dimensions based on the pixel equivalent, thereby calculating at least one geometric dimension parameter of the workpiece.

[0007] Preferably, the image acquisition trigger signal is generated by the scanning sampling signal generation module after counting, frequency division or interpolation of the displacement pulse output by the grating ruler module, so that the physical displacement interval between two adjacent one-dimensional image lines is a preset line spacing.

[0008] Preferably, the preset line spacing and the row scanning pixel spacing of the scanning electron microscope in the Y-axis direction satisfy a preset proportional relationship, so that the sampling density of the two-dimensional grayscale image in the X-axis direction and the Y-axis direction meets the spatial sampling requirements of equal proportion or anisotropy.

[0009] Preferably, constructing a two-dimensional grayscale image from a sequence of one-dimensional image lines includes associating each one-dimensional image line with its corresponding displacement pulse count value or equivalent physical displacement value, and arranging and splicing the one-dimensional image lines in the order of the displacement pulse count value or equivalent physical displacement value to form a two-dimensional grayscale image.

[0010] Preferably, during the construction of the two-dimensional grayscale image, the displacement non-uniformity between adjacent one-dimensional image lines is corrected. The correction includes resampling or interpolation alignment of the arrangement position of the one-dimensional image lines based on the displacement feedback of the laser interferometer.

[0011] Preferably, the dynamic compensation control includes calculating the instantaneous position deviation of the motion platform in the X-axis and Y-axis directions based on the real-time position offset signal, and converting the instantaneous position deviation into an electron beam deflection compensation amount to be superimposed on the deflection drive signal of the electron beam, so that the actual scanning point of the electron beam is corrected according to the instantaneous position change of the motion platform.

[0012] Preferably, the dynamic compensation control is implemented digitally, specifically by: outputting the instantaneous position deviation as a digital quantity to the scanning sampling signal generation module; the scanning sampling signal generation module performs coordinate conversion and proportional calculation, outputs a digital compensation quantity, and generates an analog compensation signal through digital-to-analog conversion; and then superimposes the analog compensation signal onto the electron beam deflection drive signal.

[0013] Preferably, the dynamic compensation control is implemented in an analog manner, specifically by converting the instantaneous position deviation into an analog compensation signal after proportional calculation, and then superimposing it with the electron beam deflection drive signal through an adder circuit before outputting it to the deflector.

[0014] Preferably, the edge extraction employs a sub-pixel edge detection algorithm, which includes curve fitting or interpolation to find peaks based on gray-level gradient distribution within the neighborhood of pixel-level edge candidate points, in order to obtain the sub-pixel position of the edge within the pixel.

[0015] The second objective of this invention is to provide a large-size precision measurement system based on CD-SEM, comprising: A motion platform is used to carry and drive the standard or the wafer under test along a preset scanning path; The motion control module is used to control the motion platform to move at a constant speed during the effective sampling segment based on the coordinate sequence and / or starting position defined by the external input. The grating ruler module is used to acquire the nominal displacement of the motion platform and output a displacement pulse corresponding to the nominal displacement; A laser interferometer correction module is used to output the displacement measurement values ​​of the motion platform in the X and / or Y directions in real time. The scanning sampling signal generation module is connected to the grating ruler module, the laser interferometer correction module, and the electron beam deflection unit of the CD-SEM. The scanning sampling signal generation module is configured to: generate an image acquisition trigger signal based on the displacement pulse to trigger the electron beam to deflect and sample line by line, thereby acquiring a one-dimensional image line sequence arranged in displacement order; form a real-time position offset based on the displacement measurement value output by the laser interferometer correction module and the position information of the motion platform, and generate a deflection compensation amount based on the real-time position offset and superimpose it on the electron beam deflection drive signal to dynamically compensate and control the electron beam scanning point; An image processing unit, connected to the acquisition output of a CD-SEM, is configured to: construct a two-dimensional grayscale image from the one-dimensional image line sequence according to the acquisition order, wherein the X-axis direction of the two-dimensional grayscale image corresponds to the physical displacement of the motion platform, and the Y-axis direction of the two-dimensional grayscale image corresponds to the row scan pixel coordinates of the CD-SEM; perform edge extraction on the two-dimensional grayscale image to obtain edge point cloud data, and convert the edge point cloud data into physical dimensions based on pixel equivalents to calculate at least one geometric dimension parameter of the workpiece; The measurement system further includes a calibration module or calibration program, which is configured to: calculate the pixel equivalent in the X direction and the pixel equivalent in the Y direction based on the pixel distance of the calibration pattern of the standard in the two-dimensional grayscale image and the actual physical length corresponding to the position feedback of the motion platform and the displacement measurement value of the laser interferometer correction module, and provide the pixel equivalent to the image processing unit for the calculation of the geometric dimension parameters.

[0016] The beneficial effects of this invention are: By using the grating ruler displacement pulse as the master clock for image acquisition, the electron beam is deflected and sampled line by line to form a one-dimensional image line sequence arranged in displacement order. The one-dimensional image line sequence is then constructed into a two-dimensional grayscale image according to the acquisition order, so that the X-axis direction of the two-dimensional grayscale image is directly correlated with the physical displacement of the motion platform. This enables imaging and measurement of continuous areas larger than the single-frame field of view in CD-SEM under the limited field of view of a single frame. This avoids the problems of long measurement cycle and accumulation of repeated positioning errors caused by traditional multi-point stop-and-go positioning and multi-frame discrete sampling, and significantly improves the continuity and throughput of large-size measurements.

[0017] To address the issues of servo following errors, speed fluctuations, and micro-vibrations in continuous motion mapping that can lead to distortions in geometric relationships between adjacent scan lines, inconsistent stitching scales, and edge jitter, this invention introduces a laser interferometer correction module to output the platform displacement in real time while triggering sampling line by line. This output is combined with platform position feedback to generate a real-time position offset. , The scanning sampling signal generation module is based on , The deflection compensation amount is generated and superimposed on the electron beam deflection drive signal, so that the electron beam scanning point position is corrected in real time according to the instantaneous deviation of the platform. This suppresses the drift of the scanning point position in the continuous motion state of the platform, ensures the consistency of the actual physical position corresponding to the one-dimensional image line, and improves the geometric stability and edge positioning stability of the constructed two-dimensional grayscale image. This makes the generated large-size two-dimensional image a reliable metrological basis rather than just a display image.

[0018] To address the issue that triggering based solely on platform nominal displacement may introduce non-strict equidistant line spacing and scale stretching / compression, this invention employs a laser interferometer correction module for real-time displacement-based dynamic compensation, preventing instantaneous positional deviations from directly translating into scanning point errors. Furthermore, in one optional implementation, each one-dimensional image line is associated with a displacement pulse count or interferometer displacement value, and the line sequence arrangement is resampled or interpolated based on interferometer displacement feedback. This further reduces scale errors introduced by platform speed fluctuations and pulse quantization, improving the scale consistency and traceability of large-scale mapping in the first direction.

[0019] System calibration is performed before measurement, and the pixel distance of the predefined features in the two-dimensional image is determined using a standard. , The actual physical length corresponding to the position feedback of the motion platform and the displacement feedback of the laser interferometer , Calculate the pixel equivalents in the X-axis and Y-axis directions. , The pixel equivalent is used for subsequent conversion of edge point cloud to physical size. Since the physical length benchmark of the pixel equivalent introduces the displacement feedback calibration of the interferometer, and the calibration acquisition process also adopts dynamic compensation to ensure geometric consistency, the pixel equivalent still has good stability and repeatability under continuous motion mapping conditions, thereby improving the absolute accuracy and consistency of the final size measurement.

[0020] After constructing a two-dimensional grayscale image, edge extraction is performed to generate edge point cloud data. In one optional implementation, sub-pixel edge detection is used to obtain sub-pixel level edge point cloud data, which is then combined with pixel equivalents. , The physical coordinate conversion is completed, and geometric dimensional parameters such as length, width, spacing, aperture, and inner and outer diameter are output through geometric fitting methods such as straight line fitting, circle fitting, ellipse fitting or spline fitting. Since the aforementioned dynamic compensation and mapping correction reduce edge jitter and geometric distortion, the geometric fitting residual of the edge point cloud is smaller and the measurement results are more stable, thereby improving the reliability and repeatability of large-size precision measurement.

[0021] In summary, this invention forms a unified metrological closed loop by integrating processes such as "displacement pulse-triggered line-by-line sampling and mapping", "interferometer real-time offset dynamic compensation and correction of scanning points", "pixel equivalent calibration and physical conversion", and "edge point cloud fitting to determine dimensions". While expanding the measurable range of CD-SEM, it suppresses geometric distortion and scale error introduced by continuous motion, taking into account measurement accuracy, stability and measurement efficiency. It is suitable for application in large-size, high-precision measurement scenarios in semiconductor manufacturing and precision machining. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a system block diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] Example 1 like Figure 1 As shown, a large-size precision measurement system based on CD-SEM includes at least a main control PC, a motion control module, a motion platform, a grating ruler module, a laser interferometer correction module, a scanning sampling signal generation module, an electron beam deflection unit, an image acquisition unit, and an image processing unit.

[0024] The main control PC is used to input the measurement coordinate sequence, image acquisition parameters, and calibration parameters, and sends the starting position definition and path parameters to the motion control module. At the same time, it sends the image acquisition trigger parameters, line spacing parameters, and compensation parameters to the scanning sampling signal generation module.

[0025] The motion control module is used to drive the motion platform into the preset scanning path and maintain a constant speed during the effective sampling segment to ensure the stability of the platform's motion during line-by-line sampling.

[0026] The grating ruler module is used to provide real-time feedback on the platform displacement and output a displacement pulse corresponding to the displacement. This displacement pulse is used as the master clock for image acquisition triggering.

[0027] The laser interferometer calibration module is used to measure the displacement of the motion platform in the X-axis and Y-axis directions in real time and output the corresponding real-time position offset information.

[0028] The platform position feedback obtained through the grating ruler module is used to characterize the nominal displacement of the moving platform and trigger line-by-line sampling, while the displacement feedback from the laser interferometer is used to characterize the actual displacement of the moving platform. Specifically, during the acquisition process, the difference between the two forms the real-time position offset. , It is used to generate deflection compensation to achieve real-time correction of scan points; in calibration calculations, both are used to determine the distance to pixels. , Corresponding actual physical length , Thus, the pixel equivalent is calculated. , And establish a measurement mapping between the pixel domain and the physical domain.

[0029] The scanning sampling signal generation module generates electron beam deflection and sampling control signals based on the received displacement pulse trigger signal. After receiving the real-time position offset information output by the laser interferometer correction module, it generates a compensation amount and superimposes it on the deflection drive signal, so that the electron beam scanning point position is corrected in real time according to the instantaneous deviation of the platform.

[0030] After each line scan sampling, the image acquisition unit forms a one-dimensional image line and outputs it to the image processing unit. The image processing unit arranges and stitches the continuous one-dimensional image lines according to the triggering order to construct a two-dimensional grayscale image. Based on the two-dimensional grayscale image, it performs edge extraction, point cloud generation and size calculation to output the final geometric size parameters.

[0031] Example 2 like Figure 2 As shown, a large-size precision measurement method based on CD-SEM is adopted. The measurement system is a combination system of CD-SEM platform and two-dimensional motion platform. The system includes at least a main control PC, motion control module, motion platform, grating ruler module, laser interferometer correction module, scanning sampling signal generation module, electron beam deflection unit, image acquisition unit and image processing unit.

[0032] The grating ruler module outputs a displacement pulse corresponding to the nominal displacement of the motion platform to generate an image acquisition trigger signal. The laser interferometer correction module outputs the real-time displacement or real-time position offset of the motion platform to generate a dynamic compensation amount. The scanning sampling signal generation module triggers the electron beam to scan line by line when the image acquisition trigger signal arrives, and superimposes the dynamic compensation amount onto the deflection drive signal during the same line scan to correct the vibration or instantaneous position deviation caused by the motion platform's movement. This ensures that the subsequently constructed two-dimensional grayscale image establishes a measurable and stable mapping relationship between the physical displacement of the motion platform in the X-axis direction.

[0033] This large-size precision measurement method based on CD-SEM specifically includes the following steps: Step 1: System Calibration First, the standard gauge block or standard wafer is fixed on the support platform of the motion platform, ensuring that the position of the standard remains unchanged during the calibration process. The standard has at least one calibration pattern area with a known physical length (e.g., a grid structure with known line spacing, a circular hole structure with known aperture, or an alignment mark structure with known edge spacing). The main control PC sets the calibration path parameters and sends them to the motion control module, causing the motion platform to move at a constant speed along a preset path and enter the effective sampling segment. The preset path is set to have a non-zero angle relative to the X-axis direction of the motion platform, so that the calibration pattern simultaneously generates pixel changes in both the X-axis and Y-axis directions in the constructed image, thereby enabling joint calibration of the pixel equivalents in the X-axis and Y-axis directions.

[0034] After the motion platform enters the effective sampling segment, the grating ruler module outputs displacement pulses. The scanning sampling signal generation module counts, divides, or interpolates the displacement pulses to obtain the image acquisition trigger signal, so that two adjacent sampled image lines correspond to a preset line spacing. And when each image acquisition trigger signal arrives, the key-size scanning electron microscope (CD-SEM) is triggered to complete a line scan sampling to obtain a one-dimensional image line.

[0035] At the same time, the laser interferometer correction module outputs the actual displacement of the motion platform in real time and forms a real-time position offset with the nominal displacement of the motion platform obtained by the grating ruler module. , The scanning sampling signal generation module generates signals based on real-time position offset. , The deflection compensation amount is generated and superimposed on the electron beam deflection drive signal, so that the electron beam scanning point position is corrected in real time according to the instantaneous deviation of the motion platform, thereby ensuring the geometric consistency of the calibration acquisition process.

[0036] The image acquisition unit outputs the continuously acquired one-dimensional image lines to the image processing unit in the triggering order. The image processing unit arranges and stitches the one-dimensional image line sequence in the acquisition order to construct a two-dimensional grayscale image. The X-axis direction of the two-dimensional grayscale image corresponds to the physical displacement of the motion platform, and the Y-axis direction of the two-dimensional grayscale image corresponds to the pixel coordinates of the CD-SEM row scan.

[0037] Subsequently, the image processing unit extracts predetermined features of the calibration pattern from the two-dimensional grayscale image and calculates its pixel distance in the image. , Simultaneously, the actual physical lengths corresponding to the position feedback of the motion platform and the displacement feedback of the laser interferometer are read. , The pixel equivalent in the X-axis direction was calculated respectively. Pixel equivalent in the Y-axis direction And write it into the measurement parameter table; where the actual physical length , The nominal displacement of the motion platform can be calibrated or corrected by feedback from the interferometer to reduce quantization error, and the pixel equivalent can be adjusted after repeated acquisition at different positions of the standard. , Take the mean or weighted mean to improve the stability and reproducibility of pixel equivalents.

[0038] Step 2: Image Acquisition and Construction for Actual Measurement After system calibration, the wafer under test is fixed on the carrier platform of the motion platform. The main control PC inputs the measurement coordinate sequence, the range of the area to be tested, and the image acquisition parameters according to customer requirements and sends them to the motion control module and the scanning sampling signal generation module. The motion control module drives the platform to enter the preset scanning path from the starting position and maintain a constant speed during the effective sampling segment. In an optional embodiment, the preset scanning path is a row-by-row reciprocating or serpentine path, and the turning segments of adjacent scanning rows are set as non-effective sampling segments, thereby avoiding inconsistent sampling density caused by acceleration, deceleration, or turning, and ensuring continuous coverage of the effective area of ​​the constructed image without missing rows.

[0039] After the motion platform enters the effective sampling segment, the grating ruler module continuously outputs displacement pulses. The scanning sampling signal generation module counts, divides, or interpolates the displacement pulses to generate an image acquisition trigger signal to determine the physical displacement interval corresponding to adjacent one-dimensional image lines. The electron beam is deflected and sampled line by line to obtain a one-dimensional image line each time an image acquisition trigger signal arrives; wherein, the physical displacement interval It can be set according to the measurement resolution and throughput requirements, and can further meet the preset ratio relationship with the pixel spacing of CD-SEM line scanning, so that the two-dimensional grayscale image meets the requirements of equal or anisotropic sampling density in the X-axis and Y-axis directions.

[0040] At the same time, the laser interferometer correction module outputs the actual displacement information of the motion platform in real time and forms a real-time position offset with the nominal displacement fed back by the position of the motion platform. , The scanning sampling signal generation module generates signals based on real-time position offset. , The deflection compensation amount is generated and superimposed on the electron beam deflection drive signal in real time to correct the influence of vibration or instantaneous position deviation caused by the continuous motion of the motion platform on the scanning point position, so that the physical displacement position corresponding to each one-dimensional image line in the X-axis direction remains consistent, thereby ensuring that the two-dimensional grayscale image after the line sequence is stitched together has measurable geometric consistency.

[0041] The essence of the above dynamic compensation is to shift the real-time position. , It is converted into a compensation amount and superimposed on the electron beam deflection drive signal to achieve scanning point correction.

[0042] In one alternative implementation, dynamic compensation is implemented digitally, i.e., real-time position offset. , The digital input is used to generate the scanning sampling signal module. The scanning sampling signal generation module performs proportional conversion to obtain the digital compensation amount, and then generates an analog compensation signal through digital-to-analog conversion, which is then superimposed on the deflection drive signal.

[0043] In another alternative implementation, dynamic compensation is achieved through simulation, i.e., real-time position offset. , The analog compensation signal is directly generated through proportional calculation and then superimposed onto the deflection drive signal through an analog adder circuit to reduce link delay and improve the response to high-frequency disturbances.

[0044] Both of the above implementation methods can achieve the purpose of dynamically compensating for the scanning points while continuously sampling and mapping line by line.

[0045] The image acquisition unit outputs a sequence of one-dimensional image lines obtained in the triggering order to the image processing unit. The image processing unit arranges and stitches the one-dimensional image line sequence according to the triggering order to construct a two-dimensional grayscale image, with the X-axis of the two-dimensional grayscale image corresponding to the platform's physical displacement and the Y-axis corresponding to the pixel coordinates of the CD-SEM row scan. In a further optional embodiment, the image processing unit associates each one-dimensional image line with its displacement pulse count value or interferometer displacement value, and performs resampling or interpolation alignment based on the interferometer displacement feedback to address the residual non-uniformity of the line spacing, thereby further reducing scale errors caused by platform speed fluctuations or pulse quantization.

[0046] Step 3: Image Processing and Edge Extraction Measurement After obtaining the two-dimensional grayscale image, the image processing unit can selectively perform filtering and / or binarization preprocessing on the two-dimensional grayscale image according to the image quality to suppress noise and enhance the gradient features of the workpiece contour edge; then, edge extraction is performed on the two-dimensional grayscale image to obtain the edge point cloud data of the workpiece contour.

[0047] In one possible implementation, the edge extraction employs a sub-pixel edge detection algorithm, which first locates pixel-level edge candidate points based on gradient magnitude and gradient direction, then constructs a grayscale profile along the gradient direction in the neighborhood of the candidate points, and performs parabolic fitting or Gaussian fitting on the data near the gradient extremum to obtain the sub-pixel position of the edge within the pixel, thereby outputting the sub-pixel edge point coordinates and forming sub-pixel precision edge point cloud data.

[0048] Subsequently, the image processing unit calls the pixel equivalent obtained from step one. , The edge point cloud is converted from pixel coordinates to physical coordinates and at least one geometric dimension parameter of the workpiece is calculated; wherein, the geometric dimension parameter includes at least one of length, width, spacing, line width, hole diameter, outer diameter, inner diameter, roundness, rectangularity or polygonal contour dimension.

[0049] In one possible implementation, the image processing unit performs geometric fitting on the physical coordinate point cloud and calculates the corresponding dimensions using fitting parameters. Geometric fitting includes line fitting, circle fitting, ellipse fitting, polyline fitting, or spline fitting. For example, the least squares circle fitting is performed on the point cloud of the circular hole contour to obtain the radius R and output the diameter D=2R. Straight lines are fitted to the two sides of the line respectively, and the vertical distance between the two straight lines is used as the line width or spacing.

[0050] After the calculation is completed, the system will output and store the measurement results along with quality indicators such as measurement location and fitting residuals for verification and traceability.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for large size precision measurement based on CD-SEM, characterized in that, Includes the following steps: System calibration: Place the standard block or standard wafer on the motion platform, control the motion platform to move at a constant speed with a preset angle and step distance, and use a scanning electron microscope to scan and sample the calibration pattern area of ​​the standard. Based on the pixel distance of the predetermined features in the calibration pattern in the image and the actual physical length corresponding to the position feedback of the motion platform and the displacement feedback of the laser interferometer, calculate the pixel equivalent including at least the X-axis direction and the Y-axis direction. Image acquisition and construction: The wafer under test is fixed and the motion platform is controlled to move along a preset scanning path at a constant speed. The grating ruler module outputs an image acquisition trigger signal based on the displacement pulse of the motion platform to trigger the scanning electron microscope electron beam to deflect and sample line by line. A one-dimensional image line sequence arranged in displacement order is obtained by acquiring line by line. At the same time, the real-time position offset signal output by the laser interferometer is acquired and the electron beam scanning point is dynamically compensated based on the real-time position offset signal to correct the influence of vibration and position deviation caused by the motion platform on the scanning point. Then, the one-dimensional image line sequence is constructed into a two-dimensional grayscale image according to the acquisition order. The X-axis direction of the two-dimensional grayscale image corresponds to the physical displacement of the motion platform, and the Y-axis direction of the two-dimensional grayscale image corresponds to the row scanning pixel coordinates of the scanning electron microscope. Image processing and edge extraction measurement: edge extraction is performed on the two-dimensional grayscale image to obtain edge point cloud data of the workpiece contour, and the edge point cloud data is converted into physical dimensions based on the pixel equivalent, thereby calculating at least one geometric dimension parameter of the workpiece.

2. The CD-SEM-based large-size precision measurement method according to claim 1, wherein The image acquisition trigger signal is generated by the scanning sampling signal generation module after counting, frequency division or interpolation of the displacement pulse output by the grating ruler module, so that the physical displacement interval between two adjacent one-dimensional image lines is a preset line spacing.

3. The CD-SEM-based large-size precision measurement method according to claim 2, wherein The preset line spacing and the row scanning pixel spacing of the scanning electron microscope in the Y-axis direction satisfy a preset proportional relationship, so that the sampling density of the two-dimensional grayscale image in the X-axis direction and the Y-axis direction meets the preset spatial sampling requirements of equal proportion or anisotropy.

4. The large-size precision measurement method based on CD-SEM according to claim 1, characterized in that, Constructing a two-dimensional grayscale image from a sequence of one-dimensional image lines involves associating each one-dimensional image line with its corresponding displacement pulse count value or equivalent physical displacement value, and arranging and stitching the one-dimensional image lines in the order of the displacement pulse count value or equivalent physical displacement value to form a two-dimensional grayscale image.

5. The large-size precision measurement method based on CD-SEM according to claim 1 or 4, characterized in that, During the construction of the two-dimensional grayscale image, the displacement non-uniformity between adjacent one-dimensional image lines is corrected. The correction includes resampling or interpolation alignment of the arrangement position of the one-dimensional image lines based on the displacement feedback of the laser interferometer.

6. The large-size precision measurement method based on CD-SEM according to claim 1, characterized in that, The dynamic compensation control includes calculating the instantaneous position deviation of the motion platform in the X-axis and Y-axis directions based on the real-time position offset signal, and converting the instantaneous position deviation into an electron beam deflection compensation amount to be superimposed on the deflection drive signal of the electron beam, so that the actual scanning point of the electron beam is corrected according to the instantaneous position change of the motion platform.

7. The large-size precision measurement method based on CD-SEM according to claim 6, characterized in that, The dynamic compensation control is implemented digitally, specifically by outputting the instantaneous position deviation as a digital quantity to the scanning sampling signal generation module. The scanning sampling signal generation module performs coordinate conversion and proportional calculation, outputs a digital compensation quantity, and generates an analog compensation signal through digital-to-analog conversion. The analog compensation signal is then superimposed on the electron beam deflection drive signal.

8. The large-size precision measurement method based on CD-SEM according to claim 6, characterized in that, The dynamic compensation control is implemented in an analog manner, specifically by converting the instantaneous position deviation into an analog compensation signal after proportional calculation, and then superimposing it with the electron beam deflection drive signal through an additive circuit before outputting it to the deflector.

9. The large-size precision measurement method based on CD-SEM according to claim 1, characterized in that, The edge extraction employs a sub-pixel edge detection algorithm, which includes curve fitting or interpolation to find peaks based on gray-level gradient distribution within the neighborhood of pixel-level edge candidate points, in order to obtain the sub-pixel position of the edge within the pixel.

10. A large-size precision measurement system based on CD-SEM, characterized in that, include: A motion platform is used to carry and drive the standard or the wafer under test along a preset scanning path; The motion control module is used to control the motion platform to move at a constant speed during the effective sampling segment based on the coordinate sequence and / or starting position defined by the external input. The grating ruler module is used to acquire the nominal displacement of the motion platform and output a displacement pulse corresponding to the nominal displacement; A laser interferometer correction module is used to output the displacement measurement values ​​of the motion platform in the X and / or Y directions in real time. The scanning sampling signal generation module is connected to the grating ruler module, the laser interferometer correction module, and the electron beam deflection unit of the CD-SEM. The scanning sampling signal generation module is configured to: generate an image acquisition trigger signal based on the displacement pulse to trigger the electron beam to deflect and sample line by line, thereby acquiring a one-dimensional image line sequence arranged in displacement order; form a real-time position offset based on the displacement measurement value output by the laser interferometer correction module and the position information of the motion platform, and generate a deflection compensation amount based on the real-time position offset and superimpose it on the electron beam deflection drive signal to dynamically compensate and control the electron beam scanning point; An image processing unit, connected to the acquisition output of a CD-SEM, is configured to: construct a two-dimensional grayscale image from the one-dimensional image line sequence according to the acquisition order, wherein the X-axis direction of the two-dimensional grayscale image corresponds to the physical displacement of the motion platform, and the Y-axis direction of the two-dimensional grayscale image corresponds to the row scan pixel coordinates of the CD-SEM; perform edge extraction on the two-dimensional grayscale image to obtain edge point cloud data, and convert the edge point cloud data into physical dimensions based on pixel equivalents to calculate at least one geometric dimension parameter of the workpiece; The measurement system further includes a calibration module or calibration program, which is configured to: calculate the pixel equivalent in the X direction and the pixel equivalent in the Y direction based on the pixel distance of the calibration pattern of the standard in the two-dimensional grayscale image and the actual physical length corresponding to the position feedback of the motion platform and the displacement measurement value of the laser interferometer correction module, and provide the pixel equivalent to the image processing unit for the calculation of the geometric dimension parameters.

Images