Patterned wafer defect detection system and method based on double-optical-comb space-time coding

Abstract

The invention discloses a patterned wafer defect detection system and method based on double-optical-comb space-time coding, and the system comprises a double-optical-comb light source module, a space coding illumination module, a time programming control module, a signal collection and processing module, and a data analysis and processing module. And the data analysis processing module receives and analyzes the restored three-dimensional shape information of the surface of the wafer to be detected, extracts defect related characteristics, establishes a mapping relation between spectral characteristics and defects, and identifies the defects on the surface of the wafer to obtain corresponding defect types and position information. According to the invention, a wafer defect detection technology based on double-optical-comb space-time coding is adopted, rapid scanning and parallel detection are realized by using a double-optical-comb interference technology, the detection efficiency is greatly improved, and absolute distance information of the wafer surface can be obtained. The result is fed back to the precise air floating motion platform, so that an automatic focusing optical path system and a feedback control part can be omitted, the detection efficiency is improved, and the size, complexity and cost of the whole wafer detection system are reduced.

Description

Technical Field

[0001] The present invention relates to the field of image technology, and in particular to a patterned wafer defect detection system and method based on dual-comb spatiotemporal coding. Background Art

[0002] As semiconductor processes continue to advance to 3 nanometers and more advanced processes, the structure of chips has become more and more complex and three-dimensional, which has brought higher challenges to defect detection technology. In these advanced processes, the structure on the surface of the wafer not only continues to shrink in lateral dimensions, but also in the vertical direction. The aspect ratio is also increasing, which has led to the traditional two-dimensional plane detection method being unable to meet the increasingly sophisticated process control needs.

[0003] Traditional two-dimensional detection technology faces many challenges, such as the cumbersome, time-consuming and inefficient manual microscope detection process, and the inability of optical plane imaging technology to provide depth information. At present, the industry generally adopts scanning electron microscopes to scan and detect the entire wafer to obtain complete three-dimensional morphological data of the wafer surface. However, this method has the problems of high detection cost and low efficiency, and it is easy to cause electron beam damage to the wafer surface. Therefore, it is mainly suitable for offline defect detection with low efficiency requirements.

[0004] In view of this, in advanced processes, developing a new detection method that can quickly and accurately obtain three-dimensional morphological information of defects, while taking into account detection efficiency and sensitivity, is of vital importance to improving the yield rate of chip manufacturing and reducing production costs. This new method not only needs to be able to accurately locate defects, but also provide detailed three-dimensional morphological data for process optimization and quality control. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a patterned wafer defect detection system and method based on dual-comb spatiotemporal encoding.

[0006] In order to solve the above technical problems, the present invention is solved by the following technical solutions: A patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding, comprising: Dual-comb light source module, providing signal light comb and reference light comb; The spatial coding illumination module includes a spatial coupling unit, a dispersion spectrometer, a beam expansion unit, a coding unit and an illumination receiving unit. The spatial coupling unit receives the signal light comb and converts it into spatial light. The dispersion spectrometer performs multi-beam interference on the spatial light and separates it longitudinally. The spatial light is then expanded transversely and enters the beam expansion unit. The coding unit forms a two-dimensional spatial coding light field. The illumination unit receives the two-dimensional spatial coding light field and illuminates it on the surface of the wafer to form a reflection signal. A time programming control module synchronizes and locks the reflected signal and the reference optical comb to form a time programming optical comb pair, performs time measurement based on the time programming optical comb pair to obtain absolute distance information, controls the relative time delay between the time programming optical comb pairs, and forms a time domain beat frequency signal with different repetition frequencies; The signal acquisition and processing module collects the time-domain beat frequency signal and extracts the spectral intensity information of the RF comb, performs phase extraction and unwrapping calculations on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; The data analysis and processing module receives the restored three-dimensional morphology information of the wafer surface to be tested and identifies the defects on the wafer surface through the defect recognition model to obtain the corresponding defect types and spectral information, and determines the location of the defects on the wafer surface based on the preset mapping relationship.

[0007] As an implementable embodiment, the spatial coupling unit includes a collimator, and the dispersion splitting unit includes a first beam splitter and a second beam splitter; After passing through the circulator and the collimator, the signal light comb is incident on the first beam splitter to form a first beam of light and a second beam of light; The first beam of light passes through the reference mirror to form reflected light as a distance reference. The second beam of light enters the beam expansion unit and passes through the encoding unit to form a two-dimensional spatial encoded light field, which is then irradiated on the surface of the wafer to form a reflected signal. The reflected signal enters the beam expansion unit, the circulator and the second beam splitter in turn. The second beam splitter divides the reflected signal into a first reflected signal and a second reflected signal. The first reflected signal enters the optical time phase detector.

[0008] As an implementable method, the encoding unit includes a cylindrical lens, a virtual imaging phased array, a lens and a blazed grating in sequence, and the second light beam enters the beam expansion unit and then passes through the cylindrical lens, the virtual imaging phased array, the lens and the blazed grating in sequence to form a two-dimensional spatial encoding light field and a reflection signal; The virtual imaging phased array includes a first mirror and a second mirror, wherein the reflectivity of the first mirror is 95%, the reflectivity of the second mirror is 99.9%, the distance between the first mirror and the second mirror is 1.686 mm, and the surface accuracy of the first mirror and the second mirror is respectively ; The number of lines of the blazed grating is 1600, and the diffraction efficiency is greater than 95%; The free spectral range of the virtual imaging phased array is expressed as follows:

[0009] in, The self-expression is represented by the spectral range, represents the refractive index of the material, The wavelength is The incident angle of the light incident on the virtual imaging phased array, represents the wavelength, Indicates time.

[0010] As an implementable embodiment, it further includes a light splitting and beam combining unit, which includes a third beam splitter and a beam combiner; The third beam splitter splits the reference light comb into a first light path and a second light path. The first light path enters the optical time phase detector and is time-phased with the first reflected signal through the optical time phase detector to obtain the absolute distance information of the entire wave packet. The beam combiner receives the second light and the second reflected signal to form a time domain beat signal.

[0011] As an implementable method, the time programming control module includes a digital locking controller, an optical time phase detector and a dual optical comb mutual scanning controller, wherein the digital locking controller locks the key parameters of the time programming optical comb pair to a common time reference, measures the repetition frequency and carrier envelope offset frequency of the time programming optical comb pair, and then locks them; The dual optical comb mutual scanning controller adjusts the phase parameters of the programmable optical comb, so that the reference optical comb is synchronized with the reflected signal and a programmable controllable phase relationship is established, thereby obtaining the relative time delay between the reference optical comb and the reflected signal; The relative time delay is expressed as follows:

[0012] in, represents the relative time delay between the reference optical comb and the signal optical comb, N represents the number of comb teeth, represent the initial phases of the reference optical comb and the signal optical comb respectively, represent the phase of the Nth comb tooth of the reference light comb and the signal light comb respectively, represents the repetition frequency of the optical comb.

[0013] As an implementable method, it also includes a precision motion control module and an automatic loading and unloading module; The precision motion control module controls the position of the wafer to be tested, and the automatic loading and unloading module clamps the wafer to be tested and places the wafer to be tested on the precision air floating platform to realize automatic loading and unloading.

[0014] As an implementable method, the time domain beat signal is collected, and the spectrum intensity information of the radio frequency comb is extracted by Fourier transform; and the phase information is obtained by performing phase unwrapping calculation, including the following steps: Acquire a time-domain beat frequency signal, and perform Fourier transform on the time-domain beat frequency signal to obtain spectral intensity information of a radio frequency spectrum; Perform phase processing and unwrapping operations on the RF spectrum to obtain the phase curve of the RF spectrum. Process the phase curve and the absolute position information of the wave packet to obtain the absolute position information of the two-dimensional spatially encoded light field area, and combine it with the spectral intensity information to obtain the restored three-dimensional morphology of the wafer surface. Based on the restored three-dimensional morphology of the wafer surface, the correspondence between the spectral frequency and the spatial position is calibrated to obtain a mapping relationship; Wherein, the time domain beat frequency signal , which is expressed as follows:

[0015] , denote the electric fields of the signal light comb and the reference light comb respectively, It represents the beat frequency, i.e. the difference in repetition frequency between the reference light and the signal light. Represents phase information, represents the phase difference, Indicates time; The time domain beat frequency signal is Fourier transformed , the RF spectrum is obtained, which is expressed as follows:

[0016] in, represents the frequency domain index, represents the total number of sampling points, represents the imaginary unit, The beat frequency signal in the time domain is represented by sampling points, Indicates Sampling points, each comb tooth corresponds to a different position in the two-dimensional spectral spatial encoding area, and the intensity reflects the reflectivity of the position. The spectrum is segmented by combining the comb tooth interval and the free spectrum region (FSR) of the spectrometer to obtain the intensity image of the illumination area. The RF spectrum is subjected to phase processing and unwrapping operation, which is expressed as follows:

[0017]

[0018] Indicates rounding down. , Indicates the phase after unwrapping; The mapping relationship is expressed as follows:

[0019] in, represents the spatial coordinates, Respectively represent the spatial coordinate position, represents the proportionality coefficient, represents the wavelength of light, Indicates the offset.

[0020] As an implementable method, receiving and analyzing the restored three-dimensional morphology information of the surface of the wafer to be tested to extract defect-related features includes the following steps: Performing differential processing on the restored three-dimensional morphology of the wafer to be tested in the current area and the restored three-dimensional morphology of the wafer in the adjacent area to obtain random defects on the wafer surface, and determining shape information and position information of the random defects through the random defects; By performing differential processing on the restored three-dimensional morphology of the wafer to be tested and the morphology of the standard sample, a systematic defect on the surface of the wafer to be tested is obtained, and shape information and position information of the systematic defect are determined based on the systematic defect; The random defects on the wafer surface are expressed as follows:

[0021] The system defects on the surface of the wafer to be tested are expressed as follows:

[0022] in, Respectively represent the position coordinates of each point on the wafer to be measured, represents the spectral intensity information at the corresponding point, is the adjacent region weight function, Represents the restoration of the three-dimensional shape, The restored 3D topography of the wafer representing the adjacent area, Indicates random defects on the wafer surface. Represents the standard sample morphology, Indicates the system defects on the surface of the wafer to be tested. represents the wavelength-dependent calibration coefficient, Indicates wavelength.

[0023] As an implementable method, the step of identifying wafer surface defects to obtain corresponding defect types includes the following steps: Acquire normal wafer images and spectral data of various defect types, including surface scratches, particle contamination, bridging, and pattern loss; Mark various defect types into categories to form a labeled data set; Constructing a defect recognition pre-training model and training it based on a labeled data set, and adjusting parameters based on a loss function, thereby obtaining a defect recognition model, wherein the defect recognition pre-training model is constructed based on a convolutional neural network capable of extracting local spatial features; Based on transfer learning, the geometric features are integrated with the spectral features. Based on the mapping relationship, the spatial position of the spectral data is mapped to the actual wafer coordinates, and the defect area is marked according to the defect classification results after spectral processing. The loss function is expressed as follows:

[0024] in, represents the number of defect categories, is the regularization coefficient, are the true labels and the predicted results respectively. Indicates defect type data, Indicates the phase offset.

[0025] A patterned wafer defect detection method based on dual-comb spatiotemporal coding is implemented by a patterned wafer defect detection system based on dual-comb spatiotemporal coding. The system includes a dual-comb light source module, a spatial coding illumination module, a time programming control module, a signal acquisition processing module and a data analysis processing module, and includes the following steps: Obtaining a signal light comb and a reference light comb provided by a dual light comb light source module; The spatial coupling unit receives the signal light comb and converts it into spatial light. The dispersion splitting unit performs multi-beam interference on the spatial light and separates it longitudinally. Then, the spatial light is expanded transversely and enters the beam expansion unit. The encoding unit forms a two-dimensional spatially encoded light field. The irradiation unit receives the two-dimensional spatially encoded light field and irradiates it on the wafer surface to form a reflected signal. The reflected signal and the reference optical comb are synchronously locked to form a time-programmed optical comb pair, and the absolute distance information is obtained by time measurement based on the time-programmed optical comb pair, and the relative time delay between the time-programmed optical comb pairs is controlled to form a time-domain beat frequency signal with different repetition frequencies; The signal acquisition and processing module acquires the time-domain beat frequency signal and extracts the spectral intensity information of the radio frequency comb, performs phase extraction and unwrapping calculation on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; The data analysis and processing module receives the restored three-dimensional morphological information of the wafer surface to be tested and identifies the defects on the wafer surface through the defect recognition model to obtain the corresponding defect types and spectral information, and determines the locations of the defects on the wafer surface based on the preset mapping relationship.

[0026] The present invention has significant technical effects due to the adoption of the above technical solutions: As an ultra-precise "optical ruler" in the time and frequency domains, the optical frequency comb has become an ideal light source for precision measurement. It not only has the accuracy that can be traced back to the frequency reference, but also has the ability to multiplex multiple channels, which is the basis for realizing high-throughput parallel optical detection. The optical frequency comb covers a wide spectral range, from ultraviolet to infrared and even terahertz bands, and its broadband characteristics allow for the simultaneous detection of multiple types of material defects. At the same time, the precise time-frequency correspondence of the optical frequency comb can be used to accurately locate the position of the defect and quantify its size or depth, so that a comprehensive evaluation can be performed whether it is surface scratches, contamination or internal structural discontinuities.

[0027] The present invention adopts dual-comb space-time coding for wafer defect detection, and uses dual-comb space-time coding technology to achieve fast scanning and parallel detection, greatly improving detection efficiency; while increasing detection bandwidth, it reduces equipment costs.

[0028] The present invention can obtain the absolute distance information of the wafer surface, and feed back the result to the precision air-floating motion platform, which can save the automatic focusing optical path system, thereby improving the detection efficiency and reducing the volume, complexity and cost of the entire wafer detection system; through the combination of spatiotemporal coding technology and the wide spectrum characteristics of the optical comb, the detection strategy can be flexibly adjusted according to different types of wafers and defect characteristics, and various types of defects, including surface scratches, particle contamination, pattern defects, etc., can be detected, with strong flexibility.

[0029] It also combines machine learning algorithms to perform intelligent analysis of the spectrum, achieving automated defect identification, classification and positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0031] Figure 1 It is a schematic diagram of the overall framework of the system of the present invention; Figure 2 It is a schematic diagram of the overall process of the method of the present invention; Figure 3 Schematic diagram of a patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding of the present invention; Figure 4 It is a schematic diagram of a spatial coding illumination module in a patterned wafer defect detection system based on dual-comb spatiotemporal coding of the present invention; Figure 5 yes Figure 4 Schematic diagram of the generated two-dimensional light spot; Figure 6 It is a schematic diagram of spatially coded illumination spot scanning in the patterned wafer defect detection method based on dual-comb spatiotemporal coding of the present invention; Figure 7 It is a schematic diagram of time programming control in a patterned wafer defect detection system based on dual-optical comb spatiotemporal coding of the present invention; Figure 8 It is a schematic diagram of the dual-comb interference signal decoding process of the data analysis and processing module in the patterned wafer defect detection method based on dual-comb spatiotemporal encoding of the present invention; Fig. 9 It is a schematic diagram of the defect feature signal differential extraction processing flow of the data analysis and processing module in the patterned wafer defect detection method based on dual-optical comb spatiotemporal encoding of the present invention; Fig.10 It is a schematic diagram of the intelligent defect classification and positioning process in the data analysis and processing module of the patterned wafer defect detection method based on dual-comb spatiotemporal encoding of the present invention. DETAILED DESCRIPTION

[0032] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0033] Embodiment 1: A patterned wafer defect detection system based on dual-comb spatiotemporal coding includes a dual-comb light source module, a spatial coding illumination module, a time programming control module, a signal acquisition and processing module, and a data analysis and processing module; The dual-optical comb light source module provides a signal optical comb and a reference optical comb; The spatial coding illumination module includes a spatial coupling unit, a dispersion spectrometer, a beam expansion unit, a coding unit and an irradiation receiving unit. The spatial coupling unit receives the signal light comb and converts it into spatial light. The dispersion spectrometer performs multi-beam interference on the spatial light and separates it longitudinally, and then performs transverse dispersion expansion to enter the beam expansion unit. The coding unit forms a two-dimensional spatial coding light field. The irradiation unit receives the two-dimensional spatial coding light field and irradiates it on the wafer surface to form a reflection signal, wherein the reflection signal has the reflectivity and depth information of the wafer surface. The time programming control module includes a digital locking controller, an optical time phase detector and a dual-optical comb mutual scanning controller. The digital locking controller synchronously locks the reflection signal and the reference optical comb to form a time programming optical comb pair. The optical time phase detector receives the time programming optical comb pair and performs time measurement to obtain absolute distance information. The dual-optical comb mutual scanning controller controls the relative time delay between the time programming optical comb pair to form a time domain beat frequency signal with different repetition frequencies. The signal acquisition and processing module acquires the beat frequency signal in the time domain, extracts the spectral intensity information of the radio frequency comb through Fourier transform, performs phase extraction and unwrapping calculation on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; The data analysis and processing module receives the restored three-dimensional morphological information of the surface of the wafer to be tested and identifies the defects on the wafer surface through a defect recognition model to obtain the corresponding defect types and spectral information. Based on a preset mapping relationship, the location of the defects on the wafer surface is determined, wherein the defect-related features include geometric features, spectral features, and reflection and absorption features.

[0034] In actual operation, the detection system used is as follows Figure 3 As shown, it includes a dual-optical comb mutual scanning controller 1, a reference optical comb 2, a digital locking controller 3, a signal optical comb 4, a third optical fiber splitter 5, an optical time phase detector 6, an optical fiber combiner 7, a second optical fiber splitter 8, an optical fiber circulator 9, a collimator 10, a first optical fiber splitter 11, a reference mirror 12, a beam expansion system 13, a photodetector 14, a two-dimensional spectral space encoding module 15, a programmable gate array 16, an objective lens 17, a computer 18, a precision air-floating motion platform 19, and a robotic arm 20. Here, the optical fiber splitter is the mentioned beam splitter, and the optical fiber combiner is the mentioned combiner.

[0035] The signal light comb 4 provided is the signal light. After receiving the signal light comb 4, it enters the fiber circulator 9 and then outputs as spatial light through the collimator 10. The spatial light passes through the first fiber beam splitter 11 and is divided into two paths of light. One path of light passes through the reference mirror 12 to form reflected light and then returns along the original path. The other path of light passes through the beam expansion system 13 and enters the two-dimensional spectral spatial coding module 15 to form a two-dimensional spatial coded light field. Then, it is focused by the objective lens 17 and irradiated on the wafer to be tested to form a reflected signal. After passing through the fiber circulator 9, the reflected signal enters the second fiber beam splitter 8 and is divided into two paths of light. One path of light (i.e., the reflected signal) enters the optical time phase detector 6, and the other path of light passes through the fiber combiner 7 and enters the photodetector 14. The reference light comb 2 provided is the detection light. After receiving the reference light comb 2, it passes through the third fiber beam splitter 5 and is divided into two paths of light. One path of light enters the optical time phase detector 6, and the other path of light passes through the fiber combiner 7 and enters the photodetector 14.

[0036] The optical time phase detector 6 receives one reflected signal and one light to obtain the time phase detector with attosecond accuracy. , and then obtain the high-precision absolute position information of the wafer , which is expressed as follows:

[0037] in, Indicates time phase detection, Indicates absolute position information. Represents the speed of light, signal light and the probe light The beat frequency signal is formed on the photodetector 14, the optical frequency signal is down-converted to the radio frequency band, and the beat frequency time domain signal is collected by the photodetector 14 After entering the programmable gate array 16, Fourier transform, phase extraction, and phase unwrapping calculation are performed to obtain phase information and combine it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface, which is then input into the computer 18.

[0038] By combining the absolute distance information of the optical time phase detector 6 and the relative position information of each part of the entire coding area obtained by the programmable gate array 16 in the computer 18, the three-dimensional object morphology of the coding area can be quickly and accurately reconstructed to achieve accurate identification and positioning of defects.

[0039] By controlling the precision air-floating motion platform 19 to drive the wafer to move, a fast and high-precision scan of the wafer surface can be achieved. The wafer absolute position information calculated by the computer 18 can be used to determine whether the wafer is located at the focal position of the objective lens 17. By scanning the displacement platform in the Z direction, it can be ensured that the wafer is always located at the focal point of the objective lens during the scanning process.

[0040] In one embodiment, the time-programmable optical comb meets the following requirements: a fiber-mode-locked femtosecond optical comb is selected, with a repetition frequency of 100 MHz, a central wavelength of 780 nanometers after the frequency doubling module, and a spectral range of 15 nanometers. , pulse envelope ,in, and represent the phase offset and time offset of the optical comb pulse respectively, represents the offset phase, Indicates The complex amplitude of the pulse is realized by two phase-locked loops and Any setting of Indicates time, Indicates carrier packet, It represents the reciprocal of the repetition frequency, and i represents the imaginary unit.

[0041] The digital lock controller uses phase-locked loop (PLL) technology to lock the key parameters of the two optical combs to a common reference source (such as a CW laser) and measure the repetition frequency of the two optical combs. and carrier envelope offset frequency , one of the optical combs is used as the master optical comb, and the other as the slave optical comb. The parameters of the slave optical comb are adjusted through the dual-comb mutual scanning controller to synchronize it with the master optical comb and establish a stable phase relationship, thereby achieving coherence between the two optical combs and establishing an accurate frequency relationship, providing a stable frequency reference, eliminating system drift, and finally obtaining a stable dual-comb system with sub-hertz frequency accuracy, thereby performing high-precision spectral measurements.

[0042] The optical time phase detector is a Mach-Zehnder interferometer based on polarization-maintaining fiber. Its function is to accurately identify the time difference between two pulses, thereby obtaining high-precision distance and position information.

[0043] The two-dimensional spectral spatial encoding module consists of optical dispersive elements, such as Figure 4 and Figure 5 As shown, in one embodiment, a cylindrical lens 301, a VIPA 302, a lens 303 and a grating 304 are included. Figure 4 In The wavelength is The incident angle of the light incident on the virtual imaging phased array can use VIPA as the longitudinal dispersion spectrometer. The reflectivity of the two mirrors in the cavity is 95% and 99.9% respectively. The distance between the two mirrors is 1.686 mm. The surface accuracy is: As a preferred method, a blazed grating can be used as a lateral beam splitter, with the number of lines being 1600 and the diffraction efficiency being greater than 95%.

[0044] The dual optical comb mutual scanning controller is used to achieve precise control of the repetition frequency of the two optical combs, thereby achieving Figure 7 Dual-comb spectroscopy acquisition modes are shown for flexible selection of measurements.

[0045] The precision air-floating motion platform 19 uses a marble base to reduce the conduction of low-frequency vibrations. A vibration isolation device is arranged under the base, including but not limited to foundation reinforcement and foundation vibration isolation. The base is supported by an air-floating vibration isolation support and an active vibration isolation platform above it to ensure that the impact of sensitive frequency vibration on the measurement is within a reasonable range during measurement. An air-floating guide rail is provided on the air-floating vibration isolation support. The wafer is driven by the output power of a linear motor and moves along the air-floating guide rail, which greatly reduces the friction of the platform movement and effectively improves the positioning accuracy and movement smoothness. In one embodiment, the precision air-floating motion platform meets the following requirements: repeat positioning accuracy: ±50 nanometers / 300 mm, horizontal straightness: ±0.1μm / 300 mm, vertical straightness: ±0.1μm / 300 mm, positioning jitter: ±5 nanometers.

[0046] The robot arm 20 is located in front of the precision air-floating motion platform and is used to clamp the wafer to be tested and quickly and smoothly place the wafer accurately on the precision air-floating motion platform to achieve automatic loading and unloading. The robot arm 20 can rotate over a large range, move in height over a long distance, and extend in length to meet the requirements of moving the wafer to be tested from the production line material box to the precision air-floating motion platform.

[0047] Embodiment 2: A method for patterned wafer defect detection based on dual-comb spatiotemporal coding is implemented by a patterned wafer defect detection system based on dual-comb spatiotemporal coding, the system comprising a dual-comb light source module, a spatial coding illumination module, a time programming control module, a signal acquisition processing module and a data analysis processing module, such as Figure 2 As shown, the following steps are included: S100, obtaining a signal light comb and a reference light comb provided by a dual light comb light source module; S200, the spatial coupling unit receives the signal light comb and converts it into spatial light, the dispersion splitting unit performs multi-beam interference on the spatial light and separates it longitudinally, and then performs transverse dispersion expansion and enters the beam expansion unit, the encoding unit forms a two-dimensional spatial coded light field, and the irradiation receiving unit receives the two-dimensional spatial coded light field and irradiates it on the wafer surface to form a reflection signal, wherein the reflection signal has the reflectivity and depth information of the wafer surface; S300, the digital locking controller synchronously locks the reflected signal and the reference optical comb to form a time-programmed optical comb pair, the optical time phase detector receives the time-programmed optical comb pair and performs time measurement to obtain absolute distance information, and the dual optical comb mutual scanning controller controls the relative time delay between the time-programmed optical comb pair to form a time-domain beat frequency signal with different repetition frequencies; S400, the signal acquisition and processing module acquires the beat frequency signal in the time domain, extracts the spectral intensity information of the radio frequency comb through Fourier transform, performs phase extraction and unwrapping calculation on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; S500, the data analysis and processing module receives the restored three-dimensional morphological information of the surface of the wafer to be tested and identifies the surface defects of the wafer through a defect recognition model to obtain the corresponding defect types and spectral information, and determines the positions of the surface defects of the wafer based on a preset mapping relationship, wherein the defect-related features include geometric features, spectral features, and reflection and absorption features.

[0048] In actual operation, this can be achieved through the following steps: Step 1: Place the wafer to be tested on the precision air-floating platform based on the automatic loading module. The specific steps are: take out the wafer to be tested from the wafer box through the robotic arm, and after grasping, rotating, lifting, and retracting, quickly and steadily place it on the precision air-floating motion platform measurement table.

[0049] Step 2: Through the linkage of the spatial coding lighting module and the precision air-floating motion platform, the wafer to be tested is scanned and tested in different areas, including: 1) First, ensure that the inspection equipment has been correctly installed and calibrated, and adjust the position of the wafer to be tested so that it is within the two-dimensional spectral spatial encoding illumination range.

[0050] 2) The signal light comb 4 provided is the signal light. After receiving the signal light comb 4, it enters the optical fiber circulator 9 and then outputs as spatial light through the collimator 10. The spatial light is divided into two paths of light through the first optical fiber beam splitter 11. One path of light passes through the reference mirror 12 to form reflected light and then returns along the original path. The other path of light passes through the beam expansion system 13 and enters the two-dimensional spectral spatial coding module 15 to form a two-dimensional spatial coded light field, which can be understood as the attached Figure 5 The two-dimensional light spot shown in the figure is then focused by the objective lens 17 and irradiated onto the wafer to be tested to form a reflection signal.

[0051] 3) After the reflected light returns along the original path, it is received by the optical fiber circulator 9 and then split into two paths by the second optical fiber beam splitter 8, one path enters the optical time phase detector 6, and the other path enters the optical fiber combiner 7.

[0052] 4) After the output light of the reference light comb 2 passes through the optical fiber beam splitter 1, one path enters the optical time phase detector 6 and performs time phase detection with the return light of the signal light comb 4, thereby obtaining high-precision absolute distance information of the entire wave packet. The absolute distance information is fed back to the precision air-floating motion platform 19 through the computer 18 for vertical automatic focusing. Another path passes through the optical fiber combiner 7 and the other return signal of the signal light comb 4 to form a time domain beat frequency signal on the photodetector 14.

[0053] Step 3: Through the signal acquisition and processing module, collect the dual-comb time-domain interference signal and extract the RF spectrum intensity and phase information, such as Figure 8 As shown, specifically: 31) The time domain beat signal obtained in step 2 is stored and processed by the programmable gate array 16.

[0054] 32) After the dual-comb time-domain interference signal is Fourier transformed (FFT), the RF spectrum is obtained, in which each comb tooth corresponds to a different position in the two-dimensional spectral spatial encoding area, and its intensity reflects the reflectivity of the position. The spectrum is segmented by combining the comb tooth interval and the free spectrum range FSR of the spectrometer to obtain the intensity image of the illumination area and the spectral intensity information. The number of horizontal and vertical pixels of the image is determined by the free spectrum range of the spectrometer. As a preferred method, VIPA is selected as the spectrometer element, and its free spectrum range is , the repetition frequency of the optical comb ,bandwidth , horizontal pixel number ,Vertical pixels Total number of pixels , that is, a single field of view consists of 90,000 parallel imaging focuses, greatly improving the detection throughput.

[0055] 33) The phase curve of the RF spectrum is obtained by performing phase extraction and unwrapping operations on the RF spectrum, which reflects the relative position information of different points in the two-dimensional spectral spatial encoding area, that is, the depth information of different points, combined with the absolute position information of the wave packet obtained in step 202 , the absolute position information of the entire two-dimensional spectral spatial encoding area can be obtained , combined with the spectral intensity information obtained in the previous step , the three-dimensional morphology of the wafer surface can be restored .

[0056] 34) Use a single-frequency laser and a standard resolution plate (USAF 1951) to calibrate the correspondence between spectral frequency and spatial position.

[0057] Step 4: Differ the test results from the adjacent areas and the standard sample results to extract random defect and system defect information. The specific steps are as follows: 41) By combining the three-dimensional morphology test results of the wafer to be tested With adjacent areas Make a difference and extract random defects on the wafer surface shape and position information, ; 42) By combining the three-dimensional morphology test results of the wafer to be tested With standard sample Make a difference and extract the systematic defects on the wafer surface shape and position information, .

[0058] Step 5: By analyzing the spectral data, the corresponding relationship between the geometric features (such as depth, size, shape, etc.) and spectral features (such as waveform, abnormal peaks, etc.) related to the defect is extracted. The results are as follows: Fig. 9 As shown, the specific steps include: 51) Collect spectral data of normal wafer samples and various defects including surface scratches, particle contamination, bridging, pattern missing, etc.; 52) Label different defect types to ensure balanced data sets; 53) Divide the dataset into training set (70%), validation set (20%) and test set (10%) to avoid data leakage.

[0059] Step 6: Based on machine learning, deep learning and other methods, by learning a large number of labeled samples, a mapping relationship between spectral features and defects is established to achieve rapid and automatic identification, classification and positioning of defects, such as Fig.10 As shown, the specific steps include: 61) Select a convolutional neural network (CNN) suitable for extracting local spatial features for construction and training; 62) Use the test set to evaluate the model’s accuracy, precision, and confusion matrix; 63) Use transfer learning to pre-train ImageNet and then fine-tune to improve classification accuracy. Integrate geometric features with spectral features to improve learning ability; 64) Map the spatial position of the spectral data to the actual wafer coordinates. Mark the defect area according to the defect classification results after spectral processing.

[0060] Based on machine learning, deep learning and other methods, combined with the constructed mapping relationship, and then combined with the data set collected by this solution, it is not only possible to identify and classify various types of defects, including surface scratches, particle contamination, pattern defects, etc., but also to obtain specific location information on the wafer to be tested based on the mapping relationship. In other words, this application is able to obtain accurate defect classification and location information, providing accurate data for some subsequent work.

[0061] 65) After completing the above steps, follow the preset scanning direction as follows Figure 6 As shown, the wafer is moved to the next scanning area. The advantage of this scanning method is that it can achieve dense scanning of the wafer surface. Repeat steps 2 to 6 until the corresponding wafer detection is completed.

[0062] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also belong to the scope of the present invention.

[0063] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0064] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0066] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0068] It should be noted that: The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0069] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A patterned wafer defect detection system based on dual-comb spatiotemporal encoding, characterized in that: include: Dual-comb light source module, providing signal light comb and reference light comb; The spatial coding illumination module includes a spatial coupling unit, a dispersion spectrometer, a beam expansion unit, a coding unit and an illumination receiving unit. The spatial coupling unit receives the signal light comb and converts it into spatial light. The dispersion spectrometer performs multi-beam interference on the spatial light and separates it longitudinally and then performs transverse dispersion expansion before entering the beam expansion unit. The coding unit forms a two-dimensional spatial coding light field. The illumination receiving unit receives the two-dimensional spatial coding light field and irradiates it on the wafer surface to form a reflection signal. A time programming control module receives the reflected signal and synchronizes the reflected signal with a reference optical comb to form a time programming optical comb pair, performs time measurement based on the time programming optical comb pair to obtain absolute distance information, controls the relative time delay between the time programming optical comb pairs, and forms a time domain beat frequency signal with different repetition frequencies; The signal acquisition and processing module collects the time-domain beat frequency signal and extracts the spectral intensity information of the RF comb, performs phase extraction and unwrapping calculations on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; The data analysis and processing module receives the restored three-dimensional morphology information of the wafer surface to be tested and identifies the defects on the wafer surface through the defect recognition model to obtain the corresponding defect types and spectral information, and determines the location of the defects on the wafer surface based on the preset mapping relationship.

2. The patterned wafer defect detection system based on dual-comb spatiotemporal encoding according to claim 1, characterized in that: The spatial coupling unit includes a collimator, and the dispersion splitting unit includes a first beam splitter and a second beam splitter; After passing through the circulator and the collimator, the signal light comb is incident on the first beam splitter to form a first beam of light and a second beam of light; The first beam of light passes through the reference mirror to form reflected light as a distance reference. The second beam of light enters the beam expansion unit and passes through the encoding unit to form a two-dimensional spatial encoded light field, which is then irradiated on the surface of the wafer to form a reflected signal. The reflected signal enters the beam expansion unit, the circulator and the second beam splitter in turn. The second beam splitter divides the reflected signal into a first reflected signal and a second reflected signal. The first reflected signal enters the optical time phase detector.

3. The patterned wafer defect detection system based on dual-comb spatiotemporal encoding according to claim 2, characterized in that: The encoding unit includes a cylindrical lens, a virtual imaging phased array, a lens and a blazed grating in sequence. The second light beam enters the beam expansion unit and then passes through the cylindrical lens, the virtual imaging phased array, the lens and the blazed grating in sequence to form a two-dimensional spatial encoding light field and a reflection signal. The virtual imaging phased array includes a first mirror and a second mirror, wherein the reflectivity of the first mirror is 95%, the reflectivity of the second mirror is 99.9%, the distance between the first mirror and the second mirror is 1.686 mm, and the surface accuracy of the first mirror and the second mirror is respectively ; The number of lines of the blazed grating is 1600, and the diffraction efficiency is greater than 95%; The free spectral range of the virtual imaging phased array is expressed as follows: in, The self-expression is represented by the spectral range, represents the refractive index of the material, The wavelength is The incident angle of the light incident on the virtual imaging phased array, represents the wavelength, Indicates time.

4. The patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding according to claim 2, characterized in that: It also includes a light splitting and beam combining unit, which includes a third beam splitter and a beam combiner; The third beam splitter splits the reference light comb into a first light path and a second light path. The first light path enters the optical time phase detector and is time-phased with the first reflected signal through the optical time phase detector to obtain the absolute distance information of the entire wave packet. The beam combiner receives the second light and the second reflected signal to form a time domain beat signal.

5. The patterned wafer defect detection system based on dual-comb spatiotemporal encoding according to claim 1, characterized in that: The time programming control module includes a digital locking controller, an optical time phase detector and a dual optical comb mutual scanning controller. The digital locking controller locks the key parameters of the time programming optical comb pair to a common time reference, measures the repetition frequency and carrier envelope offset frequency of the time programming optical comb pair, and then locks them; The dual optical comb mutual scanning controller adjusts the phase parameters of the programmable optical comb, so that the reference optical comb is synchronized with the reflected signal and a programmable controllable phase relationship is established, thereby obtaining the relative time delay between the reference optical comb and the reflected signal; The relative time delay is expressed as follows: in, represents the relative time delay between the reference optical comb and the signal optical comb, N represents the number of comb teeth, represent the initial phases of the reference optical comb and the signal optical comb respectively, represent the phase of the Nth comb tooth of the reference light comb and the signal light comb respectively, represents the repetition frequency of the optical comb.

6. The patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding according to claim 1, characterized in that: It also includes precision motion control module and automatic loading and unloading module; The precision motion control module controls the position of the wafer to be tested, and the automatic loading and unloading module clamps the wafer to be tested and places the wafer to be tested on the precision air floating platform to realize automatic loading and unloading.

7. The patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding according to claim 1, characterized in that: The method of collecting the time domain beat frequency signal and extracting the spectral intensity information of the radio frequency comb; and obtaining the phase information by performing phase unwrapping calculation comprises the following steps: Acquire a time-domain beat frequency signal, and perform Fourier transform on the time-domain beat frequency signal to obtain spectral intensity information of a radio frequency spectrum; Perform phase processing and unwrapping operations on the RF spectrum to obtain the phase curve of the RF spectrum. Process the phase curve and the absolute position information of the wave packet to obtain the absolute position information of the two-dimensional spatially encoded light field area, and combine it with the spectral intensity information to obtain the restored three-dimensional morphology of the wafer surface. Based on the restored three-dimensional morphology of the wafer surface, the correspondence between the spectral frequency and the spatial position is calibrated to obtain a mapping relationship; Wherein, the time domain beat frequency signal , which is expressed as follows: , denote the electric fields of the signal light comb and the reference light comb respectively, It represents the beat frequency, i.e. the difference in repetition frequency between the reference light and the signal light. Represents phase information, represents the phase difference, represents a proportional relationship and ignores constant factors, Indicates time; The time domain beat frequency signal is Fourier transformed , the RF spectrum is obtained, which is expressed as follows: in, represents the frequency domain index, represents the total number of sampling points, represents the imaginary unit, The beat frequency signal in the time domain is represented by sampling points, Indicates Sampling points, each comb tooth corresponds to a different position in the two-dimensional spectral spatial encoding area, and the intensity reflects the reflectivity of the position. The spectrum is segmented by combining the comb tooth interval and the free spectrum region (FSR) of the spectrometer to obtain the intensity image of the illumination area. The RF spectrum is subjected to phase processing and unwrapping operation, which is expressed as follows: Indicates rounding down. , Indicates the phase after unwrapping; The mapping relationship is expressed as follows: in, represents the spatial coordinates, Respectively represent the spatial coordinate position, represents the proportionality coefficient, represents the wavelength of light, Indicates the offset.

8. The patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding according to claim 1, characterized in that: The method of receiving and analyzing the restored three-dimensional morphology information of the surface of the wafer to be tested and extracting defect-related features includes the following steps: Performing differential processing on the restored three-dimensional morphology of the wafer to be tested in the current area and the restored three-dimensional morphology of the wafer in the adjacent area to obtain random defects on the wafer surface, and determining shape information and position information of the random defects through the random defects; By performing differential processing on the restored three-dimensional morphology of the wafer to be tested and the morphology of the standard sample, the surface system defects of the wafer to be tested are obtained; The random defects on the surface of the wafer to be tested are expressed as follows: The system defects on the surface of the wafer to be tested are expressed as follows: in, Respectively represent the position coordinates of each point on the wafer to be measured, represents the spectral intensity information at the corresponding point, is the adjacent region weight function, Represents the restoration of the three-dimensional shape, The restored 3D topography of the wafer representing the adjacent area, Indicates random defects on the wafer surface. Represents the standard sample morphology, Indicates the system defects on the surface of the wafer to be tested. represents the wavelength-dependent calibration coefficient, Indicates wavelength.

9. The patterned wafer defect detection system based on dual-optical comb spatiotemporal encoding according to claim 1, characterized in that: The method of identifying the surface defects of the wafer to obtain the corresponding defect types includes the following steps: Acquire normal wafer images and spectral data of various defect types, including surface scratches, particle contamination, bridging, and pattern loss; Mark various defect types into categories to form a labeled data set; Constructing a defect recognition pre-training model and training it based on a labeled data set, and adjusting parameters based on a loss function, thereby obtaining a defect recognition model, wherein the defect recognition pre-training model is constructed based on a convolutional neural network capable of extracting local spatial features; Based on transfer learning, the geometric features are integrated with the spectral features. Based on the mapping relationship, the spatial position of the spectral data is mapped to the actual wafer coordinates, and the defect area is marked according to the defect classification results after spectral processing. The loss function is expressed as follows: in, represents the number of defect categories, is the regularization coefficient, Represent the true label and the predicted result respectively. Indicates defect type data, Indicates the phase offset.

10. A method for patterned wafer defect detection based on dual-optical comb spatiotemporal coding implemented by the system according to any one of claims 1 to 9, comprising the following steps: Obtaining a signal light comb and a reference light comb provided by a dual light comb light source module; The spatial coupling unit receives the signal light comb and converts it into spatial light, which then enters the beam expansion unit. The dispersion splitting unit performs multi-beam interference on the spatial light and separates it longitudinally, and then performs transverse dispersion expansion. The coding unit forms a two-dimensional spatial coded light field. The irradiation receiving unit receives the two-dimensional spatial coded light field and irradiates it on the wafer surface to form a reflected signal. The reflected signal and the reference optical comb are synchronously locked to form a time-programmed optical comb pair, and the absolute distance information is obtained by time measurement based on the time-programmed optical comb pair, and the relative time delay between the time-programmed optical comb pairs is controlled to form a time-domain beat frequency signal with different repetition frequencies; The signal acquisition and processing module acquires the time-domain beat frequency signal and extracts the spectral intensity information of the radio frequency comb, performs phase extraction and unwrapping calculation on the spectral intensity information, obtains the phase information and combines it with the spectral intensity information to obtain the restored three-dimensional morphology information of the wafer surface; The data analysis and processing module receives the restored three-dimensional morphological information of the wafer surface to be tested and identifies the defects on the wafer surface through the defect recognition model to obtain the corresponding defect types and spectral information, and determines the locations of the defects on the wafer surface based on the preset mapping relationship.

Images