Mask defect detection system

Abstract

The invention discloses a mask plate defect detection system, which relates to the technical field of mask plate defect detection, and comprises an objective table for placing a mask plate to be detected, an automatic focusing module, a laser scanning module, a signal collecting and imaging module and a signal processing module, the automatic focusing module comprises a first light source, a first receiver, a first lens group, a first mechanical shell, a first signal processing unit, a first motor and a driving part of the first motor; compared with a traditional pure vision scheme, the method has the advantages that the sensitivity is better, the algorithm pressure can be greatly reduced, and the system stability is improved; the automatic focusing module is matched with high-resolution imaging and laser scanning, so that micron-order and even submicron-order defects can be accurately detected; and meanwhile, the modular design can flexibly adapt to mask plates of different sizes and types, and is widely applied to the high-end manufacturing fields of semiconductors, photoelectrons, precision optics and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mask plate defect detection, in particular to a mask plate defect detection system. BACKGROUND

[0002] The mask plate is a core key component in high-end manufacturing fields such as semiconductors, optoelectronics, and precision optics, and the surface quality thereof directly determines the yield and performance of downstream products. For example, in the semiconductor lithography process, micron-level or even sub-micron-level micro defects (such as scratches, impurities, and pattern deviations) on the surface of the mask plate will be copied to the surface of the wafer, ultimately leading to chip functional failure. In the manufacturing of precision optical devices, mask plate defects can seriously affect the key indicators of the optical system, such as light transmission and imaging accuracy. Therefore, efficient and accurate detection of micron and sub-micron defects on the surface of the mask plate is an indispensable quality control link in the high-end manufacturing industry chain, and is directly related to the production efficiency and product competitiveness of the industry.

[0003] Currently, the mainstream mask plate defect detection scheme in the industry mainly uses a "camera + lens" pure vision detection architecture. The core principle of this scheme is to capture a two-dimensional image of the surface of the mask plate through a high-resolution camera combined with an optical lens, and then use image processing algorithms (some schemes introduce deep learning models) to identify and determine the defect features in the image. The main advantage of this scheme is that the field of view of the optical system is relatively large, and it can quickly cover the surface of the mask plate, so the detection speed is relatively fast. This scheme is widely used in the detection of obvious defects with a size greater than 0.5 microns, and can meet the detection needs of some medium and low precision mask plates.

[0004] However, as the precision requirements of mask plates in high-end manufacturing fields continue to increase (such as the need to stably detect sub-micron defects in the semiconductor field), the inherent defects of traditional pure vision detection schemes gradually become apparent, making it difficult to adapt to the high-precision and high-stability needs of actual production. The specific limitations are as follows:

[0005] Insufficient detection precision due to limitations of the optical system: The defect recognition capability of the traditional pure vision scheme is highly dependent on the optical imaging performance of the camera and lens, and is limited by the inherent contradiction between "field of view - resolution". If the field of view is expanded to ensure detection speed, the actual size corresponding to each pixel will increase, making it impossible to distinguish sub-micron-level micro defects. If the field of view is reduced to improve resolution, the detection efficiency will be greatly reduced, making it difficult to adapt to the detection rhythm in mass production scenarios. Even if the lens parameters are optimized or the camera hardware is upgraded, the optical system still lacks the ability to capture weak light signals (such as differences in reflected light and scattered light) generated by sub-micron-level defects, making it impossible to break through the detection precision bottleneck.

[0006] Misjudgment and missed judgment problem is prominent: the scheme has very high requirements for recognition algorithm and lighting conditions. In order to distinguish between small defects and background noise, the algorithm needs to have complex feature extraction and classification logic, which not only increases the system development cost and operation load, but also is easy to fail due to environmental light interference, mask surface material differences (such as different plating layers, reflectivity) and other factors; At the same time, for different patterns and different size mask, the light angle, intensity and color temperature need to be repeatedly adjusted to ensure that the defect features are clearly presented, which not only consumes time, but also further reduces the stability and consistency of the detection process. In practical application, misjudgment (judging normal area as defect) and missed judgment (not identifying real defect) still frequently occur.

[0007] Cannot meet the sub-micron detection requirement: the design of the traditional scheme is aimed at defects larger than 0.5 microns, and the optical system resolution and signal processing accuracy are not adapted to the detection of sub-micron defects. When facing the detection requirement of 0.1-0.5 microns sub-micron defects in high-end manufacturing field, the traditional scheme cannot effectively capture the light signal characteristics of defects, and cannot accurately determine the position, size and type of defects, which cannot meet the detection requirements after the industry technology upgrade.

[0008] In summary, the traditional "camera + lens" visual detection scheme has been difficult to adapt to the high-precision and high-stability requirements of mask defect detection in high-end manufacturing fields such as semiconductors and optoelectronics due to the limitations of the optical system, high misjudgment and missed judgment rate, and insufficient precision. The industry urgently needs a new detection system that can break through the above limitations.

[0009] Therefore, a mask defect detection system is provided to overcome the above problems. SUMMARY

[0010] The purpose of the present application is to provide a mask defect detection system to solve the problems raised in the above background art.

[0011] To solve the above technical problems, the mask defect detection system provided by the present application is used to detect micron and sub-micron mask defects, which includes a sample stage for placing the mask to be detected, and further includes an automatic focusing module, a laser scanning module, a signal collection and imaging module, and a signal processing module.

[0012] The automatic focusing module comprises a first light source, a first receiver, a first lens group, a first mechanical shell, a first signal processing unit and a first motor and a driving part thereof, the first lens group comprises a collimating lens, a focusing lens and a first turning mirror, the first signal processing unit comprises a first chip and a corresponding first circuit board, light emitted by the first light source is collimated by the collimating lens, and then reflected by the first turning mirror to a light outlet of the first mechanical shell, light received by a light inlet of the first mechanical shell is focused by the focusing lens, and then transmitted to the first receiver, and the first signal processing unit outputs a control signal to the driving part of the first motor according to a signal input by the first receiver, so as to drive the first motor to work.

[0013] The laser scanning module comprises a second light source, a second lens group, a reflecting mirror, a scanning mirror, a sleeve lens, an objective lens and a second motor and a driving part thereof, the second lens group comprises a light splitter and a dichroic mirror, the second motor drives a rotating shaft to rotate through the driving part thereof, the reflecting mirror is installed on the rotating shaft, light emitted by the second light source is irradiated to the reflecting mirror through the second turning mirror, and light reflected by the reflecting mirror is irradiated on a mask plate on the object table in sequence through the scanning mirror, the sleeve lens, the second lens group and the objective lens.

[0014] The signal collection and imaging module comprises a signal collector and a signal receiver, the signal collector comprises a scattered light collector and a third turning mirror, the signal receiver comprises a photomultiplier tube and a CMOS camera, the third turning mirror reflects light output by the scattered light collector to the photomultiplier tube, and the CMOS camera receives light reflected on the surface of the mask plate.

[0015] The signal processing module comprises a photoelectric conversion circuit and a terminal image processing device, the photoelectric conversion circuit amplifies, filters and denoises a signal output by the photomultiplier tube, the terminal image processing device comprises a signal acquisition card and a software algorithm, the signal acquisition card collects signal data output by the photoelectric conversion circuit and data output by the CMOS camera, and the software algorithm processes the data to determine whether the mask plate has defects, a position, a size and a type of the defects.

[0016] Further, the first light source of the automatic focusing module is a 650nm laser, and the first receiver is a CMOS imaging chip.

[0017] Further, the first mechanical shell of the automatic focusing module is an aluminum black part.

[0018] Further, the second light source of the laser scanning module is a 532nm laser, the reflecting mirror is an 8-face turning mirror, and the scanning mirror is a telecentric lens group.

[0019] Further, the sleeve lens of the laser scanning module has a focal length of 200mm, and the objective lens has a magnification of 50 times.

[0020] Further, the signal collection and imaging module is coaxially installed with the objective lens, and the inner side wall of the scattering light collector is a mirror surface made of aluminum.

[0021] Further, the object table is made of high-rigidity aluminum alloy material and is equipped with high-precision linear guide rails.

[0022] Further, in the automatic focusing module, the first motor drives the objective lens to move up and down until the diameter of the laser spot received on the CMOS imaging chip reaches ±1% of the preset threshold, and the preset threshold is 0.01 mm.

[0023] Further, the software algorithm of the signal processing module includes an edge detection algorithm, a morphological processing algorithm and a machine learning model, and the software algorithm performs Gaussian filter denoising processing on the image collected by the CMOS camera to enhance the contrast of the image.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] 1. Higher sensitivity, significantly reduced algorithm pressure, and greatly improved system stability:

[0026] Compared with the traditional "camera + lens" pure vision detection scheme, the defect recognition sensitivity of the present system is better, and the dependence of the algorithm on the defect judgment can be reduced, avoiding the misjudgment and omission problems commonly seen in pure vision schemes, and improving the long-term operation stability.

[0027] The traditional pure vision scheme needs to compensate for the precision limitations of the optical system through complex algorithms (even combined with deep learning) due to the "wide field of view" characteristic, and is easily affected by light and lens distortion; and the present system uses a combination of "point-by-point laser scanning + CMOS imaging", the laser scanning module (532nm laser, 0.01mm spot) can accurately position the detection area, and the CMOS imaging chip (2048x2048 pixels) can capture high-resolution details, both of which reduce the dependence on "algorithm correction of optical errors", thereby improving the basic precision of defect recognition from the hardware level, and reducing the algorithm operation pressure, and reducing the system instability caused by algorithm overload or misjudgment.

[0028] 2. The detection accuracy breaks through the micron level and can identify sub-micron level small defects:

[0029] It can stably detect micron (μm) defects on the surface of the mask, and even cover sub-micron (<1μm) defects, far exceeding the limitation of the traditional pure vision scheme "difficult to break through the accuracy below 0.5 microns".

[0030] The precision improvement relies on the cooperation of three modules:

[0031] Auto-focusing module: Real-time calculation of defocus amount by 650nm laser and CMOS sensor (10000 frames / s sampling), driving motor to adjust the height of the objective lens, ensuring that the detection area is always in the "best focus state" (spot diameter is stable at 0.01mm±1%), avoiding the loss of details caused by defocus;

[0032] Laser scanning module: 532nm laser is optimized by "8-face rotating mirror (45000RPM high-speed rotation) + sleeve lens (200mm focal length) + 50x objective lens", forming a scanning spot with uniform energy and extremely small diameter, which can accurately cover the micro defect area;

[0033] Signal collection module: Scattered light collector (coaxial installation, aluminum mirror inner wall) cooperates with photomultiplier tube (gain up to 10 6 times) to capture weak scattered light signals generated by micro defects, and then amplify and filter through the signal processing module to finally realize signal extraction of sub-micron defects.

[0034] 3、Modular design flexible adaptation to multiple scenarios, wide application field:

[0035] According to the different size and type of mask, it is flexible to adjust and adapt to semiconductor, optoelectronic, precision optical and other high-end manufacturing fields, solving the problem of "fixed structure and poor adaptability" of traditional solutions.

[0036] The system adopts "auto-focusing, laser scanning, signal collection and imaging, signal processing" four independent module design, and the core parameters (such as laser wavelength, objective lens magnification, scanning speed) of each module can be adjusted individually — for example, when detecting large-size mask, the rotation angle of 8-face rotating mirror can be optimized to expand the scanning range; when detecting mask of special material, different wavelength laser or gain of photomultiplier tube can be replaced. This modular architecture breaks the limitation of traditional pure visual solution "one set of equipment corresponds to one type of mask", realizing flexible application across fields. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The principle diagram of the mask defect detection system of the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0039] Please refer to Figure 1 , the present application provides a technical solution:

[0040] Referring to Figure 1 An embodiment of a reticle defect detection system is shown:

[0041] I. Sample preparation and system startup:

[0042] In the implementation process, first, the 4-inch reticle to be detected (reticle 1) is placed on the stage 2 of the system. The stage 2 is made of high-rigidity aluminum alloy material, equipped with high-precision linear guide rails, and can achieve positioning accuracy of ±0.001 mm. The operator needs to ensure that the surface of the reticle is free of dust, oil stains and other pollutants, so as to avoid affecting the detection results. After completing the sample placement, the operator starts the detection system through the human-machine interface, and the system modules enter the working state in turn.

[0043] II. Working process of the automatic focusing module:

[0044] After starting, the automatic focusing module 3 immediately starts working. The module is built-in with a semiconductor laser 31 with a wavelength of 650 nm. The laser beam emitted by the laser 31 is focused through the lens group 32 and the objective lens 45, forming a laser spot with a diameter of about 0.01 mm, which is irradiated on the surface of the reticle 1. The imaging chip uses a high-sensitivity CMOS sensor 33, which can capture reflected signals in real time at a speed of 10,000 frames per second.

[0045] The imaging chip transmits the feedback signal to the signal processing unit, which analyzes the signal using an algorithm and calculates the defocus amount. The system automatically controls the motor to move the objective lens 45 up and down according to the size of the defocus amount, and adjusts the height of the objective lens in real time until the diameter of the laser spot on the imaging chip reaches ±1% (i.e. 0.01 mm) of the preset threshold. At this time, the system confirms that the best focusing state has been reached, providing a basis for subsequent high-precision detection.

[0046] III. Cooperative work of the laser scanning module:

[0047] After the automatic focusing module completes the focusing, the laser scanning module 4 starts working. The 532 nm laser beam emitted by the laser 41 is adjusted in power to ensure that the output power is stable at 30 mW. The laser beam is first irradiated on one face of the 8-face rotating mirror 42, which rotates at a speed of 45,000 RPM to ensure that the laser beam covers the entire reticle surface within 5 seconds.

[0048] The laser beam passes through the scanning mirror 43, the sleeve lens 44 and the objective lens 45, and finally scans on the reticle surface with a spot diameter of 0.01 mm. The design of this optical system ensures that the energy distribution of the laser beam on the reticle surface is uniform, and the interference and overlap of the spots are minimized.

[0049] IV. Signal acquisition and spectral processing:

[0050] When the 532nm laser beam is incident on the reticle surface, part of the laser is reflected by the sample, and part of the laser is scattered. The reflected light passes through the objective lens 45, the sleeve lens 44 and the beam splitter 46, and enters the CMOS imaging chip 51. The resolution of the CMOS imaging chip 51 is 2048x2048 pixels, which can capture the light intensity information corresponding to each pixel, ensuring high resolution and clear details of the image.

[0051] At the same time, the scattered light passes through the scattered light collector 52 and the hole mirror 47, and is guided to the photomultiplier tube 53. The photomultiplier tube 53 amplifies the scattered light signal by a high multiple, with a gain of up to 10^6, and converts the weak light signal into an electrical signal. The electrical signal then enters the back-end signal processing circuit, which is amplified, filtered and analog-to-digital converted by multiple stages, ensuring the integrity and anti-interference ability of the signal.

[0052] Five, signal processing and defect judgment:

[0053] The processed electrical signal is sent to the data acquisition card 61, and the data acquisition card 61 works synchronously with the CMOS imaging chip 51 to transmit the collected image signal and electrical signal to the host computer. The image processing software built-in the host computer uses multiple algorithms, including edge detection, morphological processing and machine learning model, to deeply analyze the collected signal.

[0054] Image preprocessing: denoising processing is performed on the image collected by the CMOS, and Gaussian filtering algorithm is used to reduce background noise and enhance image contrast.

Claims

1. A reticle defect detection system for detecting micrometer and sub-micrometer scale reticle defects, comprising a stage for placing a reticle to be inspected, characterized in that, The automatic focusing module, the laser scanning module, the signal collecting and imaging module, and the signal processing module are further included. The automatic focusing module includes a first light source, a first receiver, a first lens group, a first mechanical housing, a first signal processing unit, and a first motor and a driving part thereof, the first lens group includes a collimating lens, a focusing lens, and a first turning mirror, the first signal processing unit includes a first chip and a corresponding first circuit board, light emitted by the first light source is collimated by the collimating lens, and then reflected by the first turning mirror to a light outlet of the first mechanical housing, light received by a light inlet of the first mechanical housing is focused by the focusing lens, and then transmitted to the first receiver, and the first signal processing unit outputs a control signal to the driving part of the first motor according to a signal input by the first receiver, so as to drive the first motor to work. The laser scanning module includes a second light source, a second lens group, a reflecting mirror, a scanning mirror, a sleeve lens, an objective lens, and a second motor and a driving part thereof, the second lens group includes a light splitter and a dichroic mirror, the second motor drives the rotating shaft to rotate through the driving part thereof, and the reflecting mirror is installed on the rotating shaft, light emitted by the second light source is irradiated to the reflecting mirror through the second turning mirror, and light reflected by the reflecting mirror is sequentially irradiated to the mask plate on the object table through the scanning mirror, the sleeve lens, the second lens group, and the objective lens. The signal collecting and imaging module includes a signal collector and a signal receiver, the signal collector includes a scattered light collector and a third turning mirror, the signal receiver includes a photomultiplier tube and a CMOS camera, the third turning mirror reflects light output by the scattered light collector to the photomultiplier tube, and the CMOS camera receives light reflected by the surface of the mask plate. The signal processing module includes a photoelectric conversion circuit and a terminal image processing device, the photoelectric conversion circuit amplifies, filters, and denoises a signal output by the photomultiplier tube, the terminal image processing device includes a signal acquisition card and a software algorithm, the signal acquisition card collects signal data output by the photoelectric conversion circuit and data output by the CMOS camera, and the software algorithm processes the data to determine whether the mask plate has defects, a position, a size, and a type of the defects.

2. A reticle defect detection system as claimed in claim 1, wherein: The first light source of the automatic focusing module is a 650 nm laser, and the first receiver is a CMOS imaging chip.

3. A reticle defect detection system as claimed in claim 1 or 2, characterized in that: The first mechanical housing of the automatic focusing module is an aluminum black part.

4. A reticle defect detection system as claimed in claim 1, wherein: The second light source of the laser scanning module is a 532 nm laser, the reflecting mirror is an 8-face turning mirror, and the scanning mirror is a telecentric lens group.

5. A reticle defect detection system as claimed in claim 1 or 4, characterized in that: The sleeve lens of the laser scanning module has a focal length of 200 mm, and the objective lens has a magnification of 50 times.

6. A reticle defect detection system as claimed in claim 1, wherein: The scattered light collector of the signal collecting and imaging module is coaxially installed with the objective lens, and an inner side wall of the scattered light collector is a mirror surface made of aluminum.

7. A reticle defect detection system as claimed in claim 1, wherein: The object table is made of high-rigidity aluminum alloy material and is equipped with a high-precision linear guide rail.

8. A reticle defect detection system as claimed in claim 2, wherein: In the automatic focusing module, the first motor drives the objective lens to move up and down until the diameter of a laser spot received by the CMOS imaging chip reaches ±1% of a preset threshold value, and the preset threshold value is 0.01 mm.

9. A reticle defect detection system as claimed in claim 2, wherein: The software algorithm of the signal processing module includes an edge detection algorithm, a morphological processing algorithm, and a machine learning model, and the software algorithm performs Gaussian filter denoising processing on images collected by the CMOS camera to enhance the contrast of the images.

Images