A system and method for super-resolution optical detection of surface defects of a material

Abstract

The application relates to the technical field of optical detection, and discloses a material surface defect super-resolution optical detection system and method. The system and method utilize an evanescent wave chip to form a structured evanescent wave field on a surface as a near-field illumination probe, change the relative position of the probe and a sample through an angle or wavelength tuning non-mechanical scanning mode, and acquire a set of far-field polarization resolution diffraction patterns through a diffraction signal acquisition module. A central control and data processing unit carries out iterative calculation on the diffraction data set through a vector coherent diffraction imaging algorithm, and simultaneously reconstructs the complex amplitude distribution of the probe and a high-resolution Jones matrix image of the sample. Through analysis of the Jones matrix, quantitative detection of the morphology, size and anisotropic physical properties of the surface defect can be completed. The application avoids vibration caused by mechanical scanning, and realizes high-stability and high-resolution nondestructive detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optical detection, in particular to a super-resolution optical detection system and method for material surface defects. BACKGROUND

[0002] In the fields of semiconductor manufacturing, optical component production, and new material research, it is crucial to accurately detect the integrity of material surfaces. Small defects on the surface, such as scratches, particles, contaminants, or local material property variations caused by stress, can affect the final performance and reliability of devices.

[0003] Currently, optical microscopes are widely used for surface detection due to their non-destructive and atmospheric operation characteristics. However, the detection capability of traditional far-field optical microscopes is limited by the fundamental physical limit of optical diffraction, and their spatial resolution is usually limited to about half of the wavelength of the light used. Therefore, when the size of the defect enters the nanometer scale, this method cannot effectively identify and distinguish it.

[0004] To break through the diffraction limit, various high-resolution detection techniques have been developed, such as scanning electron microscopes (SEM) and scanning probe microscopes (SPM), including atomic force microscopes (AFM). These techniques can provide nanometer or even atomic level spatial resolution, but they also have certain limitations. For example, SEM usually needs to operate in a high vacuum environment, and the high-energy electron beam can cause damage to some sensitive samples. SPM technology can work in normal environment, but its scanning speed is slow, and the interaction between the physical probe and the sample surface poses a risk of damaging the sample.

[0005] Furthermore, these high-resolution techniques, including some existing super-resolution optical techniques, mainly provide information about the sample's topography, topology, or light intensity distribution. They usually do not have the ability to directly measure the local anisotropic optical properties of the sample. However, in some cases, the manifestation of defects is not a change in geometric topography, but a change in local birefringence or dichroism caused by internal stress, lattice distortion, etc. The existing technology lacks a means to simultaneously obtain super-high resolution topography information and such vector optical property information.

[0006] In addition, existing scanning detection methods usually use fixed scanning step and sampling time to uniformly collect data on the entire preset area. This process is time-consuming. When a suspicious area is found in a global scan, if a more in-depth and detailed analysis of the specific area is needed, there is a lack of an integrated and dynamic mechanism to automatically redistribute subsequent detection resources (such as photon flux or sampling time) to the specific area. A new and more accurate scan is usually required, which reduces the overall efficiency and flexibility of the detection.

[0007] Therefore, there is a need in the art for a new technical solution that can achieve non-contact detection in an atmospheric environment, with spatial resolution not limited by the diffraction limit, and the ability to simultaneously obtain the topography and anisotropic physical properties of the sample, while also having the ability to dynamically optimize the data acquisition process. SUMMARY

[0008] In view of the deficiencies of the prior art, the present application provides a system and method for super-resolution optical detection of material surface defects, which solves the technical problems of insufficient resolution and difficulty in obtaining complex physical information such as polarization anisotropy of material surface defects in existing optical detection techniques.

[0009] To achieve the above object, the present application is implemented by the following technical solutions:

[0010] The present application provides, in a first aspect, a system for super-resolution optical detection of material surface defects.

[0011] The system includes a coherent light source, a plasmonic evanescent wave chip, a sample positioning module, a diffraction signal acquisition module, and a central control and data processing unit.

[0012] The coherent light source is used to generate light beams with a determined phase relationship. In one embodiment, the coherent light source is a laser.

[0013] The plasmonic evanescent wave chip includes a metal thin film and a two-dimensional periodic nanostructure arranged on the surface of the metal thin film. The plasmonic evanescent wave chip is functionally configured to form a structured evanescent wave field on the surface area of the plasmonic evanescent wave chip when it is excited by the light beams emitted by the coherent light source under specific incidence conditions.

[0014] The sample positioning module is used to carry a material sample to be detected and precisely position it within the action range of the structured evanescent wave field.

[0015] The diffraction signal acquisition module is arranged in the far field region and is used to acquire the diffraction pattern formed by the structured evanescent wave field interacting with the material sample and propagating to the far field.

[0016] The central control and data processing unit is configured to perform the following functions: based on a plurality of relative positions between the structured evanescent wave field and the material sample, a plurality of far-field diffraction patterns acquired by the diffraction signal acquisition module, a coherent diffraction imaging algorithm is used to reconstruct the surface information of the material sample; and further through analysis of the surface information, the detection of the material surface defects is completed.

[0017] In a preferred embodiment, the two-dimensional periodic nanostructure of the plasmonic evanescent wave chip is configured to interact with surface plasmons generated at the surface of the metal thin film, thereby forming the structured evanescent wave field at the surface of the plasmonic evanescent wave chip.

[0018] In a specific embodiment, the system further comprises a dynamic wavefront modulation module. The dynamic wavefront modulation module is disposed in the light path of the coherent light source, and is configured to dynamically and programmably modulate the wavefront or amplitude distribution of the light beam emitted by the coherent light source.

[0019] Further, the central control and data processing unit is further configured to perform an adaptive focal region detection procedure. The procedure comprises: first, the system performs a preliminary scan, and reconstructs a preliminary surface information based on the collected data; then, the central control and data processing unit analyzes the preliminary surface information, and identifies one or more specific regions containing potential defects or structural abnormalities; finally, the central control and data processing unit controls the dynamic wavefront modulation module to mainly spatially distribute and concentrate the illumination energy of the light beam emitted by the coherent light source on the one or more identified specific regions in subsequent scan detection.

[0020] In a specific embodiment, the diffraction signal collection module comprises a polarization resolution unit. The polarization resolution unit is configured to spatially separate the collected far-field diffraction patterns into at least two orthogonal polarization components, and record their intensity distributions by different detector arrays, respectively.

[0021] In a preferred embodiment, when the polarization resolution unit is configured, the coherent diffraction imaging algorithm employed by the central control and data processing unit is a vector coherent diffraction imaging algorithm. By running the vector coherent diffraction imaging algorithm, the central control and data processing unit reconstructs the Jones matrix (J) of the material sample from the intensity data of the plurality of far-field diffraction patterns corresponding to the at least two orthogonal polarization components as the surface information through an iterative calculation process. ) of the material sample from the intensity data of the plurality of far-field diffraction patterns corresponding to the at least two orthogonal polarization components as the surface information through an iterative calculation process.

[0022] The physical model and mathematical implementation of the iterative process are as follows:

[0023] When a vector probe light field is incident on a position of the sample, the exit light field can be described by the following formula:

[0024] ;

[0025] wherein, is an index of the scanning position; is a two-dimensional real space coordinate vector of the sample plane; is a two-dimensional real space coordinate vector of the scanning position indexed by ; is a vector probe field, i.e. a structured evanescent wave field, centered at ; is represented by a Jones vector with components representing the complex amplitudes of the light field in two orthogonal directions.

[0026] is the intensity of the diffraction pattern measured on the far field detector is proportional to the modulus square of the Fourier transform of the outgoing wave field , where . Here, denotes the Fourier transform operator, is a two-dimensional reciprocal space coordinate vector.

[0027] The iterative procedure comprises applying an update rule for the sample Jones matrix to correct the current estimate of the Jones matrix. In the th iteration, the update rule can be expressed as:

[0028] ;

[0029] where is an index of the iteration number; and are the estimates of the sample Jones matrix in the th and th iteration, respectively; is the first outgoing wave calculated based on the current estimate and ; is the second outgoing wave updated by imposing the amplitude constraints from the far field diffraction pattern on the Fourier transform of the first outgoing wave, followed by an inverse Fourier transform; is an error signal representing the difference between the second outgoing wave and the first outgoing wave; is the element-wise complex conjugate of the probe field ; is the maximum value of the probe field intensity, as a normalization factor; is a sample update step size parameter to control the convergence speed; is a small positive number to prevent the denominator from being zero.

[0030] Further, the iterative process further comprises applying a probe field update rule to correct the current estimate of the vector probe field of the structured evanescent wave field in synchronization with the correction of the current estimate of the Jones matrix. The update rule can be expressed as:

[0031] ;

[0032] wherein, and are the estimates of the vector probe field at the first and the second iteration, respectively; is the conjugate transpose of the sample Jones matrix estimate ; is the maximum value of a certain norm of the sample Jones matrix , as a normalization factor; is a probe update step size parameter to control the convergence speed; and the remaining parameters , , , , , , are defined as before.

[0033] In a preferred embodiment, the change of the relative position is achieved by the central control and data processing unit controlling the coherent light source to fine tune the incident angle or wavelength.

[0034] The second aspect of the present application provides a method for super-resolution optical detection of surface defects of a material, which is applied to the super-resolution optical detection system of any of the preceding embodiments. The method comprises the following steps:

[0035] S1, exciting a plasmonic evanescent wave chip with a coherent light source to form a structured evanescent wave field on the surface of the plasmonic evanescent wave chip;

[0036] S2, changing the relative position between the structured evanescent wave field and a material sample to generate a set of far-field diffraction patterns formed by the interaction between the structured evanescent wave field and the material sample;

[0037] S3, collecting the set of far-field diffraction patterns generated in the preceding step;

[0038] S4, first, reconstructing the surface information of the material sample based on the collected set of far-field diffraction patterns using a coherent diffraction imaging algorithm; and then, detecting the surface defects by analyzing the surface information.

[0039] The present application provides a material surface defect super-resolution optical detection system and method. The following advantages are provided:

[0040] 1、The present application uses structured evanescent wave field for near-field illumination, and combines coherent diffraction imaging algorithm to calculate and reconstruct the collected far-field diffraction pattern. This technical path makes the final imaging resolution depend on the range of collected signals in reciprocal space, rather than the wavelength of the illumination light, thereby breaking through the diffraction limit of traditional optical microscopes, and enabling super-resolution detection of micro-nano scale defects on the material surface.

[0041] 2、The present application sets up a polarization resolution unit in the diffraction signal collection module, and applies a vector coherent diffraction imaging algorithm for data processing. This scheme can reconstruct the Jones matrix representing the anisotropic properties of the sample from the separated polarization component data. Therefore, the present application can not only obtain the topographic information of the sample, but also obtain its polarization-related physical information, thereby being able to detect and characterize surface defects with birefringence or dichroism properties.

[0042] 3、The present application introduces a dynamic wavefront modulation module, and configures a corresponding adaptive detection process, which can concentrate the energy of the illumination light beam on the identified specific area after completing the preliminary scanning. This technical solution can increase the signal intensity and signal-to-noise ratio of the key area without increasing the total detection time, thereby improving the data quality for accurately characterizing the defects in the area. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The structural schematic diagram of the material surface defect super-resolution optical detection system of an embodiment of the present application;

[0044] Figure 2 The cross-sectional structure and light field forming principle schematic diagram of the plasmonic evanescent wave chip 103 region in an embodiment of the present application;

[0045] Figure 3 The flowchart of the vector coherent diffraction imaging reconstruction algorithm of an embodiment of the present application;

[0046] Figure 4 The schematic diagram of the adaptive key area detection process of an embodiment of the present application;

[0047] Figure 5 The flowchart of the material surface defect super-resolution optical detection method of an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0049] Referring to the drawings Figure 1 , Figure 1 Figure 1 is a schematic diagram of a material surface defect super-resolution optical detection system according to an embodiment of the present application. The system provided by the present application can include a coherent light source 101, a dynamic wavefront modulation module 102, a plasmonic evanescent wave chip 103, a sample positioning module 104, a diffraction signal acquisition module 105, and a central control and data processing unit 106.

[0050] The coherent light source 101 is configured to generate a light beam with a determined phase relationship and polarization state. In a specific implementation, the coherent light source 101 is a single-mode tunable laser, and the wavelength and intensity of the output light beam are controlled by the instructions of the central control and data processing unit 106.

[0051] The dynamic wavefront modulation module 102 is arranged on the light beam output path of the coherent light source 101, and is configured to modulate the wavefront or amplitude of the incident light beam in space according to the instructions from the central control and data processing unit 106. In a specific implementation, the dynamic wavefront modulation module 102 can be a liquid crystal spatial light modulator (LCSLM) or a digital micromirror device (DMD).

[0052] The plasmonic evanescent wave chip 103 includes a transparent substrate, a metal film deposited on the substrate, and a two-dimensional periodic nanostructure prepared on the surface of the metal film. The plasmonic evanescent wave chip 103 is configured to form a spatially structured evanescent wave field on the surface of the chip when receiving the excitation light beam from the coherent light source 101 and modulated by the dynamic wavefront modulation module 102.

[0053] The sample positioning module 104 is configured to carry the material sample 104a to be detected, and accurately place the surface of the material sample 104a in the action range of the structured evanescent wave field generated by the plasmonic evanescent wave chip 103. The positioning process needs to achieve nanometer-level positioning accuracy to ensure effective near-field interaction between the evanescent wave field and the sample surface.

[0054] When the structured evanescent wave field interacts with the material sample 104a, the outgoing wave field can be mathematically described. In a scanning position index The outgoing wave field sample is given by the product of the transfer function (represented by a Jones matrix ), and the illumination probe field (represented by a Jones vector ):

[0055] ;

[0056] where, is the index of the scanning position; is the two-dimensional real space coordinate vector of the sample plane; is the probe center coordinate vector corresponding to the scanning position identified by the index ; is the vector probe field centered at , i.e., the structured evanescent wave field, represented as a Jones vector, whose components represent the complex amplitudes of the light field in two orthogonal directions is the Jones matrix of the material sample 104a, a complex matrix that describes the transformation of the incident light polarization state at the spatial position of the sample. is the outgoing wave field after passing through the sample, a Jones vector; this outgoing wave propagates to the far field and forms a diffraction pattern.

[0057] The diffraction signal acquisition module 105 is arranged in the far field region and is used to acquire the far field diffraction pattern generated after the above interaction. In a specific embodiment, the diffraction signal acquisition module 105 includes a polarization resolution unit 105a and one or more detector arrays 105b (such as a CCD or CMOS camera). The polarization resolution unit 105a, such as a polarization beam splitter, functions to spatially separate the far field diffraction beam carrying sample information into two orthogonal polarization components and project these two components onto different detector arrays 105b for intensity recording. In an alternative embodiment, the polarization resolution unit 105a can be a Wollaston prism or a Glan-Thompson prism, which also achieves spatial polarization separation. In another embodiment, this function can be achieved by placing a rotatable linear polarizer in front of the detector and sequentially acquiring diffraction patterns at at least three different polarization angles (such as 0 degrees, 45 degrees, and 90 degrees), and then calculating and synthesizing the complete polarization information. This method only requires one detector array. The detector array 105b converts the recorded light intensity distribution into a digital signal and transmits it to the central control and data processing unit 106.

[0058] The central control and data processing unit 106, usually a computer equipped with a data acquisition card and a high-performance computing unit (such as a GPU). It is configured to perform the following functions:

[0059] Control the timing of the entire system, including sending control signals to the coherent light source 101 and the dynamic wavefront modulation module 102 to achieve scanning illumination of different regions of the sample, and synchronously triggering the diffraction signal acquisition module 105 to perform data acquisition.

[0060] Receive and store a set (or multiple sets corresponding to different polarization components) of far-field diffraction pattern data collected by the diffraction signal acquisition module 105 at multiple scanning positions.

[0061] Based on the collected diffraction pattern data, perform a vector coherent diffraction imaging algorithm to reconstruct the complex Jones matrix of the material sample 104a through iterative calculation As its surface information.

[0062] By analyzing the reconstructed Jones matrix For example, by analyzing its amplitude and phase distribution, the positioning, size measurement and physical property characterization of the material surface defect are completed.

[0063] The overall workflow of the super-resolution optical detection system is coordinated and controlled by the central control and data processing unit 106. The core working principle is to use a structured near-field evanescent wave to scan the sample, encode the high-frequency spatial information of the sample into a diffraction pattern that can be recorded by a far-field detector, and then decode these information through a computational imaging algorithm to reconstruct a sample image beyond the diffraction limit in digital space.

[0064] At the beginning of a detection cycle, the central control and data processing unit 106 sends instructions to the coherent light source 101 to generate a beam with specific parameters. The beam is shaped by the dynamic wavefront modulation module 102 in turn, and is incident on the plasmonic evanescent wave chip 103. On the surface of the plasmonic evanescent wave chip 103, the beam energy is coupled into a structured evanescent wave field, which acts as an illumination probe to illuminate a local region of the material sample 104a within its range. After the interaction of the light field and the material sample 104a, an outgoing wave carrying sample information is formed.

[0065] The outgoing wave propagates to the far field and is received by the diffraction signal acquisition module 105. If the system is configured with polarization resolution, the polarization resolution unit 105a will decompose the outgoing wave into orthogonal polarization components, and the detector array 105b will record the intensity distribution of the diffraction pattern formed in the far field by each component to form a data frame of a single scan. This process converts and stores the complex transmission function information (including amplitude and phase) of the sample into measurable far-field light intensity information.

[0066] Subsequently, the central control and data processing unit 106 changes the spatial distribution of the structured evanescent wave field on the surface of the plasmonic evanescent wave chip 103 by controlling the parameters of the dynamic wavefront modulation module 102 or the coherent light source 101, thereby realizing the relative lateral displacement of the illumination probe on the surface of the material sample 104a and illuminating the adjacent area of the sample. The system repeatedly performs the above-mentioned illumination and acquisition process until the entire area to be detected is covered, thereby obtaining a set of far-field diffraction pattern data sets acquired at multiple overlapping scanning positions.

[0067] The data set is transmitted to the data processing part of the central control and data processing unit 106. The central control and data processing unit 106 runs a vector coherent diffraction imaging reconstruction algorithm, uses the data set as input, and simultaneously reconstructs the complex amplitude distribution information (i.e. of the illumination probe and the high-resolution Jones matrix distribution (i.e. ) of the surface of the material sample 104a through iterative calculation. The finally reconstructed Jones matrix image is the surface information containing the sample topography and the polarization anisotropy information, and the detection of surface defects can be completed by analyzing the information.

[0068] Referring to the accompanying drawings Figure 2 , Figure 2 is a schematic diagram of the cross-sectional structure of the plasmonic evanescent wave chip 103 region and the principle of light field formation according to an embodiment of the present application. The diagram shows the incident light beam, the substrate 201, the metal film 202, the two-dimensional periodic nanostructure 203, and the structured evanescent wave field 204 formed on the surface of the chip.

[0069] The formation of the structured evanescent wave field 204 is based on the excitation of surface plasmons and their interaction with the two-dimensional periodic nanostructure 203. In a typical Kretschmann excitation configuration, the P-polarized light beam from the coherent light source 101 passes through a high refractive index transparent substrate 201 (for example, a coupling prism) and is incident on the interface between the substrate 201 and the metal film 202 at an incident angle larger than the critical angle of total internal reflection.

[0070] At the interface, only when the wave vector component of the incident light in the interface plane is equal to the wave vector of the surface plasmon supported by the metal-dielectric interface When matched, the energy of the incident light is efficiently coupled to the surface of the metal film 202, exciting surface plasmons. This momentum matching condition can be expressed as:

[0071]

[0072] where the parallel component of the incident light wave vector is given by:

[0073]

[0074] and the magnitude of the surface plasmon wave vector is determined by:

[0075]

[0076] In the above equations:

[0077] is the wavelength of the coherent light source 101 in vacuum; is the refractive index of the transparent substrate 201; is the angle of incidence of the light beam on the substrate metal interface; is the dielectric constant of the metal film 202, which is a complex number with a negative real part; is the dielectric constant of the medium (e.g., air or liquid) in contact with the other side of the metal film 202, by precisely adjusting the angle of incidence or the wavelength , the above momentum matching condition can be satisfied, thus exciting surface plasmon waves that propagate along the interface on the surface of the metal film 202.

[0078] The excited surface plasmon waves scatter with the two-dimensional periodic nanostructure 203 pre-prepared on the surface of the metal film 202 during propagation along the surface of the metal film 202. The two-dimensional periodic nanostructure 203 can be a series of nanoholes, nanodisks, or other shaped nanoscattering bodies arranged in a specific lattice (such as a square or hexagonal lattice). In addition, the two-dimensional periodic nanostructure 203 can also be a quasi-periodic structure (such as a Penrose tiling structure), or in some embodiments, it can be simplified to a one-dimensional grating structure. In other embodiments, the structure can be a Fresnel zone plate structure or a photonic sieve structure for generating a specific focusing evanescent field. These different nanostructure designs can be used to generate structured evanescent wave fields with different spatial distribution characteristics to meet the needs of different detection tasks. These nanostructures act as secondary wave sources, scattering the propagating surface plasmon waves.

[0079] ​​​Because the positions of all scattering units in the two-dimensional periodic nanostructure 203 have a definite phase relationship, the electromagnetic fields scattered by each unit coherently superimpose (interfere) in the near-field region of the chip surface. This interference results in the formation of a non-propagating electromagnetic field above the chip surface, characterized by a spatially non-uniform distribution and specific intensity and phase modulation in the transverse plane—a structured evanescent wave field 204. The spatial structure of this field (e.g., periodicity, shape) is jointly determined by the geometric parameters of the two-dimensional periodic nanostructure 203 (such as lattice constant, unit shape, and size) and the parameters of the excitation light.

[0080] The resulting structured evanescent wave field 204 is an evanescent field whose field strength increases with distance from the normal direction of the chip surface. The decay rate decreases exponentially with increasing value, and its decay characteristic can be expressed as follows: ,in Let be the component of the wave vector in the normal direction. In the medium in contact with the metal thin film 202 (whose dielectric constant is...), The refractive index is In ), the normal wave vector component With the wave vector in the plane The relationship given by the wave equation Because the momentum matching condition is required when exciting surface plasmons. This led to It is a negative value, therefore It is a purely imaginary number. If we denote... Then the attenuation constant Given a positive real number, the field strength is... With distance Present Its form decays exponentially. It is precisely because of this evanescent characteristic that its effective range is strictly limited to a nanoscale region immediately adjacent to the chip surface, making it a high spatial resolution near-field illumination probe.

[0081] See attached document Figure 3 , Figure 3 This is a schematic flowchart of a vector coherent diffraction imaging reconstruction algorithm according to an embodiment of the present invention. The reconstruction process is executed by a central control and data processing unit 106, whose purpose is to calculate and recover the two-dimensional Jones matrix of the material sample 104a from a series of polarization-resolved far-field diffraction intensity patterns acquired by the diffraction signal acquisition module 105. .

[0082] First, in the data acquisition phase, for data collected from indexes... For each identified scan position, a polarization resolving unit 105a within the diffraction signal acquisition module 105 separates the diffracted beam into two orthogonal polarization components (e.g. horizontal and vertical polarization components). A detector array 105b records the far-field diffraction intensity of these two components separately, resulting in two sets of diffraction pattern data sets, and where are the reciprocal space coordinates. These data sets are the input of the entire reconstruction algorithm.

[0083] At the start of the reconstruction algorithm (step 301), the unknown sample Jones matrix and the illumination probe's Jones vector are set to initial estimates, denoted as and can be set to a unit matrix, indicating an initial assumption that the sample is completely transparent and isotropic; can be a Gaussian or Airy spot calculated from the system's optical parameters.

[0084] Subsequently, the algorithm enters an iterative loop. In the th iteration, the algorithm sequentially goes through all scan positions' indices . For the scan position corresponding to the current index , the following calculation steps are performed:

[0085] Step 302: Based on the sample estimate and the probe estimate of the th iteration, the current exit wavefield is calculated.

[0086] Step 303: A two-dimensional Fourier transform is applied to the exit wavefield , resulting in its complex amplitude distribution in reciprocal space (the detector plane) .

[0087] Step 304: Fourier domain constraints are applied. This step replaces the modulus of the calculated complex amplitude with the square root of the experimentally measured diffraction intensity data, while keeping its phase unchanged. This generates an updated reciprocal space complex amplitude . The modulus of its two polarization components are and , respectively.

[0088] Step 305: A two-dimensional inverse Fourier transform is applied to the updated , resulting in a corrected real-space exit wavefield .

[0089] Step 306: Update the estimates of the sample Jones matrix and the probe Jones vector based on the difference between the pre-corrected exit wave and the corrected exit wave, respectively. The update rule for the sample Jones matrix is as follows:

[0090] ;

[0091]

[0092] and are the estimates of the sample Jones matrix at the th and th iteration, respectively;

[0093] is the element-wise complex conjugate of the probe field ;

[0094] is the maximum value of the probe field intensity, as a normalization factor;

[0095] is a tiny positive number to prevent the denominator from being zero;

[0096] is a vector representing the difference between the corrected exit wave and the current model-predicted exit wave, which is the error signal driving the function update;

[0097] as a global term, is the operator that projects the above error signal back to the sample function domain;

[0098] The update rule for the probe Jones vector is as follows, synchronized with the sample update:

[0099] ;

[0100]

[0101]

[0102] is a preset real update step size parameter to control the convergence rate of the probe estimates;

[0103] is the estimate of the sample Jones matrix ​​​​​​​​​​the conjugate transpose of

[0104] is the Jones matrix of the sample the maximum of some norm (e.g. Frobenius norm) of

[0105] is the operator that projects the error signal back into the probe function domain.

[0106] The algorithm repeats steps 302-306 for all the indices of the scanning positions , and completes one full iteration . Subsequently, the algorithm proceeds to the next iteration (step 307) until a pre-defined convergence condition is met (step 308), e.g. the error function value between the calculated diffraction pattern and the experimentally measured diffraction pattern is smaller than a threshold value, or the number of iterations reaches an upper limit. When the iteration terminates, the final is the reconstructed high-resolution Jones matrix image of the sample surface (step 309).

[0107] Referring to the accompanying Figure 4 , Figure 4 is a schematic diagram of the adaptive critical area detection procedure according to an embodiment of the present application. The procedure is implemented by the dynamic wavefront modulation module 102 in collaboration with the central control and data processing unit 106, and its function is to selectively concentrate the illumination energy into specific areas containing potential surface defects, so as to improve the detection data quality of these areas.

[0108] The dynamic wavefront modulation module 102, e.g. a liquid crystal spatial light modulator, is arranged in the optical path between the coherent light source 101 and the plasmonic evanescent wave chip 103. The dynamic wavefront modulation module 102 contains a two-dimensional array of pixel elements, each of which is capable of imposing an independent, programmable phase delay on the wavefront of the light beam passing through it according to an electrical signal from the central control and data processing unit 106. By loading a complete two-dimensional phase map into the dynamic wavefront modulation module 102, the overall wavefront of the incident light beam can be accurately shaped.

[0109] The adaptive critical area detection procedure specifically comprises the following steps:

[0110] Step 401: System initialization, perform a global preliminary scan. In this stage, the central control and data processing unit 106 loads a pre-determined reference phase map into the dynamic wavefront modulation module 102, which is used to generate a standard illumination probe (e.g. a diffraction-limited focused spot). The system then performs a relatively low-resolution scan covering the entire area of the sample to be measured, and collects the corresponding far-field diffraction pattern data set;

[0111] Step 402: The central control and data processing unit 106 uses the data set collected in step 401 to perform a preliminary image reconstruction. The reconstruction process can employ the vectorial coherence diffractive imaging algorithm described above to generate a Jones matrix image of the sample with low resolution covering the whole scanned area .

[0112] Step 403: The central control and data processing unit 106 executes an automated region identification algorithm on the preliminary reconstructed Jones matrix image . The algorithm searches and locates one or more specific regions in the image according to pre-set numerical criteria. The criteria can include: the deviation of the amplitude or phase values of the Jones matrix elements in a region from the average value of the background region exceeds a pre-set threshold; or the values of the off-diagonal elements in a region (characterizing the polarization rotation or cross-talk) are greater than a pre-set threshold. The output of the algorithm is a set of coordinates of the identified specific regions;

[0113] Step 404: The central control and data processing unit 106 calculates a new, target-oriented phase map according to the set of region coordinates determined in step 403. The calculation process employs, for example, the Gerchberg-Saxton algorithm or the weighted Fourier transform algorithm, with the goal of generating a phase map that, when loaded onto the dynamic wavefront modulation module 102, can redistribute the energy of the incident light beam to form a non-uniform illumination field on the surface of the plasmonic evanescent wave chip 103, with high intensity at the coordinates of the identified specific regions and low intensity in other regions;

[0114] Step 405: The central control and data processing unit 106 loads the new phase map calculated in step 404 to the dynamic wavefront modulation module 102. Subsequently, the system initiates one or more targeted fine scans. In this scan, due to the concentration of illumination energy, the signal strength scattered from the specific regions of the sample surface is enhanced, resulting in higher signal-to-noise ratio of the diffraction patterns collected by the diffraction signal acquisition module 105 in the corresponding regions;

[0115] Step 406: The central control and data processing unit 106 uses the high signal-to-noise ratio diffraction pattern data collected in step 405 to perform the vectorial coherence diffractive imaging algorithm again to reconstruct one or more specific regions with high resolution and high precision, and finally outputs the Jones matrix images of these key regions to complete the accurate characterization of surface defects.

[0116] The scanning process performed on the sample surface in this invention, namely the process of changing the relative position between the structured evanescent wave field and the material sample 104a, can be achieved in a non-mechanical manner. This method avoids moving the sample or chip using mechanical components such as piezoelectric ceramic displacement stages, thereby eliminating the influence of mechanical vibration and drift on detection accuracy. Its implementation principle is based on the sensitivity of surface plasmon excitation conditions to incident light parameters.

[0117] See attached document Figure 2 As mentioned earlier, the excitation of surface plasmons requires strict satisfaction of the momentum matching condition, i.e., the wave vector component of the incident light in the direction of the interface plane. It must be equal to the surface plasmon wave vector supported by the interface. .

[0118] ;

[0119] The two items in this condition are determined by the following formulas:

[0120] ;

[0121] ;

[0122] in, The wavelength of light The refractive index of the substrate, Angle of incidence and are the dielectric constants of the metal and the dielectric, respectively.

[0123] The first non-mechanical scanning method is angle scanning. In this method, the wavelength of the coherent light source 101 is maintained. The central control and data processing unit 106 controls the incident angle of the beam incident on the plasmonic evanescent wave chip 103 by controlling a beam deflection device (e.g., an electrically controlled galvanometer) disposed in the optical path. Make small and precise changes. According to The expression for the angle of incidence tiny changes This will cause a change in its wave vector component in the interface plane direction. This change in wave vector space directly leads to the lateral displacement of the surface plasmon excitation location in real space. Therefore, the angle of incidence is controlled by programming the central control and data processing unit 106. Switching between a series of predetermined values ​​allows the structured evanescent wave field 204 to be driven to perform step-by-step or continuous scanning on the surface of the material sample 104a.

[0124] The second non-mechanical scanning method is wavelength scanning. In this method, the incident angle is maintained. Constant. This method requires the coherent light source 101 to be a wavelength-tunable laser. The dielectric constant of the metal thin film 202... Wavelength The function of the surface plasmon wave vector (i.e., material dispersion) is therefore the surface plasmon wave vector. For wavelength Sensitive. When the central control and data processing unit 106 sends a command to the coherent light source 101 to cause a slight change in its output wavelength... At that time, the original momentum matching conditions This will be disrupted. Under illumination from a converging beam with an angular distribution, changes in wavelength will cause the incident light satisfying the new momentum-matching condition to originate from different spatial portions of the beam, resulting in a lateral displacement of the surface plasmon excitation sites. The output wavelength is precisely controlled by the central control and data processing unit 106. The sequence can also realize the scanning motion of the structured evanescent wave field 204 on the sample surface.

[0125] In either approach, the central control and data processing unit 106 pre-calibrates the quantitative relationship between the change in incident angle or wavelength and the displacement of the evanescent wave field based on system parameters. During the scanning task, the central control and data processing unit 106 calculates the target angle or target wavelength required for each scanning step based on preset scanning path points, and issues control commands to the corresponding functional modules, thereby achieving precise, vibration-free scanning coverage of the specified path on the sample surface.

[0126] Furthermore, in embodiments equipped with the dynamic wavefront modulation module 102, non-mechanical scanning can also be achieved through this module. The central control and data processing unit 106 can dynamically calculate and load a series of phase maps onto the dynamic wavefront modulation module 102. Each phase map causes the incident beam to form a focused spot on the plasmonic evanescent wave chip 103, and the position of the spot moves as the phase map is switched. This wavefront modulation-based illumination point movement method also constitutes an implementation of the non-mechanical scanning of the present invention.

[0127] See attached document Figure 5 , Figure 5 This is a flowchart of a super-resolution optical detection method for material surface defects according to an embodiment of the present invention. A second aspect of the present invention provides a super-resolution optical detection method for material surface defects, which is applied to the super-resolution optical detection system of the aforementioned embodiment. The main steps of this method are as follows.

[0128] First, in step 501, the system uses the light beam emitted by the coherent light source 101 to excite the plasmonic evanescent wave chip 103 under certain excitation conditions, so as to form a structured evanescent wave field on the surface of the chip as a near-field illumination probe.

[0129] Next, in step 502, the sample surface is scanned by changing the relative position between the structured evanescent wave field and the material sample 104a placed in the near-field action range thereof. The change of the relative position can be realized by the aforementioned non-mechanical scanning mode (i.e. angle scanning or wavelength scanning). At each scanning position, the evanescent wave field interacts with the sample to generate an outgoing wave propagating to the far field, which carries the structural and physical property information of the sample at the position.

[0130] Then, in step 503, a set of far-field diffraction patterns generated in step 502 are collected by the far-field diffraction signal collection module 105. If the system is configured with polarization resolution, at least two diffraction intensity patterns of orthogonal polarization components are collected at each scanning position to form a corresponding data set.

[0131] Finally, in step 504, the central control and data processing unit 106 receives and processes the entire set of diffraction pattern data collected in step 503. The central control and data processing unit 106 first reconstructs the surface information of the material sample 104a, such as its high-resolution Jones matrix image, based on the data set by using a coherent diffraction imaging algorithm (such as the aforementioned vector coherent diffraction imaging algorithm). Subsequently, the central control and data processing unit 106 completes the detection, positioning and characterization of the material surface defects by analyzing the reconstructed surface information, such as the numerical distribution and variation of its matrix elements.

[0132] The execution steps of the super-resolution optical detection method shown in FIG. 1 will be described in detail below. Figure 5

[0133] Step S1: Forming a structured evanescent wave field

[0134] This step is the initialization and preparation phase of the system. The central control and data processing unit 106 sends instructions to the coherent light source 101 to set the center wavelength and power of the output light beam thereof. At the same time, the unit 106 sends instructions to the beam control element in the optical path to guide the light beam to the substrate 201 of the plasmonic evanescent wave chip 103 at a precisely calculated incident angle . The incident angle and wavelength ​The selection of the incident angle

[0135] Before performing the formal detection, the method can include a calibration sub-step: the system scans the incident angle or the wavelength while monitoring the reflected light intensity at the far end of the metal film 202. When a sharp drop in the reflected light intensity occurs, the corresponding angle or wavelength is the optimal surface plasmon excitation parameter. This parameter is stored in the central control and data processing unit 106 for subsequent steps.

[0136] Step S2: generating a far-field diffraction pattern

[0137] In this step, the system scans the material sample 104a placed on the sample positioning module 104. The central control and data processing unit 106 executes a series of illumination interaction processes according to the pre-set scanning path (e.g., a two-dimensional grid covering the area to be detected). For each position point on the scanning path identified by index , the central control and data processing unit 106 shifts the center of the structured evanescent wave field 204 to the corresponding coordinate on the sample surface by fine-tuning the incident angle or the wavelength . At this position, the evanescent wave field interacts with the sample in the near field to form an outgoing wave field that carries the complex transmission function information of the local area of the sample. This process is repeated at all pre-set scanning position points, and there is spatial overlap between the illumination areas of adjacent two scans to meet the data redundancy requirement of the coherent diffraction imaging algorithm.

[0138] Step S3: collecting the far-field diffraction pattern

[0139] This step is performed synchronously with step S2. For each scanning position identified by index , after the evanescent wave field interacts with the sample, the generated outgoing wave propagates to the far field and is captured by the diffraction signal acquisition module 105. The central control and data processing unit 106 drives the probe to move to this position (with the center coordinate ) immediately sends a synchronization trigger signal to the diffraction signal acquisition module 105. The detector array 105b integrates the received photons within a set exposure time according to the signal, recording the two-dimensional intensity distribution of the far-field diffraction pattern . If the system is configured with a polarization resolving unit 105a, the diffraction intensity patterns of the two orthogonal polarization components and are recorded simultaneously. After recording, the digitized intensity data along with its corresponding scanning position index are transmitted and stored in the memory of the central control and data processing unit 106. The acquisition process repeats with the scanning step until the data acquisition of all scanning positions is completed, forming a complete diffraction pattern dataset.

[0140] Step S4: Reconstruction and Defect Detection

[0141] In this step, the central control and data processing unit 106 utilizes its computing unit to perform offline processing on the complete diffraction pattern dataset acquired in step S3. First, the central control and data processing unit 106 executes the aforementioned vectorial coherent diffraction imaging algorithm. The algorithm takes the diffraction dataset as the amplitude constraint in the Fourier domain, and simultaneously solves for the complex amplitude distribution of the illuminating probe (structured evanescent wave field) and the two-dimensional Jones matrix of the material sample through iterative computation. The iteration process terminates upon meeting the convergence condition, outputting the final high-resolution Jones matrix image.

[0142] Subsequently, the central control and data processing unit 106 analyzes the reconstructed Jones matrix to complete defect detection. The analysis includes:

[0143] Defect Localization and Morphology Characterization: By analyzing the amplitude and phase of the diagonal elements of the Jones matrix. Pixel points with amplitudes lower than the average value of the surrounding area correspond to absorptive defects, and pixel points with different phases from the surrounding area correspond to defects with different refractive indices or thicknesses. The two-dimensional morphology and size of the defects can be directly obtained from the distribution and number of these pixel points.

[0144] Defect physical property characterization: By analyzing the magnitude of the off-diagonal elements of the Jones matrix. A non-zero off-diagonal element indicates that the sample causes polarization cross-talk at that point, and its magnitude is directly related to the birefringence or dichroism strength of the sample. By analyzing these values, the anisotropic physical properties of the defects can be characterized. For example, a physical scratch or groove usually appears as a continuous line in the reconstructed Jones matrix image, with low magnitude of the diagonal elements and a significant change in phase. A residual stress area, on the other hand, has no significant change in topography (the magnitude of the diagonal elements is close to 1), but the off-diagonal elements will exhibit non-zero values, and their distribution pattern is related to the distribution of the stress tensor. Finally, the system outputs quantitative detection results about the location, size, topography, and physical properties of the surface defects.

Claims

1. A super-resolution optical detection system for material surface defects, characterized in that, The system comprises: a coherent light source; a plasmonic evanescent wave chip comprising a metal film and a two-dimensional periodic nanostructure on the surface of the metal film, the plasmonic evanescent wave chip being configured to form a structured evanescent wave field on the surface of the plasmonic evanescent wave chip when excited by a light beam emitted by the coherent light source; a sample positioning module for placing a material sample to be detected within the range of action of the structured evanescent wave field; a diffraction signal acquisition module for acquiring a far-field diffraction pattern formed after the structured evanescent wave field interacts with the material sample; a central control and data processing unit configured to reconstruct surface information of the material sample by a coherent diffraction imaging algorithm based on a plurality of far-field diffraction patterns acquired at a plurality of relative positions between the structured evanescent wave field and the material sample, and to complete the detection of surface defects of the material by analyzing the surface information; the diffraction signal acquisition module comprises a polarization resolution unit configured to separate the far-field diffraction pattern into at least two orthogonal polarization components; the coherent diffraction imaging algorithm is a vector coherent diffraction imaging algorithm, which describes the interaction between the structured evanescent wave field and the material sample based on a mathematical model: ； wherein is a scan position index is an exit wavefield of the scan position identified by the scan position index is a Jones matrix of the material sample, is a vector probe field of the structured evanescent wavefield centered at is a two-dimensional real space coordinate vector, is a two-dimensional real space coordinate vector, is the scan position index is a corresponding probe center coordinate vector; the central control and data processing unit reconstructs the Jones matrix as the surface information via an iterative process by the vector coherent diffraction imaging algorithm. the iterative process comprises correcting the current estimate of the Jones matrix by a sample Jones matrix update rule: ； in, and The Jones matrix of the samples respectively In the Second and third The estimated value of the next iteration. It is the sample update step size parameter that controls the convergence speed; It is a tiny positive number set to prevent the denominator from being zero. probe field Element-level complex conjugation, It is the maximum value of the probe field intensity. The first emitted wave is calculated based on the current estimate. The second emitted wave is obtained by applying constraints from the far-field diffraction pattern to the first emitted wave; the iterative process further comprises synchronously correcting the current estimate of the vector probe field of the structured evanescent wave field by a probe field update rule: ； in, and These are for the vector probe field. In the Second and third The estimated value of the next iteration; It is the probe update step size parameter that controls the convergence speed; Estimates of the Jones matrix of the sample The conjugate transpose of; It is the sample Jones matrix The maximum value of a certain norm; the plurality of relative positions are changed by the central control and data processing unit fine-tuning the incident angle or wavelength of the coherent light source.

2. The system for super-resolution optical detection of surface defects of a material according to claim 1, wherein, The two-dimensional periodic nanostructure of the plasmonic evanescent wave chip is configured to interact with surface plasmons generated on the surface of the metal film, thereby forming the structured evanescent wave field on the surface of the plasmonic evanescent wave chip.

3. The system for super-resolution optical detection of surface defects of a material according to claim 1, wherein, Further comprising a dynamic wavefront modulation module arranged on the light path of the coherent light source for dynamically regulating the wavefront or amplitude distribution of the light beam emitted by the coherent light source.

4. The system for super resolution optical detection of surface defects of a material according to claim 3, wherein, The central control and data processing unit is further configured to identify one or more specific regions based on preliminary surface information obtained from a preliminary reconstruction, and control the dynamic wavefront modulation module to mainly distribute the illumination energy of the light beam emitted by the coherent light source on the one or more specific regions for subsequent focused detection.

5. A method for super-resolution optical detection of defects on a material surface, characterized in that A material surface defect super-resolution optical detection system according to any one of claims 1-4, comprising the following steps: S1, exciting a plasmonic evanescent wave chip by a coherent light source to form a structured evanescent wave field on the surface of the plasmonic evanescent wave chip; S2, generating a set of far-field diffraction patterns formed by the interaction between the structured evanescent wave field and a material sample by changing the relative position between the structured evanescent wave field and the material sample; S3, collecting a set of far-field diffraction patterns generated in the step S2; S4, first, based on the collected set of far-field diffraction patterns, reconstructing the surface information of the material sample by using a coherent diffraction imaging algorithm; then, detecting the surface defects by analyzing the surface information.

Images