Dual-comb based pattern wafer multi-parameter synchronous detection system and method

Abstract

The application discloses a kind of based on double optical comb's pattern wafer multi-parameter synchronous detection system and method, belong to wafer quality detection field.The system includes signal light path, reference light path, detection light path and a synchronous detection module.After interference, probe light beam and reference light beam are decomposed into H, V two ways by polarizing beam splitter, by first camera and second camera synchronous detection.The application also provides a kind of method: synchronous acquisition two-way time-domain interference data and;For each pixel point, Fourier transform is executed, and complex spectrum pair is obtained;Processing unit is based on the spectrum pair parallel decoding, and the three-dimensional topography of pattern wafer, stress, film parameter and so on multi-dimensional physical parameter are synchronously inversed;Finally, contrast identification defect.The application solves the problem of low detection efficiency and single information dimension in the prior art, realizes wide field, high speed, high precision and multi-parameter synchronous comprehensive characterization.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wafer quality detection, and particularly relates to a pattern wafer multi-parameter synchronous detection system and method based on a double optical comb. BACKGROUND

[0002] In the semiconductor manufacturing process, accurate and rapid detection of defects of a pattern wafer is a key link to ensure the yield of a chip. With the continuous reduction of the feature size of the chip and the increasing complexity of the three-dimensional structure (such as FinFET and GAA), the sources and forms of defects are also various. In addition to the traditional particle contamination and pattern error, residual stress, slight film thickness unevenness and slight deviation of material composition introduced by film deposition, chemical mechanical polishing (CMP) and other processes can all become fatal hidden defects.

[0003] The industry has developed various microscopic detection technologies, but each has obvious limitations:

[0004] 1. Scanning probe microscopy (such as AFM): can provide nanoscale surface three-dimensional topography information, but its imaging depends on the point-by-point scanning of a mechanical probe, which is extremely slow and cannot meet the detection needs of large area and high throughput.

[0005] 2. Electron microscopy (such as SEM): has extremely high spatial resolution, but usually needs to be carried out in a vacuum environment, which may cause damage to the sample, and it is difficult to obtain stress, film thickness and other non-topography physical information of the sample.

[0006] 3. Traditional optical microscopy: although it realizes fast and non-contact wide-field imaging, it mainly relies on light intensity information and is not sensitive to defects caused by slight phase changes (such as film thickness unevenness), material anisotropy (such as residual stress) or subsurface structures.

[0007] 4. Advanced optical metrology technology:

[0008] 1) Scanning interferometry / confocal microscopy: can measure the phase information through interference to obtain high-precision surface profile, but is still limited by the efficiency bottleneck of point-by-point scanning.

[0009] 2) Spectral ellipsometry: is the "gold standard" for measuring film parameters (thickness, refractive index, etc.), but is usually a single-point measurement or limited-area imaging, and it is difficult to realize high-resolution wide-field imaging.

[0010] 3) Dual-comb microscopy: In recent years, a scanning microscope based on dual-comb (S-DCSM) has been proposed, which verifies the feasibility of synchronously acquiring multiple information such as amplitude, phase, polarization and time of flight from a single measurement. It realizes large-range and high-precision depth measurement by combining the time-of-flight information of the interference envelope with the carrier phase information. However, the existing dual-comb microscopy mainly adopts a confocal scanning mode, and its imaging speed and field of view range are still limited by the performance of the scanning device, and it has not yet realized the true sense of high-speed, wide-field, multi-modal synchronous imaging.

[0011] In summary, the prior art is either insufficient in "detection efficiency" or single in "information dimension", or there is an inherent contradiction between "measurement range and precision", and there is a lack of an ideal solution that can simultaneously realize wide-field, high-speed, high-precision and multi-dimensional information acquisition. The detection scheme can quickly acquire multiple physical properties of the wafer (such as topography, thickness, stress, material, etc.) at one time, so as to realize comprehensive and efficient characterization of complex defects. SUMMARY

[0012] The present application provides a pattern wafer multi-parameter synchronous detection system and method based on dual-comb to realize fast, comprehensive and high-precision comprehensive characterization of wafer defects, aiming at the technical defects of low detection efficiency, single information dimension and difficulty in balancing measurement range and precision in the prior art.

[0013] The present application first provides a pattern wafer multi-parameter synchronous detection system based on dual-comb, comprising:

[0014] A signal light path comprising a first optical comb, a first polarizer and a first beam expander arranged along the light path, the signal light path being used to generate a probe light beam;

[0015] A reference light path comprising a second optical comb, an acousto-optic frequency shift module, a second polarizer and a second beam expander arranged along the light path, the acousto-optic frequency shift module being used to perform optical frequency shift on the light beam, the reference light path being used to generate a reference light beam;

[0016] A beam splitting unit comprising a main beam splitter and a sampling beam splitter;

[0017] A detection light path comprising an imaging lens, an objective lens and a displacement stage arranged along the light path, the probe light beam is reflected by the main beam splitter and irradiates the pattern wafer placed on the displacement stage through the detection light path and generates reflected return light, which reaches the sampling beam splitter through the objective lens, the lens and the main beam splitter, and interferes with the reference light beam reaching the sampling beam splitter to form an interference light beam;

[0018] The synchronous detection module includes a polarization beam splitter, a first camera, and a second camera. The interference beam is reflected by the sampling beam splitter and then incident on the polarization beam splitter, where it is decomposed into two orthogonal polarization components: a horizontal polarization component H and a vertical polarization component V, which are detected by the first camera and the second camera, respectively.

[0019] A dual-comb light source consisting of a first optical comb and a second optical comb, wherein the first optical comb (signal optical comb) and the second optical comb (reference optical comb) contain two mutually phase-locked optical combs with slightly different repetition frequencies. The mode-locked lasers are used as probe beams (frequency...). ) and reference light (frequency) = + This frequency difference is the basis for achieving fast equivalent time delay scanning without mechanical scanning.

[0020] Before illuminating the sample (patterned wafer), the probe light passes through a first polarizer to precisely control its polarization state to a preset state, preferably set to 45° linear polarization relative to the subsequent detection coordinate system. This setting ensures that the light subsequently detected by the two cameras, i.e., the H and V orthogonally polarized components, have equal energy, ensuring that both cameras can obtain sufficiently strong signals, laying the foundation for subsequent high-precision amplitude ratio and phase difference calculations.

[0021] After polarization initialization, the probe light is expanded by the first beam expander to form a large-aperture, uniform parallel beam. This beam is then used by the objective lens system to cover and illuminate a region of interest (ROI) on the patterned wafer. This enables synchronous and parallel measurement of the entire field of view, which is the key to this invention's high-throughput detection, distinguishing it from confocal scanning technology.

[0022] The reflected or transmitted light from the patterned wafer sample carries the comprehensive physical information of each pixel within the field of view. This reflected light is collected and guided to a polarization analysis module, the core component of which is a polarization beam splitter (PBS). The PBS spatially decomposes the vector optical field of the reflected light into two independent, mutually orthogonal linear polarization components: the horizontal polarization component (H component) and the vertical polarization component (V component).

[0023] Preferably, the acousto-optic frequency shifting module includes a first acousto-optic modulator and a second acousto-optic modulator arranged in series along the reference optical path. The first acousto-optic modulator and the second acousto-optic modulator are driven by the same arbitrary waveform generator. By controlling the driving frequency difference applied to the first acousto-optic modulator and the second acousto-optic modulator, an optical frequency shift is applied to the light beam.

[0024] More preferably, the patterned wafer multi-parameter synchronous detection system based on dual optical combs further includes a synchronous controller / frequency division circuit and an acquisition and processing module, which includes an image acquisition card and a processing unit;

[0025] The high-frequency synchronization signals from the first and second optical combs are split into two paths: one path is used to synchronize the arbitrary waveform generator, and the other path is processed by the synchronization controller / frequency divider circuit to generate a low-frequency trigger signal to synchronize the operation of the first camera, the second camera, and the image acquisition card.

[0026] The image sequences acquired by the first and second cameras are transmitted to the processing unit via the image acquisition card. The processing unit then processes the data to generate detection results for patterned wafer defects.

[0027] More preferably, the patterned wafer multi-parameter synchronous detection system based on dual optical combs further includes a beat frequency monitoring module, which is used to receive the interference beam transmitted from the sampling beam splitter and to use real-time monitoring of the beat frequency signal to provide feedback adjustment for the acousto-optic frequency shifting module.

[0028] More preferably, a pulse width broadening module is provided upstream of the synchronous detection module, and the interference beam is broadened in the time domain by the pulse width broadening module before being incident on the polarization beam splitter.

[0029] More preferably, a fixed reflector is provided on one side of the main beam splitter. After the probe beam passes through the main beam splitter, it is reflected by the fixed reflector and then passes through the main beam splitter again before entering the sampling beam splitter as an internal reference beam. The internal reference beam is used to provide an absolute position reference for the sample. The reference mirror has a peak in the time domain signal, and the position of the peak and the sample peak can characterize the absolute distance between them, serving as a stable reference.

[0030] This invention also provides a method for simultaneous multi-parameter detection of patterned wafers based on a dual-optical comb. The method, using the dual-optical comb-based system, includes the following steps:

[0031] S1, a probe beam is generated by the signal optical path, and a reference beam is generated by the reference optical path. The probe beam is reflected by the main beam splitter and then shines on the patterned wafer placed on the displacement stage through the detection optical path, generating reflected return light. After passing through the objective lens, lens, and main beam splitter, it reaches the sampling beam splitter and interferes with the reference beam to form an interference beam. The interference beam is reflected by the sampling beam splitter and then incident on the polarization beam splitter, where it is decomposed into two orthogonal polarization components, H and V, which are detected and recorded simultaneously by the first camera and the second camera, respectively. and Two time-domain interferometric data, of which tFor time series values, x and y The value of a pixel;

[0032] S2, after receiving the two-channel temporal interferometric data, the processing unit processes each pixel. Time series data (Time series data of H-path) and (V-channel time series data) undergoes a fast Fourier transform to convert it from the time domain to the frequency domain, yielding the core complex spectral pairs;

[0033] S3, the processing unit decodes multiple computational paths in parallel based on the complex spectrum obtained in step S2, so as to synchronously retrieve the multi-dimensional physical parameters of the patterned wafer.

[0034] S4, the processing unit compares and analyzes the various physical parameter maps calculated in step S3 with the preset process control standards or design library, and identifies and locates the non-compliant areas as defects.

[0035] Preferably, the multi-dimensional physical parameters of the patterned wafer in step S3 include three-dimensional surface morphology, internal defect location, surface stress distribution, film thickness, refractive index, extinction coefficient, and broadband reflectance spectrum of the material.

[0036] More preferably, in steps S2 and S3, after the processing unit receives the two channels of time-domain interferometric data, it starts two main processing paths in parallel, namely the time-domain direct analysis path and the frequency-domain transform analysis path.

[0037] In the direct time-domain analysis path, the processing unit performs peak finding and localization of the interference envelope of the time-domain signal, determines the peak time position of the interference envelope generated by reflections from the surface and internal interfaces of the patterned wafer, and then calculates the 2π-unambiguous time delay between each peak time position. And obtain rough depth information. Optical time delay = / M, where M is the compression ratio / Based on optical time delay This allows for the calculation of the three-dimensional surface morphology of the patterned wafer and the three-dimensional spatial location of internal defects; if the sample is a semi-transparent multilayer structure, multiple interference envelopes will appear in the time-domain signal, according to = The optical thickness can then be calculated. This allows us to characterize the layered structure of the sample;

[0038] In the frequency domain transform analysis path, the processing unit first performs a Fourier transform on the time domain signal to obtain complex spectral data pairs containing complete amplitude and phase information. Based on the obtained complex spectral data pairs, the processing unit performs parallel, multi-path decoding calculations to simultaneously extract multi-dimensional physical quantity parameters.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] 1. Multi-parameter synchronization and information fusion: This invention breaks through the limitation of traditional instruments being "one machine, one function." By deeply decoding the vector light field, it can obtain multi-dimensional physical properties of the sample, such as morphology, material, spectrum, and mechanics, in a single measurement, constructing a "physical property data cube." This allows the analysis of defects to move from "observing phenomena" to "exploring the physical root cause," greatly improving the diagnostic capability for complex defects.

[0041] 2. High-efficiency wide-field imaging: This invention employs a "snapshot" wide-field imaging mode, rather than point-by-point scanning, fundamentally overcoming the speed bottleneck of traditional high-precision microscopy techniques. Its detection throughput (area detected per unit time) is expected to be several orders of magnitude higher than scanning methods, enabling high-throughput screening and online process control.

[0042] 3. Combining large range and high precision: This invention innovatively adopts a "coarse-to-fine" strategy to calculate three-dimensional topography. It utilizes the large range and unambiguous characteristics of the time-of-flight method in the time domain to solve the 2π periodic ambiguity of the phase method, while using the high sensitivity of the frequency domain phase method to ensure nanometer-level measurement accuracy, perfectly solving the bottleneck problem of mutual constraint between range and accuracy in traditional interferometry.

[0043] 4. Non-destructive and non-contact measurement: The entire measurement process is based entirely on optical principles and has no physical contact with the sample, ensuring completely non-destructive testing of precision and fragile samples such as semiconductor wafers.

[0044] 5. System Simplification and High Stability: This invention integrates the core functions of multiple instruments, such as traditional microscopes, profilometers, ellipsometers, and spectrometers, into a single system. Its core depth and spectral scanning functions are achieved through dual-comb technology in a non-mechanical manner, thus avoiding the vibration, wear, and instability problems associated with mechanical scanning, resulting in higher overall system stability and robustness. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0046] Figure 2 This is a flowchart of a method according to an embodiment of the present invention.

[0047] Figure 3This is a block diagram illustrating the principle of simultaneously extracting multiple physical quantity parameters from complex spectra according to an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0049] I. System Structure

[0050] To better understand this invention, the following will be combined with Figure 1 The system structure of a preferred embodiment of the present invention will be described in detail below.

[0051] Figure 1 A schematic diagram of a patterned wafer multi-parameter synchronous detection system based on a dual-optical comb, according to an embodiment of the present invention, is shown. The system can be functionally divided into: a dual-optical comb light source module, an illumination and polarization control module, an acousto-optic frequency shifting module, a microscopic interference optical path, a beat frequency monitoring module, a pulse width broadening module, a synchronous detection module, and an acquisition and processing module.

[0052] 1. Preparation of light source and optical path

[0053] The dual-comb light source module includes a first optical comb 101 and a second optical comb 102, used to generate two beams with a small difference in repetition frequency. The system uses two coherent pulsed lasers, one as a probe beam and the other as a reference beam (local beam). The two independent mode-locked lasers are precisely controlled and locked in frequency and phase via an electronic feedback system. In this embodiment, the first optical comb 101 serves as the signal optical comb (red optical path in the figure), and its beam is used to probe the sample; the second optical comb 102 serves as the reference optical comb (blue optical path in the figure), and its beam is used for interferometric detection.

[0054] After exiting the first optical comb 101, the signal beam enters the illumination and polarization control module. This module is responsible for processing the probe light. In this embodiment, the probe light emitted from the first optical comb 101 first passes through the first polarizer 201 in this module, where its polarization state is precisely set, for example, to be linearly polarized at 45° relative to the subsequent detection coordinate system. Subsequently, the beam passes through the first beam expander 203 to form a uniform, wide-aperture parallel beam, preparing for subsequent wide-field illumination.

[0055] After the reference beam exits from the second optical comb 102, it directly enters the acousto-optic frequency shifting module.

[0056] 2. Acousto-optic frequency shifting and reference optical path

[0057] The acousto-optic frequency shift module is used to precisely shift the optical frequency of the reference light to achieve optical downsampling. The reference beam passes sequentially through a first acousto-optic modulator 301 and a second acousto-optic modulator 302 connected in series. These two acousto-optic modulators (AOMs) are driven by the same arbitrary waveform generator 303 (AWG). By controlling the difference in driving frequencies applied to the first acousto-optic modulator 301 and the second acousto-optic modulator 302, a precise and controllable optical frequency shift can be applied to the reference light. .

[0058] The frequency-shifted reference beam also enters the illumination and polarization control module, where it is processed by the second polarizer 202 and the second beam expander 204 to ensure that its beam characteristics match the signal light.

[0059] 3. Microscopic Interference and Imaging

[0060] The micro-interference optical path is the core imaging part of the system. The beam (red) from the first optical comb 101 (signal optical comb) serves as the main light source. After being processed by the first polarizer 201 and the first beam expander 203, it reaches the main beam splitter 401.

[0061] Sample arm: A portion of the beam is reflected by the main beam splitter 401, passes through the imaging lens 403, and enters the back focal plane of the objective lens 404, thereby providing wide-field vertical illumination to the patterned wafer 405 placed on the high-precision displacement stage 406.

[0062] Reference arm: Another portion of the beam passes through the main beam splitter 401, reaches the fixed reflector 402, then returns along the same path and is reflected by the main beam splitter 401 into the sampling beam splitter 501. This is the "internal reference beam". The reference beam can provide a position reference and benchmark for the signal beam. It is equivalent to giving a starting position, and the measured position information is the relative distance with the plane mirror. After subtracting the calibrated initial position, the surface undulations of the sample can be obtained.

[0063] The returned signal light (return light) reflected by the patterned wafer 405 is collected by the objective lens 404 and enters the imaging lens 403. The imaging lens 403 is a key component of the infinity-corrected microscope system, used to focus the parallel light collected by the objective lens to form a clear, wide-field image on the subsequent camera target surface. After passing through the main beam splitter 401, the return light reaches the sampling beam splitter 501. At the same time, the prepared reference beam (blue) also reaches the sampling beam splitter 501, where it spatially overlaps with the signal beam, forming interference.

[0064] 4. Signal monitoring, processing and detection

[0065] The beat frequency monitoring module is used for real-time monitoring. The sampling beam splitter 501 transmits a small portion of the interference beam to the high-speed photodetector 502. The output signal of the high-speed photodetector 502 can be used to monitor the beat frequency signal in real time, providing feedback for the frequency setting of the arbitrary waveform generator 303 and ensuring the accuracy of downsampling.

[0066] After being reflected by the sampling beam splitter 501, the main interference beam enters the subsequent detection optical path. First, the beam passes through a pulse width broadening module, namely a narrowband filter 601. This filter is used to broaden the pulse width of the interference signal in the time domain to improve the signal-to-noise ratio of the camera's acquisition within a limited exposure time.

[0067] The synchronous detection module is responsible for the final signal acquisition. The broadened interference beam enters the polarization beam splitter 701 (PBS) and is decomposed into two orthogonal polarization components, H and V. These two components are detected by the first camera 702 and the second camera 703, respectively.

[0068] 5. Electronic control and data acquisition

[0069] The high-frequency synchronization signal from the light source module is split into two paths: one path is used to synchronize the arbitrary waveform generator 303, and the other path is processed by the synchronization controller / frequency divider circuit 704 to generate a low-frequency trigger signal. This strictly synchronizes the operation of the first camera 702, the second camera 703, and the image acquisition card 801, ensuring that both cameras expose and read out at exactly the same time and frame rate, thereby achieving the synchronization of the two time-domain interference signals. and Precise synchronous acquisition. The high-frequency synchronization signal is an electrical signal, typically 10MHz, emitted by two optical comb light sources. Figure 1 The dashed lines connecting the first optical comb 101 and the second optical comb 102 to the arbitrary waveform generator 303 and the synchronization controller / frequency divider circuit 704 are used to represent this.

[0070] The acquisition and processing module is the system's terminal. Image sequences acquired by the cameras are transmitted at high speed to the processing unit 802 (e.g., a workstation equipped with a high-performance computing card) via the image acquisition card 801. The processing unit 802 receives the raw interferometric data streams from the two cameras and, according to the method described in this invention, performs Fourier transform, complex spectral analysis, and parallel calculation of multiple physical parameters, ultimately generating a quantitative detection report and visualization spectrum of wafer defects. Simultaneously, the software program within the processing unit 802 controls the real-time synchronous scanning of the high-precision displacement stage 406. The high-precision displacement stage can be purchased directly, and the manufacturer provides the corresponding control software.

[0071] Through the coordinated operation of the above functional modules, this system can achieve non-contact, wide-field, multi-parameter synchronous detection of patterned wafers.

[0072] II. Detection Methods and Principles

[0073] Figure 2 A flowchart of a patterned wafer multi-parameter synchronous detection method according to an embodiment of the present invention is shown. The following will refer to... Figure 1 The system architecture, for Figure 2 Each step in the process will be explained in detail.

[0074] Step S201: System settings and lighting.

[0075] First, the dual-comb light source module is activated to generate signal light from the first optical comb 101 and reference light (local light) from the second optical comb 102. The signal light (red path) is processed by the illumination and polarization control module and then reaches the main beam splitter 401, where it is split into two paths:

[0076] Sample arm: A portion of the light beam is reflected and then illuminates the patterned wafer 405 with a wide field through objective lens 404.

[0077] Reference arm: Another part of the beam is transmitted through the imaging lens 403 and reaches the fixed reflector 402 to form an internal reference optical path.

[0078] Step S202: Signal collection and polarization beam splitting.

[0079] The "sample beam" returning from the patterned wafer 405 and the "internal reference beam" returning from the mirror 402 propagate coaxially after the main beam splitter 401. These two beams converge at the sampling beam splitter 501 and undergo heterodyne interference with the reference beam (blue optical path) processed by the acousto-optic frequency shifting module. The resulting composite beam passes through the pulse width broadening module and enters the synchronous detection module, where the polarization beam splitter 701 (PBS) strictly splits it into two mutually orthogonal polarization components: a horizontal polarization (H) component and a vertical polarization (V) component.

[0080] Step S203: Synchronous time-domain interference signal acquisition.

[0081] The H and V polarization components are acquired at high speed by the first camera 702 and the second camera 703, respectively, and the two cameras record simultaneously. and Two time-domain interferometric pattern sequences, among which, t For time series values, x and y These represent the pixel values. Due to the optical path difference between the sample arm and the reference arm, each acquired temporal interferometric data... and Each iteration will contain two temporally separated interference envelopes: a reference envelope and a sample envelope. This "synchronization" is crucial for ensuring the accuracy of subsequent difference calculations.

[0082] Step S204: Frequency domain transformation.

[0083] After receiving the two channels of temporal interferometric data, the processing unit 802 processes each pixel. Time series data and Performing a Fast Fourier Transform (FFT) transforms it from the time domain to the frequency domain, yielding two core complex spectra:

[0084] in It is the complex spectrum of the H channel. This is the complex spectrum of the V channel, where A is the amplitude. It is phase. i It is the sequence number of the interface echo. It is optical frequency.

[0085] Step S205: Simultaneous solution of multiple parameters.

[0086] This is the core data processing step of the present invention. Based on the complex spectral pairs obtained in the previous step, the processing unit 802 performs parallel decoding of multiple computational paths to synchronously retrieve the multi-dimensional physical parameters of the sample (see reference). Figure 3 These parameters include, but are not limited to, three-dimensional surface morphology, internal defect location, surface stress distribution, film thickness d, and refractive index. Extinction coefficient and the broadband reflectance spectrum of the material .

[0087] Step S206: Defect identification and characterization.

[0088] Processing unit 802 processes the thin film thickness map calculated in step S205. Refractive index diagram Broadband reflectance spectrum Surface stress distribution diagram Various physical quantity parameter graphs are compared and analyzed with preset process control standards or design databases. Through algorithms such as threshold judgment, image subtraction, and pattern recognition, non-compliant areas are automatically identified and located as defects, such as excessive stress concentration points, uneven film thickness areas, surface contaminants, or underlying voids.

[0089] Step S207: Result generation and output.

[0090] Finally, the system visualizes the defect identification results, generating a user-friendly defect map that clearly marks the location, type, and related quantitative parameters of defects. It also generates detailed quantitative inspection reports for process traceability, yield analysis, and archiving.

[0091] Through the above steps, the embodiments of the present invention can complete one-time, all-round, and quantitative defect detection of patterned wafers.

[0092] III. Detailed Solution of Multiple Physical Quantity Parameters

[0093] Reference Figure 3 This figure illustrates in detail the complete calculation process from raw data acquisition to final multidimensional physical parameters. The starting point of the entire data processing flow is the time-domain interferometric signal pair directly acquired in step S203. and After receiving the raw data, the processing unit 802 starts two main processing paths in parallel: the time-domain direct analysis path and the frequency-domain transform analysis path.

[0094] In the direct analysis path in the time domain ( Figure 3 In the coarse localization section of path one, the processing unit 802 directly performs peak finding and localization of the interference envelope of the time-domain signal S(t), which can accurately determine the peak time position of the interference envelope generated by reflections from the sample surface and internal interfaces, thereby calculating the time delay without 2π ambiguity between them. And obtain rough depth information. Optical time delay = / M (M is the compression ratio) / Based on this time delay, the three-dimensional surface morphology of the sample and the three-dimensional spatial location of internal defects (such as delamination and voids) can be calculated. If the sample is a semi-transparent multilayer structure, multiple interference envelopes will appear in the time-domain signal. = The optical thickness can then be calculated. This characterizes the layered structure of the sample. In the formula, c It is the speed of light in a vacuum. w 0 Indicates the center frequency.

[0095] In the frequency domain transform analysis path, the processing unit 802 first processes the original time domain signal. and Perform a Fourier transform (FFT) to obtain complex spectral data pairs containing complete amplitude and phase information. and This pair of complex spectra forms the basis for all subsequent detailed analyses. Then, processing unit 802 performs parallel, multi-path (A, B, C, D) decoding calculations based on this pair of complex spectra to simultaneously extract multi-dimensional physical quantity parameters. This process does not calculate individual physical quantities sequentially and in isolation, but rather starts from the same raw data and, through multiple parallel calculation paths, synchronously and collaboratively solves for the multi-dimensional physical properties of the sample under test. The calculations specifically performed by processing unit 802 include, but are not limited to, the following aspects:

[0096] A. Calculation of three-dimensional morphology and internal structural parameters (coarse-fine combined high-precision method)

[0097] This computational path aims to acquire spatial structural information of the sample with extremely high precision. For high-precision morphology calculations ( Figure 3 (The precise positioning and integration part of path one), the processing unit starts from... Carrier phase information is extracted to obtain a highly sensitive relative depth. Finally, the result obtained by combining the time-domain path... and frequency domain path obtained The final depth, which combines a large range and high accuracy, is calculated. .

[0098] In optical measurement, the accuracy of depth measurement using only the time-of-flight (ToF) method based on the peak value of the interference envelope is limited by the bandwidth of the light source, and is usually on the order of micrometers. On the other hand, the accuracy of the interferometry based on the carrier phase can reach the nanometer or even sub-nanometer level, but the measurement results are subject to a periodic ambiguity of 2π, which limits the measurement range to a single wavelength and makes it impossible to measure steps or thicknesses larger than the wavelength.

[0099] To solve this technical problem, a preferred embodiment of the present invention employs a two-step method that combines coarse localization based on time-domain envelope with fine localization based on frequency-domain carrier phase.

[0100] A.1 Coarse localization based on temporal interferometric envelope

[0101] This step directly analyzes the original time-domain interference signal. The processing unit determines the time delay of the echoes at each interface using a peak-finding algorithm. (i is the sequence number of the interface echo, The time delay of the i-th echo envelope aims to perform unambiguous, wide-range depth position prediction, which directly utilizes the original time-domain interferometric signal acquired in step S203. .

[0102] 1. Data and Processing: Processing unit 802 directly analyzes time-domain interferometric signals. The peak time of the interference envelope corresponding to each reflection interface is determined by the peak-finding algorithm. ( 1, 2, 3, ...).

[0103] 2. Coarse depth calculation: based on the time delay between each peak. = - Calculate a zero Coarse depth value of phase blur :

[0104]

[0105] This coarse positioning value provides crucial integer periodic information for subsequent precise positioning. Specifically:

[0106] It is the speed of light in a vacuum.

[0107] It is the first i The measured time delay of each echo envelope relative to the first echo envelope (surface) in the radio frequency domain.

[0108] M is the frequency compression ratio of the dual optical comb system. ,in It is the repetition frequency of the optical comb. It is the difference in repetition frequency between the two optical combs.

[0109] It is the group refractive index of the sample medium at the center wavelength of the optical comb.

[0110] A.2 Precise positioning based on frequency domain carrier phase

[0111] The purpose of this step is to achieve ultra-high precision measurements at the nanometer level based on coarse positioning. It utilizes the complex spectrum obtained through Fourier transform in step S204. .

[0112] 1. Data and Processing: Processing unit 802 processes data from complex spectra. In the process, the center frequency with a high signal-to-noise ratio is extracted. carrier phase at .

[0113] 2. Relative Depth Calculation: The precise depth related to the phase 2πN, which is an integer multiple of 2π, is given by the following formula:

[0114]

[0115] in For precise depth, It is the wavelength corresponding to the center frequency.N It is an unknown phase-wound integer. For the sample medium at the center frequency The phase refractive index at that location. This step actually calculates the exact fractional part of the relative depth.

[0116] A.3 Final Depth Calculation Combining Coarse and Fine Positioning

[0117] This step is the core of achieving high-precision, large-range measurement and is completed by the processing unit 802.

[0118] 1. Determination of the integer period N: Processing unit 802 uses the coarse depth value obtained in step A.1. To determine the correct integer period corresponding to the phase measurement in step A.2. N The calculation method is as follows:

[0119]

[0120] Where round is the rounding function.

[0121] 2. Final high-precision depth calculation: Combine the determined integer N with the measured phase. Substituting the values ​​into the formula in A.2, we can obtain the final depth value that combines a large range and high accuracy. .

[0122] 3. Output results: By analyzing all pixels within the field of view By executing the above coarse-fine combined algorithm, a high-precision three-dimensional surface topography image of the sample, the precise three-dimensional coordinates of internal defects, and the physical thickness of the thin film can be output. ), and These represent the depths of the upper and lower surfaces of the film, respectively.

[0123] B. Thin film optical parameters (thickness d, refractive index) Extinction coefficient ) calculation

[0124] This calculation path is used to extract thin film optical parameters. The processing unit 802 utilizes the complex spectral information of a single channel, namely the amplitude spectrum. and phase spectrum Based on a pre-defined physical model (e.g., a multilayer optical model containing Fresnel equations), the physical thickness d of the thin film, the refractive index n(ω) as a function of frequency, and the extinction coefficient k(ω) can be calculated simultaneously and accurately through iterative fitting or direct solution.

[0125] This approach is based on the frequency domain complex spectrum S̃(ω), and uses a physical model to invert the thin film parameters. The interaction of the samples can be represented by a complex reflection (or transmission) coefficient. To describe the complex spectrum we measured It is directly proportional to this. For a matrix with thickness d and a complex refractive index of d, A single-layer thin film, the underlying substrate having a complex refractive index of Its total complex reflection coefficient Given by Fresnel equations:

[0126]

[0127] in, and These are the Fresnel reflectance coefficients of the air-film interface and the film-substrate interface, respectively. and β is a function of β. β is the complex phase delay of light during a single passage through the thin film: .

[0128] Processing unit 802 is configured to perform iterative fitting using a nonlinear least squares method (such as the Levenberg-Marquardt algorithm), by adjusting the undetermined parameters (such as physical thickness). and description and The dispersion model coefficients ( ) make the theoretical complex spectrum calculated by the above formula consistent with the experimentally measured . The amplitude spectrum and phase spectrum of the film are simultaneously optimally matched, thus allowing for the synchronous and accurate solution of various parameters of the thin film.

[0129] C. Identification of material components (broadband reflectance spectroscopy) )

[0130] This calculation path is used to extract material composition parameters and the absolute reflectance spectrum of the sample. With amplitude spectrum Directly related. The processing unit 802 extracts the amplitude spectra of the two channels respectively. and The energy is combined and compared with that from a reference sample (such as a gold or silver mirror), whose reflectivity The reference amplitude spectrum obtained from a known and approximate point (1) is normalized. This yields the absolute, frequency-dependent broadband reflectance spectrum of the sample at that pixel. By comparing this spectrum with a standard materials database, chemical or material information about the sample surface can be obtained. The identified chemical composition of the material can be used to determine contaminants, residues, or material deposition errors.

[0131] 1. Calculation of total intensity spectrum:

[0132] To obtain the total reflection intensity of the sample independent of polarization The processing unit 802 adds the energy from the two channels:

[0133] .

[0134] 2. Absolute reflectance calculation:

[0135] By comparing with the reference intensity spectrum measured on the standard reference mirror. Normalization is performed to obtain the absolute reflectance spectrum of the sample at each pixel. :

[0136]

[0137] in This is the known spectral reflectance of the standard reference mirror. The obtained... It is a spectral curve covering the bandwidth of a dual-comb light source, equivalent to performing a microscopic spectral analysis on that point. By comparing it with the built-in standard material spectral database, the material composition at that point can be identified, which can be used to determine whether there are defects such as chemical residues, material contamination, or deviations in the composition of the metal layer.

[0138] D. Calculation of polarization parameters and surface stress (σ)

[0139] This calculation path is a direct application of the polarization-sensitive characteristics of the present invention, used to quantify material anisotropy caused by mechanical stress. In this calculation path, the processing unit 802 compares the complex spectra of two orthogonal polarization channels. and To extract polarization parameters and surface stress information.

[0140] D.1 Elliptic parameters ( , )calculate

[0141] The ellipticity parameters are given directly from the amplitude ratio and phase difference of the two channels:

[0142]

[0143]

[0144] in These are the amplitude and phase of the H and V channels, respectively.

[0145] D.2 Birefringence calculate

[0146] Birefringence Due to phase difference (Also referred to as phase delay) The calculation yielded the following:

[0147]

[0148] in, It is the physical thickness of the sample, which can be precisely provided by path A or B.

[0149] D. Calculation of principal stress difference (σ1-σ2)

[0150] This is one of the key innovations of this invention: stress in the material causes birefringence, and according to the photoelastic law, the magnitude of birefringence is proportional to the principal stress difference.

[0151]

[0152] in This is the known photoelastic coefficient of the material, and the material resolution can be obtained through path C. Finally, the processing unit 802 generates a two-dimensional stress distribution map, clearly showing the stress concentration areas on the wafer surface.

[0153] In summary, this invention employs an analytical framework that coordinates time-domain and frequency-domain information processing. Through multiple parallel computing paths, it comprehensively and fully decodes the rich information in the vector dual-comb interference signal, enabling the simultaneous characterization of multiple key physical parameters of a patterned wafer in a single measurement. This ensures that the extraction of all physical parameters originates from the same measurement, truly achieving quantitative characterization of wafer defects and multiple key physical parameters, and providing a powerful new tool for advanced process control and defect analysis in the semiconductor field.

[0154] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dual-comb based pattern wafer multi-quantity synchronous detection system, characterized in that, Comprise: A signal light path comprising a first optical comb, a first polarizer, a first beam expander arranged along the light path, the signal light path being used to generate a probe light beam; A reference light path comprising a second optical comb, an acousto-optic frequency shift module, a second polarizer, a second beam expander arranged along the light path, the acousto-optic frequency shift module being used to perform optical frequency shift on the light beam, the reference light path being used to generate a reference light beam; A beam splitting unit comprising a main beam splitter and a sampling beam splitter; A detection light path comprising an imaging lens, an objective lens, a displacement stage arranged along the light path, the probe light beam after being reflected by the main beam splitter irradiates a pattern wafer placed on the displacement stage through the detection light path and generates reflected return light, the reflected return light after passing through the objective lens, the lens, and the main beam splitter reaches the sampling beam splitter, and interferes with the reference light beam reaching the sampling beam splitter to form an interference light beam; A synchronous detection module comprising a polarization beam splitter and a first camera and a second camera, the interference light beam after being reflected by the sampling beam splitter enters the polarization beam splitter and is decomposed into two orthogonal polarization components of horizontal polarization component H and vertical polarization component V, which are detected by the first camera and the second camera respectively.

2. The dual-comb based pattern wafer multi-parameter synchronous detection system according to claim 1, wherein, The acousto-optic frequency shift module comprises a first acousto-optic modulator and a second acousto-optic modulator arranged in series along the reference light path, the first acousto-optic modulator and the second acousto-optic modulator are driven by the same arbitrary waveform generator, and an optical frequency shift is applied to the light beam by controlling the driving frequency difference applied to the first acousto-optic modulator and the second acousto-optic modulator.

3. The dual-comb based pattern wafer multi-parameter synchronous detection system according to claim 2, wherein, Further comprising a synchronous controller / frequency division circuit and a collection and processing module, the collection and processing module comprising an image collection card and a processing unit; After the high-frequency synchronization signals from the first optical comb and the second optical comb are split, one is used to synchronize the arbitrary waveform generator, and the other is processed by the synchronous controller / frequency division circuit to generate a low-frequency trigger signal, which synchronizes the operation of the first camera, the second camera, and the image collection card; The image sequences collected by the first camera and the second camera are transmitted to the processing unit through the image collection card, and the processing unit generates the detection result of the pattern wafer defects after processing the data.

4. The dual-comb based pattern wafer multi-parameter synchronous detection system according to claim 3, wherein, Further comprising a beat frequency monitoring module for receiving the interference light beam transmitted from the sampling beam splitter, and for feedback adjusting the acousto-optic frequency shift module by monitoring the beat frequency signal in real time.

5. The dual-comb based pattern wafer multi-parameter synchronous detection system according to claim 3, wherein, The upstream of the synchronous detection module is further provided with a pulse width expansion module, and the interference light beam after being expanded in pulse width in time domain by the pulse width expansion module enters the polarization beam splitter again.

6. The dual-comb based pattern wafer multi-parameter synchronous detection system according to claim 3, wherein, One side of the main beam splitter is provided with a fixed mirror, and the probe light beam after being transmitted through the main beam splitter is reflected by the fixed mirror and then passes through the main beam splitter again and then enters the sampling beam splitter as an internal reference light beam.

7. A dual-comb based pattern wafer multi-parameter synchronous detection method, characterized in that, The pattern wafer multi-parameter synchronous detection method based on the double optical comb of any one of claims 3-6 comprises the following steps: S1, a probe light beam is generated by a signal light path, a reference light beam is generated by a reference light path, the probe light beam is reflected by a main beam splitter, and then irradiates a pattern wafer placed on a displacement table through a detection light path and generates reflected return light, the return light passes through the objective lens, the lens and the main beam splitter, reaches the sampling beam splitter, and interferes with the reference light beam to form an interference light beam; the interference light beam is reflected by the sampling beam splitter, and then enters the polarization beam splitter to be decomposed into two orthogonal polarization components H and V, which are detected by a first camera and a second camera respectively and simultaneously record and two-way time-domain interference data, wherein t is a time series value, x and y is a pixel value; S2, after the processing unit receives two-way time-domain interference data, for each pixel point time series data and perform fast Fourier transform, from time domain to frequency domain, get the core complex spectrum pair; S3, the processing unit performs decoding on multiple calculation paths in parallel based on the complex spectrum pairs obtained in step S2 to inversely calculate the multi-dimensional physical parameters of the pattern wafer synchronously; S4, the processing unit compares the physical parameter maps calculated in step S3 with the preset process control standards or design library, and identifies and locates the non-compliant areas as defects.

8. The dual-comb based pattern wafer multi-parameter synchronous detection method according to claim 7, characterized in that, The multi-dimensional physical parameters of the pattern wafer in step S3 include three-dimensional surface topography, internal defect positioning, surface stress distribution, film thickness, refractive index, extinction coefficient, and broadband reflection spectrum of the material.

9. The dual-comb based pattern wafer multi-parameter synchronous detection method according to claim 7, wherein, In steps S2 and S3, after receiving the two-way time-domain interference data, the processing unit starts two main processing paths in parallel, namely a time-domain direct analysis path and a frequency-domain transformation analysis path. In the time-domain direct analysis path, the processing unit performs peak searching and positioning of the interference envelope of the time-domain signal, determines the peak time position of the interference envelope generated by the reflection of each interface on the surface and inside of the pattern wafer, and calculates the time delay between each peak time position without 2π ambiguity , and obtains rough depth information . Optical time delay = / M, M is the compression ratio / , based on the optical time delay , the three-dimensional surface topography of the pattern wafer and the three-dimensional spatial position of the internal defects can be calculated; if the sample is a semi-transparent multi-layer structure, multiple interference envelopes will appear in the time-domain signal, and based on = , the optical thickness can be calculated, thereby representing the layered structure of the sample, wherein is the group refractive index, is the center frequency, is the speed of light in vacuum; In the frequency-domain transformation analysis path, the processing unit first performs Fourier transform on the time-domain signal to obtain a complex spectrum data pair containing complete amplitude and phase information, and then performs parallel and multi-path decoding calculation based on the obtained complex spectrum data pair to synchronously extract multi-dimensional physical quantity parameters.

Images