A method and apparatus for ultra-low dose coherent diffraction imaging

Abstract

This application belongs to the field of coherent diffraction imaging technology, specifically disclosing an ultra-low dose coherent diffraction imaging method and apparatus. Based on a blind source separation strategy, this application uses principal component analysis to process the acquired diffraction signals, performing noise separation and updating in reciprocal space. This effectively separates mixed noise energy into noise components, thus avoiding crosstalk to the reconstruction process and significantly improving the convergence stability and robustness of coherent diffraction imaging when reconstructing diffraction signals with extremely low signal-to-noise ratios under ultra-low exposure doses. Simultaneously, this application constructs noise components separately for each scanning position and correlates noise components at different positions through low-dimensional spatial projection, achieving non-stationary noise separation. This enables more effective handling of random noise caused by the low quantum efficiency of ultra-short band detectors, thus maintaining extremely high noise robustness and reconstruction accuracy even under ultra-low exposure doses, achieving an effective improvement in resolution.

Description

Technical Field

[0001] This application belongs to the field of coherent diffraction imaging technology, and more specifically, relates to an ultra-low dose coherent diffraction imaging method and apparatus. Background Technology

[0002] Coherent diffraction imaging is a lensless quantitative phase imaging technique that uses algorithms to alternately project and reconstruct the complex amplitude information of the illumination probe and the sample under test between reciprocal space and real space. It boasts advantages such as simple equipment, ease of use, and no reliance on high-quality optical imaging devices. The lensless nature of coherent diffraction imaging makes it widely applicable to radiation sources such as extreme ultraviolet (EUV), X-rays, and electron beams, where it is difficult to fabricate high-quality imaging elements.

[0003] However, such ultra-short wavelength radiation sources, especially desktop ultra-short wavelength light sources, typically suffer from low flux and insufficient stability; simultaneously, some samples have low radiation dose tolerance. These factors severely limit the effective exposure dose to the sample surface, causing high-intensity noise to alias in the acquired signal, reducing the signal-to-noise ratio of the detector signal and limiting the system's effective numerical aperture. Furthermore, the low quantum efficiency of ultra-short wavelength detectors introduces significant time-varying random noise. These problems disrupt the consistency between the forward model in the algorithm and the actual physical process, easily causing iterations to get trapped in local optima or even fail to converge, resulting in limited resolution of the imaging system in ultra-low exposure dose scenarios.

[0004] To address this issue, researchers have proposed a series of improved algorithms to suppress noise crosstalk in ultra-low dose scenarios. For example, the least squares stacked diffraction imaging method based on maximum likelihood estimation (LSQML) uses a pre-defined noise model combined with maximum likelihood estimation to suppress the influence of noise during the iteration process, thus improving the noise robustness of imaging to some extent. However, it cannot inherently separate noise from the acquired signal, making it still unable to handle low signal-to-noise ratio signals under ultra-low exposure doses, resulting in limited resolution of the imaging system in ultra-low exposure dose scenarios. To solve the above problems, a stacked diffraction computational imaging method with real-time noise separation (publication number CN115201110A) has been proposed. By constructing a virtual noise probe with a wavelength different from the illumination probe, it adaptively decouples the signal and noise in real space, achieving robust imaging even when the signal-to-noise ratio of the diffraction signal decreases by two orders of magnitude. However, it requires the introduction of computationally expensive spatial transformation calculations between the reciprocal space and the real space, resulting in an exponential increase in imaging time overhead. Meanwhile, all of the above methods assume that the noise signal is stationary noise that does not change with time, and are only suitable for imaging in environments with stable lighting conditions. They cannot effectively handle non-stationary noise caused by complex imaging conditions such as light sources and detectors in ultra-short wavelength imaging, making it difficult for the imaging system to achieve high-resolution reconstruction in ultra-low exposure dose scenarios. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide an ultra-low dose coherent diffraction imaging method and apparatus, which aims to effectively improve the resolution of coherent diffraction imaging systems under ultra-low exposure doses and achieve high-resolution reconstruction.

[0006] To achieve the above objectives, in a first aspect, this application provides an ultra-low dose coherent diffraction imaging method, applied to a coherent diffraction imaging system, the method comprising: Step S1: Through initialization, obtain the complex amplitude of the illumination probe, the complex amplitude of the sample under test, and the guessed noise component in the first iteration—the complex amplitude of the illumination probe, the complex amplitude of the sample under test, and the guessed noise component in the first iteration are denoted as follows: , and As the object updated in the first iteration, the guessed noise component has the same dimension as the diffraction field acquired at different scanning positions; Step S2: Randomly select a scanning position, and based on the complex amplitude of the illumination probe and the complex amplitude of the sample to be tested, multiply the two to form an exit wave and input it into the forward diffraction propagation model to propagate it to the detector plane to obtain the guessed coherent diffraction components. Step S3: Based on the collected diffraction field light intensity, the guessed coherent diffraction components and the guessed noise components, the separated coherent diffraction components and the separated noise components are obtained through principal component analysis. Step S4: Based on the separated noise components, the updated guessed noise components are obtained in the reciprocal space by introducing a noise field update step size. Step S5: Repeat steps S2 to S4 to obtain the separated coherent diffraction components and the updated guessed noise components at all scanning positions. Step S6: Based on the separated coherent diffraction components obtained in step S5, the updated complex amplitude of the illumination probe and the updated complex amplitude of the sample under test are obtained through phase recovery optimization. The updated complex amplitude of the illumination probe and the updated complex amplitude of the sample under test in this iteration are used as the complex amplitude of the illumination probe and the complex amplitude of the sample under test in the next iteration, that is, as the objects to be updated in the next iteration. Step S7: Project the updated guessed noise components in the reciprocal space onto the dimension-reduced space to extract the common parts of the noise and perform a relaxation update to obtain the guessed noise components in the next iteration (i.e., the objects to be updated in the next iteration). Step S8, based on the guessed coherent diffraction components ( ), guessing noise components ( The system collects the diffraction field light intensity and determines whether the conditions for ending the iteration are met. If not, it continues to execute steps S2 to S7 to proceed to the next iteration.

[0007] In this application, ultra-low dose refers to doses lower than [amount missing]. The dosage.

[0008] It should be noted that at the end of the iteration, the updated complex amplitude of the illumination probe and the updated complex amplitude of the sample under test obtained in the current iteration are output as the final updated complex amplitude of the illumination probe. ( ) and the complex amplitude of the sample to be tested ( ).

[0009] It is understood that steps S3 and S4 of this application are based on a blind source separation strategy, using principal component analysis to process the acquired diffraction signal, and performing noise separation and updating in the reciprocal space to effectively separate the mixed noise energy into the noise component, thereby avoiding crosstalk to the reconstruction process. Compared to the least squares stacked diffraction imaging method based on maximum likelihood estimation, which can only suppress the contribution of noise energy to the reconstructed sample and probe based on assumed prior knowledge of the noise model, this application effectively avoids the interference of noise energy on reconstruction, significantly improving the convergence stability and robustness of coherent diffraction imaging when reconstructing diffraction signals with extremely low signal-to-noise ratios under ultra-low exposure doses; compared to the stacked diffraction computational imaging method that separates noise in real space in real time, this application effectively avoids the computationally expensive spatial transformation calculation between the reciprocal space and the real space, greatly improving imaging efficiency. Meanwhile, this application constructs a noise component for each scanning position separately, and associates the noise components at different positions through low-dimensional spatial projection in step S7 above, thereby achieving non-stationary noise separation. This enables the algorithm to more effectively handle random time-varying noise caused by the low quantum efficiency of ultra-short band detectors, thereby improving the algorithm's noise robustness and reconstruction accuracy, and enabling high-resolution reconstruction to be achieved even under ultra-low exposure doses.

[0010] As can be seen, this application separates the acquired diffraction signal into coherent diffraction components and noise components through principal component analysis, and constructs noise components separately for each scanning position and correlates them through low-dimensional spatial projection. This effectively isolates noise crosstalk and overcomes the limitations of stationary noise processing without the need for prior noise information, ultimately achieving highly robust imaging under ultra-low exposure doses.

[0011] In one possible implementation, step S3 above includes obtaining the separated coherent diffraction component and the separated noise component using the following formula: ; in, Indicates the principal component analysis method; Indicates the first The intensity of the diffraction field collected at each scanning position The sequence number representing the scan position. Represents the number of iterations. Represents reciprocal space coordinates; For includes and The set, Indicates the first In the nth iteration The guessed coherent diffraction components corresponding to each scanning position; Indicates the first In the nth iteration The guessed noise components corresponding to each scan position; For includes and The set, This represents the coherent diffraction components after separation. This represents the separated noise components.

[0012] In one possible implementation, step S4 above includes obtaining the updated guessed noise components in the reciprocal space using the following formula: ; in, The sequence number representing the scan position. Represents the number of iterations. This represents the guessed noise component updated in the reciprocal space. Indicates the first In the nth iteration The guessed noise component corresponding to each scan position Indicates the noise field update step size. The separated noise components.

[0013] In one possible implementation, step S7 above includes removing noise components from all scan positions using the following formula. dimensionality reduction to In 3D space, the common components of the noise are obtained. : ; in, The sequence number representing the scan position. Represents the number of iterations. Represents the common components of noise. This represents the set of noise components at all scan positions. , This represents the guessed noise component updated in the reciprocal space. express Dimensions This indicates dimensionality reduction operations (including but not limited to singular value decomposition, canonical correlation analysis, group factor analysis, etc.). , This indicates the total number of scan positions. ( The dimension is , The dimension is ).

[0014] For example, Specifically, it could be , This indicates a truncated singular value decomposition.

[0015] In one possible implementation, step S7 above involves a relaxation update to obtain the guessed noise components for the next iteration, including: Determine the relaxation factor and will As The corresponding weight item will As The corresponding weighting terms; Based on the weight term, and Perform a weighted summation to obtain the guessed noise components for the next iteration; in, Represents the number of iterations. This represents the set of noise components at all scan positions. , This represents the guessed noise component updated in the reciprocal space. , This indicates the total number of scan positions. This represents the common component of the noise.

[0016] In one possible implementation, the condition for ending the iteration is that the root mean square error is less than a threshold or the maximum number of iterations is reached.

[0017] Alternatively, the root mean square error can be calculated using the following formula: ; in, The sequence number representing the scan position. Represents the number of iterations. This represents the calculated root mean square error. Indicates the first In the nth iteration The guessed coherent diffraction components corresponding to each scan position Indicates the first In the nth iteration The guessed noise component corresponding to each scan position Indicates the first The intensity of the diffraction field collected at each scanning position Represents reciprocal space coordinates.

[0018] In a second aspect, this application provides an ultra-low dose coherent diffraction imaging device that applies the method described in the first aspect or any possible implementation of the first aspect, wherein the device uses at least one of the following as a radiation source: electron beam, X-ray, extreme ultraviolet light, ultraviolet light, visible light, infrared light, microwave and radio waves. The device is a transmission-type coherent diffraction imaging device or a reflection-type coherent diffraction imaging device. It includes, but is not limited to, conventional coherent diffraction imaging devices, stacked diffraction imaging devices, coherent modulation imaging devices, and multi-distance coherent diffraction imaging devices.

[0019] It should be noted that this application is applicable to different types of coherent diffraction imaging devices that use various electromagnetic waves as light sources, demonstrating strong applicability. This application aims to solve the problem of decreased imaging resolution under ultra-low exposure doses in coherent diffraction imaging. Traditional coherent diffraction imaging devices, stacked diffraction imaging devices, coherent modulation imaging devices, and multi-distance coherent diffraction imaging devices, all developed based on coherent diffraction imaging and using ultra-short wavelengths such as electron beams, X-rays, and EUV, as well as ultraviolet, visible, infrared, microwave, and radio wave illumination sources, all suffer from the inability to achieve high-resolution reconstruction under ultra-low exposure doses. Therefore, this application is applicable to all of them. Generally, the wavelength range of X-rays is 0.01 nm to 10 nm; the wavelength range of extreme ultraviolet (EUV) light is 10 nm to 100 nm; the wavelength range of ultraviolet light is 100 nm to 380 nm; the wavelength range of visible light is 380 nm to 750 nm; the wavelength range of infrared light is 750 nm to 1 mm; the wavelength range of microwaves is 1 mm to 1 m; and the wavelength range of radio waves is > 1 m.

[0020] It is understandable that the beneficial effects of the second aspect mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0021] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: This application employs a blind source separation strategy, utilizing principal component analysis to process the acquired diffraction signals. Noise separation and updating are performed in the reciprocal space, effectively separating mixed noise energy into noise components, thus avoiding crosstalk to the reconstruction process. Compared to least-squares stacked diffraction imaging methods based on maximum likelihood estimation, which can only suppress the contribution of noise energy to the reconstructed sample and probe based on assumed prior knowledge of the noise model, this application effectively avoids noise energy interference with reconstruction. This significantly improves the convergence stability and robustness of coherent diffraction imaging when reconstructing diffraction signals with extremely low signal-to-noise ratios under ultra-low exposure doses, thereby effectively enhancing the resolution of the coherent diffraction imaging system under ultra-low exposure doses.

[0022] Furthermore, this application constructs a separate noise component for each scanning position and correlates the noise components at different positions through low-dimensional spatial projection. Compared to existing stacked diffraction computational imaging methods that can only handle stationary noise in real time, this application achieves non-stationary noise separation, enabling it to more effectively handle random noise caused by the low quantum efficiency of ultra-short band detectors, thereby effectively improving the resolution of coherent diffraction imaging systems at ultra-low exposure doses.

[0023] Therefore, this application effectively reduces the requirement for sample exposure dose in coherent diffraction imaging systems, thereby reducing the need for high-throughput, high-stability light sources and low-noise, high-dynamic-range detectors, among other high-specification hardware. It effectively shortens the exposure time for coherent diffraction imaging, thus increasing imaging throughput. Furthermore, for biological samples susceptible to radiation damage, this application also promises to reduce the exposure dose to below the sample loss dose, thereby achieving non-destructive, high-resolution biological tissue imaging. In addition, this application is applicable across the entire electromagnetic spectrum, not only in the EUV band but also in imaging fields using X-rays, electron beams, and other extremely short wavelengths, as well as ultraviolet, visible, infrared, microwave, and electromagnetic wave bands. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the optical path of the coherent diffraction imaging system provided in the embodiments of this application; Figure 2 This is a schematic flowchart of the ultra-low dose coherent diffraction imaging method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the sample to be tested and the illumination probe provided in the embodiments of this application; Figure 4 These are diffraction patterns with different signal-to-noise ratios collected according to embodiments of this application; Figure 5 This is a schematic diagram of the reconstruction result using a diffraction signal with a signal-to-noise ratio of 40dB provided in an embodiment of this application; Figure 6This is a schematic diagram of the reconstruction result using a diffraction signal with a signal-to-noise ratio of 30dB, provided in an embodiment of this application. Figure 7 This is a schematic diagram of the reconstruction result using a diffraction signal with a signal-to-noise ratio of 20dB provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0025] In all the accompanying drawings, the same reference numerals are used to denote the same elements or structures, wherein: 1 is a desktop high-harmonic EUV light source, 2 is a filter module, 3 is an aperture, 4 is a multilayer film focusing mirror, 5 is the sample to be tested, and 6 is a vacuum EUV photon detector. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0028] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0029] The embodiments of this application are described below with reference to the accompanying drawings.

[0030] Example 1.

[0031] This application provides an ultra-low dose coherent diffraction imaging method, which is applied to a coherent diffraction imaging system and includes the following steps S21 to S27.

[0032] Step S21: Construct a coherent diffraction imaging system (or coherent diffraction imaging device), including but not limited to traditional coherent diffraction imaging systems, stacked diffraction imaging systems, coherent modulation imaging systems, and multi-distance coherent diffraction imaging systems. This application embodiment uses a desktop EUV transmission-reflection stacked diffraction imaging system as an example.

[0033] Figure 1This is a schematic diagram of the optical path of the desktop EUV coherent diffraction imaging system provided in this application embodiment. Figure 1 Part (a) is a desktop EUV transmission-type stacked diffraction imaging system. Figure 1 Part (b) is a desktop EUV reflective stacked diffraction imaging system. The stacked diffraction system is configured as follows: desktop high-harmonic EUV light source 1, filter module 2, aperture 3, multilayer film focusing mirror 4, sample under test 5, displacement stage (not shown), and vacuum EUV photon detector 6.

[0034] Step S22: Adjust the collimation system optical path. The EUV beam is emitted from a desktop high-harmonic EUV source. After the driving laser is filtered out by the filter module, the beam is modulated and shaped by the aperture and then converged onto the sample 5 by the multilayer film focusing mirror 4 to form an illumination probe. The sample 5 is placed near the back focal plane of the multilayer film focusing mirror 4 and is moved by the displacement stage for scanning. After the illumination probe interacts with the sample, the resulting exit wave propagates a distance in free space and reaches the target surface of the vacuum EUV photon detector 6. The vacuum EUV photon detector 6 collects the corresponding diffraction field intensity information. The angle and position of the vacuum EUV photon detector 6 are adjusted so that its target surface is perpendicular to the optical axis and the zero-order diffracted light is located in the center region of its target surface. A suitable filter module is selected according to actual needs to ensure that the diffraction field collected by the vacuum EUV photon detector has sufficient dynamic range. Meanwhile, the size of the aperture 3 or the distance between the multilayer film focusing mirror 4 and the sample 5 to be tested can be adjusted according to the experimental requirements to adjust the specific size of the illumination probe. Generally, the diameter of the illumination probe illuminating the sample 5 is 0.2mm~2mm.

[0035] Step S23: The displacement stage moves the sample 5 to be tested along the set trajectory (including but not limited to grating curves, concentric circle curves, Fermat curves, etc.) in the plane of the sample to be tested, while ensuring that there is more than 60% overlap between adjacent illumination probes, that is, satisfying the real space overlap constraint.

[0036] Step S24: At a certain distance after the sample 5, the vacuum EUV photon detector 6 records the intensity of a series of diffraction fields generated and propagated to the detector target plane after the illumination probe interacts with the sample 5 at each position. .in , The sequence number representing the scan position. This indicates the total number of scan positions. Represents the spatial coordinates of the reciprocal lattice.

[0037] Step S25: Obtain the diffraction field intensity information measured in step S24 Substitute into step S25 (e.g.) Figure 2 (As shown), separate noise components Simultaneously update the complex amplitude of the sample under test. Complex amplitude of illumination probe ,in, Represents real space coordinates.

[0038] Step S25 includes the following steps S25.1 to S25.7.

[0039] Step S25.1 Initialize noise components Complex amplitude of the sample to be tested and illumination probe complex amplitude . As the first In the nth iteration, the noise component is guessed, and in the nth iteration... In the next iteration Based on this, we update and guess the noise components. As the first The complex amplitude of the sample under test in the nth iteration, in the nth iteration In the next iteration The complex amplitude of the sample under test is updated based on this. As the first The complex amplitude of the illumination probe in the next iteration. In the next iteration The complex amplitude of the illumination probe is updated based on this. This can be achieved using, but is not limited to, random matrices, all-one matrices, or dark-field images acquired under no illumination as initial estimates of the noise components; it can also use, but is not limited to, random matrices and all-one matrices as initial estimates of the sample under test; and it can use, but is not limited to, circular light spots with uniform intensity, Gaussian light spots, and circular light spots with randomly distributed intensity as initial estimates of the illumination probe.

[0040] Step S25.2 in the In the next iteration ( (Initial value 0), randomly select the first... Each scan position is calculated. corresponding areas and The output wave after action : ; in, Indicates the first In the nth iteration The exit wave at each scan position, Indicates the first Each scan position relative to real space coordinates The translation vector, Indicates the first The sample to be tested in the next iteration Complex amplitude at each scan position, Indicates the first Complex amplitude of the illumination probe in the next iteration.

[0041] Step S25.3, the exit wave obtained in step S25.2 The input forward diffraction propagation model propagates to the plane containing the detector target surface (reciprocal space), yielding the predicted coherent diffraction components. : ; in, Indicates the first In the nth iteration The exit wave at each scan position, Indicates the first In the nth iteration The guessed coherent diffraction components at each scan position. Forward diffraction propagation models of light fields in free space, including but not limited to Fresnel diffraction propagation models, Fraunhofer diffraction propagation models, and angular spectrum propagation models.

[0042] Step S25.4 involves performing principal component analysis on the acquired diffraction signal in reciprocal space to update coherent diffraction separation and noise separation. The principal component analysis method that can be used is... This includes, but is not limited to, coherent mode decomposition, Karhunen-Loève transform, and low-rank sparse decomposition. In this embodiment, principal component analysis using coherent mode decomposition is taken as an example. The predicted coherent diffraction components obtained in step S25.3 are used as a basis. Guessing noise components The diffraction pattern acquired in step S24 Constructing a minimum density functional analysis of diffraction signals in reciprocal space : ; in, Indicates the first The intensity of the diffraction field collected at each scanning position Indicates the first In the nth iteration The guessed coherent diffraction components at each scan position. No. In the nth iteration The guessed noise component corresponding to each scan position. These are the separated coherent diffraction components. The separated noise components are represented. The minimum density functional analysis is solved using the gradient descent method. The separated coherent diffraction components were obtained. and the separated noise components .

[0043] Step S25.5: Utilize the separated noise components obtained in step S25.4 Update noise components within the reciprocal space. : ; in, This represents the guessed noise component updated in the reciprocal space. Indicates the first In the nth iteration The guessed noise component corresponding to each scan position Indicates the noise field update step size. The separated noise components.

[0044] Step S25.6 Randomly select the next scan position, and repeat steps S25.2 to S25.5 until all scans are completed. The coherent diffraction components separated at each scanning position ,all Updated guessed noise components at each scan location .

[0045] Step S25.7: The separated coherent diffraction components obtained in step S25.6 are... Input phase recovery optimization problem: ; ; in, Indicates the first In the nth iteration The exit wave at each scan position, For the updated export wave, , This refers to the reverse diffraction propagation model of the light field in free space (including but not limited to Fresnel diffraction propagation model, Fraunhofer diffraction propagation model, angular spectrum propagation model, etc.). Indicates the first Complex amplitude of the illumination probe in the next iteration Indicates the first The sample to be tested in the next iteration Complex amplitude at each scan position. For the first The updated complex amplitude of the illumination probe in the next iteration. Indicates the first In the nth iteration The updated complex amplitude of the sample under test at each scanning position. and These represent the reciprocal space amplitude constraint and the real space overlap constraint, respectively. The two phase retrieval optimization problems are solved using algorithms such as gradient descent and differential mapping to update the complex amplitudes of the sample under test and the illumination probe. , As the first The complex amplitude of the sample under test and the complex amplitude of the illumination probe in the nth iteration, in the nth iteration In the next iteration The amplitude of the sample under test and the illumination probe is updated based on this.

[0046] Step S26: Update all position noise components obtained in step S25.5. Dimensionality reduction is performed to extract common components of the noise to constrain noise updates. Dimensionality reduction methods that can be used include, but are not limited to, singular value decomposition (SVD), canonical correlation analysis (CAN), and group factor analysis (CFA). This embodiment below uses SVD as an example to illustrate dimensionality reduction.

[0047] Step S26 specifically includes the following steps (e.g.) Figure 2 (as shown) Step S26.1: Update all position noise components obtained in step S25.5. Input truncated singular value decomposition algorithm, from Projecting 3D space to 3D space, in which : ; in, This indicates a truncated singular value decomposition. It represents the noise components in low-dimensional space, and represents the common components of noise; .

[0048] Step S26.2, based on the common components of the noise in step S26.1 Relax update noise components : ; in, This represents the noise component after relaxation and update. This represents the relaxation factor. As the first The guessed noise component in the nth iteration. In the next iteration Based on this, we update and guess the noise components.

[0049] Step S27 calculates the predicted diffraction field distribution and the measured diffraction pattern at each scanning position. The root mean square error (RMSE) is considered when the RMS error is less than a set threshold or reaches the preset maximum number of iterations. Figure 2 As shown, the output and The final updated complex amplitude of the test sample and illumination probe complex amplitude That is, the reconstruction result provided by the method of this application.

[0050] Example 2.

[0051] Figure 3 These are the amplitude and phase patterns of the sample under test and the illumination probe used in the simulation process provided in this application embodiment. In the simulation experiment, a desktop high-harmonic EUV light source emits a Gaussian beam with a wavelength of 13.5 nm and a beam diameter of 128 pixels. After passing through a 2.5 mm focal length lens, the beam propagates 5 mm to the plane of the sample under test to form the illumination probe. Its amplitude and phase are as follows: Figure 3 As shown in (c) and (d), the sample size is 256×256 pixels, and its amplitude is as follows. Figure 3 The Siemens star shown in (a) has the following phase: Figure 3 The target resolution is shown in (b). The sample under test is moved by a displacement stage to scan a 10×10 grid, with an interval of 20 pixels between adjacent scan positions. A certain random offset is added to each position in the scan grid to avoid periodic artifacts. A detector with 256×256 pixels is set at 60 mm behind the sample to collect the diffraction fields corresponding to all 100 scan positions.

[0052] Figure 4 These are diffraction fields acquired by the stacked diffraction imaging system provided in this application embodiment under different exposure doses. The diffraction field unaffected by noise contamination is as follows: Figure 4 As shown in (a). To simulate the effects of photon shot noise, dark current noise, and readout noise during actual data acquisition, Poisson noise, Gaussian noise, and background noise were added to the signal, respectively. A signal-to-noise ratio of 40 dB was obtained under different exposure doses (e.g., ...). Figure 4 (as shown in (b)), 30dB (as shown in (b)) Figure 4 (as shown in (c)), 20dB (as shown in (c)) Figure 4 The diffraction field is shown in (d). The acquired diffraction pattern is input into the proposed ultra-low dose coherent diffraction imaging method. A Gaussian spot is used as the initial illumination probe guess, and a random matrix is ​​used as the initial sample guess for multiple rounds of iterative iteration.

[0053] The first set of simulation experiments was carried out using a diffraction field signal with a signal-to-noise ratio of 40dB. At the same time, the traditional LSQML method was compared with the ultra-low dose coherent diffraction imaging method proposed in this application. Figure 5 This application provides a method for reconstructing the complex amplitude of the sample under test and the illumination probe using a diffraction signal with a signal-to-noise ratio of 40dB through stacked diffraction. Figure 5 Images (a) to (d) represent the reconstructed amplitude and phase images of the test sample and illumination probe after 500 iterations using the LSQML method, respectively. The results show that although the LSQML method can successfully reconstruct the test sample and illumination probe after 500 iterations, the reconstructed sample becomes blurred due to noise artifacts. In particular, the line pairs of the first group of elements with the smallest linewidth in the reconstructed phase image of the test sample are no longer clearly distinguishable. Simultaneously, significant noise is present in the reconstructed illumination probe. This indicates that the traditional LSQML method cannot separate the noise energy in the diffraction field, causing noise to couple into the reconstruction results and limiting image clarity. Figure 5 Images (e) through (h) represent the reconstructed amplitude and phase images of the sample under test and the illumination probe after 500 iterations of the proposed ultra-low-dose coherent diffraction imaging method. The results show that the proposed ultra-low-dose coherent diffraction imaging method successfully eliminates noise artifacts in the reconstruction results, allowing the first group of elements with the smallest linewidth in the phase image of the sample under test to be clearly distinguished. Experiments demonstrate that the method of this application significantly improves imaging resolution while maintaining structural integrity.

[0054] The second set of experiments was conducted using a diffraction field signal with a signal-to-noise ratio of 30 dB, and the traditional LSQML method was compared with the ultra-low dose coherent diffraction imaging method proposed in this application. Figure 6 This application provides a method for reconstructing the complex amplitude of the sample under test and the illumination probe using a diffraction signal with a signal-to-noise ratio of 30dB through stacked diffraction. Figure 6 Images (a) to (d) represent the amplitude and phase images of the sample under test and the illumination probe reconstructed after 500 iterations of the LSQML method, respectively. The results show that the LSQML method experiences a severe performance degradation under noise enhancement, barely reconstructing the approximate outline of the Siemens star while completely losing internal details; significant speckle noise appears in the reconstructed probe, and its Gaussian structure is disrupted. In contrast, Figure 6 Images (e) through (h) represent the reconstructed amplitude and phase images of the sample and illumination probe after 500 iterations of the proposed ultra-low-dose coherent diffraction imaging method. The results show that the proposed ultra-low-dose coherent diffraction imaging method successfully separates noise components from the diffraction signal, enabling stacked diffraction to clearly reconstruct the complex amplitude information of the sample and illumination probe. The line pairs with the smallest linewidth in the sample phase are still clearly identifiable. These results demonstrate that the proposed method exhibits excellent robustness even under low signal-to-noise ratio conditions.

[0055] The third set of experiments was conducted using a diffraction field signal with a signal-to-noise ratio of 20 dB. The traditional momentum-accelerated stacked diffraction iteration engine (LSQML method) was used for comparison with the ultra-low dose coherent diffraction imaging method proposed in this application. Figure 7 This application provides a method for reconstructing the complex amplitude of the sample under test and the illumination probe using a diffraction signal with a signal-to-noise ratio of 20dB through stacked diffraction. Figure 7 Images (a) through (d) represent the reconstructed amplitude and phase images of the test sample and illumination probe after 500 iterations of the LSQML method, respectively. The results show that the LSQML method fails to converge completely when the signal-to-noise ratio is only 20 dB. In contrast, Figure 7 Images (e) through (h) represent the reconstructed amplitude and phase images of the sample under test and the illumination probe after 500 iterations of the ultra-low-dose coherent diffraction imaging method proposed in this application. These images are very similar to the reconstruction results in the first set of experiments, and the smallest elements in the first group can still be distinguished. These results demonstrate that traditional methods completely fail when noise is increased by a factor of 100. Under the same conditions, the method proposed in this application maintains extremely high noise robustness and reconstruction accuracy.

[0056] In summary, this application divides the acquired diffraction signal into coherent diffraction components and noise components through principal component analysis. This eliminates the need for prior noise information, effectively isolating noise energy from crosstalk in imaging and significantly improving the algorithm's noise robustness. Furthermore, this application constructs a separate noise component for each scanning position and correlates noise components from different positions through low-dimensional spatial projection, overcoming the limitation of traditional methods that can only handle stationary noise that does not change over time. Compared to existing technologies, the method and apparatus proposed in this application can significantly reduce the exposure dose requirements for coherent diffraction imaging and greatly reduce the reliance on high-throughput, high-stability light sources and low-noise, high-dynamic-range detectors, among other high-performance hardware.

[0057] The ultra-low dose coherent diffraction imaging device provided in this application is described below. The ultra-low dose coherent diffraction imaging device described below can be referred to in correspondence with the ultra-low dose coherent diffraction imaging method described above.

[0058] This application also provides an ultra-low dose coherent diffraction imaging device, which applies any of the above-mentioned ultra-low dose coherent diffraction imaging methods, and the device uses at least one of the following as a radiation source: electron beam, X-ray, extreme ultraviolet light, ultraviolet light, visible light, infrared light, microwave and radio waves; The device is either a transmission coherent diffraction imaging device or a reflection coherent diffraction imaging device.

[0059] It is understood that the detailed functional implementation of each of the above units / modules can be found in the description in the aforementioned method embodiments, and will not be repeated here.

[0060] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0061] Based on the methods in the above embodiments, this application provides an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute the methods in the above embodiments.

[0062] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0063] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0064] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0065] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0066] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0067] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0068] It is understood that the various numerical designations used in the embodiments of this application are merely for the convenience of description and are not intended to limit the scope of the embodiments of this application.

[0069] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for ultra-low dose coherent diffraction imaging, characterized in that, include: Step S1: Through initialization, obtain the complex amplitude of the illumination probe, the complex amplitude of the sample to be tested, and the guessed noise component in the first iteration; Step S2: Randomly select a scanning position, and based on the complex amplitude of the illumination probe and the complex amplitude of the sample to be tested, multiply the two to form an exit wave and input it into the forward diffraction propagation model to propagate it to the detector plane to obtain the guessed coherent diffraction components. Step S3: Based on the collected diffraction field light intensity, the guessed coherent diffraction components and the guessed noise components, the separated coherent diffraction components and the separated noise components are obtained through principal component analysis. Step S4: Based on the separated noise components, the updated guessed noise components are obtained in the reciprocal space by introducing a noise field update step size. Step S5: Repeat steps S2 to S4 to obtain the separated coherent diffraction components and the updated guessed noise components at all scanning positions. Step S6: Based on the separated coherent diffraction components obtained in step S5, the updated complex amplitude of the illumination probe and the updated complex amplitude of the sample under test are obtained through phase recovery optimization. Step S7: Project the updated guessed noise components in the reciprocal space onto the dimension-reduced space to extract the common parts of the noise and perform a relaxation update to obtain the guessed noise components in the next iteration. Step S8: Based on the guessed coherent diffraction component, the guessed noise component, and the collected diffraction field light intensity, determine whether the condition for ending the iteration is met. If not, continue to execute steps S2 to S7 to proceed to the next iteration.

2. The ultra-low dose coherent diffraction imaging method according to claim 1, characterized in that, Step S3 includes obtaining the separated coherent diffraction component and the separated noise component using the following formula: ； in, Indicates the principal component analysis method; Indicates the first The intensity of the diffraction field collected at each scanning position The sequence number representing the scan position. Represents the number of iterations. Represents reciprocal space coordinates; For includes and The set, Indicates the first In the nth iteration The guessed coherent diffraction components corresponding to each scanning position; Indicates the first In the nth iteration The guessed noise components corresponding to each scan position; For includes and The set, This represents the coherent diffraction components after separation. This represents the separated noise components.

3. The ultra-low dose coherent diffraction imaging method according to claim 1, characterized in that, Step S4 includes obtaining the updated guessed noise components in the reciprocal space using the following formula: ； in, The sequence number representing the scan position. Represents the number of iterations. This represents the guessed noise component updated in the reciprocal space. Indicates the first In the nth iteration The guessed noise component corresponding to each scan position Indicates the noise field update step size. The separated noise components.

4. The ultra-low dose coherent diffraction imaging method according to claim 1, characterized in that, Step S7 includes using the following formula to extract all noise components from the scanned locations. dimensionality reduction to In 3D space, the common components of the noise are obtained. : ； in, The sequence number representing the scan position. Represents the number of iterations. Represents the common components of noise. This represents the set of noise components at all scan positions. , This represents the guessed noise component updated in the reciprocal space. This indicates a dimensionality reduction operation. , This indicates the total number of scan positions. .

5. The ultra-low dose coherent diffraction imaging method according to claim 1, characterized in that, In step S7, a relaxation update is performed to obtain the guessed noise component in the next iteration, including: Determine the relaxation factor and will As The corresponding weight item will As The corresponding weighting terms; Based on the weight term, and Perform a weighted summation to obtain the guessed noise components for the next iteration; in, Represents the number of iterations. This represents the set of noise components at all scan positions. , This represents the guessed noise component updated in the reciprocal space. , This indicates the total number of scan positions. This represents the common component of the noise.

6. An ultra-low dose coherent diffraction imaging device, characterized in that, When using the ultra-low dose coherent diffraction imaging method as described in any one of claims 1-5, the device employs at least one of the following as a radiation source: electron beam, X-ray, extreme ultraviolet light, ultraviolet light, visible light, infrared light, microwave, and radio waves; The device is a transmission coherent diffraction imaging device or a reflection coherent diffraction imaging device.

Images