Dynamic synchronization correction system and method for incident light monitoring based on annular scintillation crystal

Abstract

The application provides an incident light monitoring dynamic synchronous correction system and method based on a ring-shaped scintillation crystal, which comprises a ring-shaped scintillation crystal coaxially arranged with a light source, a sample experimental device, and a transmitted light photon counter, an off-axis photon counter deviated from the optical axis and aligned with the side of the ring-shaped scintillation crystal facing the sample experimental device, a trigger acquisition board connected with the transmitted light photon counter and the off-axis photon counter, and an upper computer connected with the trigger acquisition board. The system of the application places the ring-shaped scintillation crystal on the main optical axis in the experimental cavity, and the ring directly receives X-rays with the same energy and the same path as the experimental light. The experimental light passes through the center channel of the ring-shaped scintillation crystal, and the outer ring collects signals with the same energy and the same path, thereby ensuring the energy consistency of the reference light and the experimental light and overcoming the problem of unstable correlation of different energy light drift.

Description

Technical Field

[0001] This invention belongs to the field of synchrotron radiation X-ray intensity monitoring and dynamic correction technology, specifically relating to an incident light monitoring dynamic synchronous correction system and method based on a ring scintillation crystal, which is suitable for high-precision, sub-millisecond time-resolved experiments in scanning transmission microscopy (STXM). Background Technology

[0002] Synchrotron radiation is electromagnetic radiation emitted by charged particles moving along curved orbits under the influence of electromagnetic fields. The emitted radiation has a wide range of wavelengths, covering wavelengths from far-infrared to X-rays.

[0003] With the development of synchrotron radiation technology, from the first-generation to the third-generation synchrotron radiation sources, photon flux has increased by several orders of magnitude. This has provided the possibility of using new methods in many physics experiments, two of which are sub-millisecond fast absorption spectroscopy (QXAFS) and soft X-ray scanning transmission microscopy (STXM). Existing synchrotron radiation source facilities in China include the Beijing Synchrotron Radiation Facility (BSRF), the National Synchrotron Radiation Laboratory (NSRL) in Hefei, and the Shanghai Synchrotron Radiation Facility (SSRF). Among them, the SSRF is the only third-generation source, and its BL08U beamline is the only beamline in China capable of performing soft X-ray spectroscopic microscopy.

[0004] However, due to factors such as electron orbital drift in the storage ring, beam energy attenuation, and electron injection, experiments using synchrotron radiation sources can experience light intensity fluctuations or even rapid attenuation, leading to instability in the generated X-ray intensity. For example, in actual operation, the Shanghai Synchrotron Radiation Facility (SSRF) injects electron bundles when the beam intensity decreases to maintain a constant beam intensity. This injection causes significant fluctuations in X-ray intensity. The frequency of electron bundle injection is 2 Hz, and the number of injected electron bundles can be 12, 24, or 36 or even more. Sometimes, bundle injection can cause the X-ray intensity to decrease by 60% within 1 ms, which can severely impact beamstations with data acquisition at the sub-millisecond level. Furthermore, slow drift of the storage ring beam orbit and fluctuations in cooling water temperature can cause low-frequency band-like fluctuations in X-ray intensity (band noise), resulting in large areas of bright and dark fringes in the imaging background. Therefore, a high-speed and accurate light intensity monitoring and correction system is needed to eliminate this effect.

[0005] Currently, the mainstream light intensity monitoring systems at home and abroad can be divided into gas ionization chamber type, wire scanning type, diamond blade type, diamond surface type and semiconductor type. These methods all have certain problems in the light intensity measurement of soft X-rays.

[0006] Some beamlines at the Shanghai Synchrotron Radiation Facility (SSRF) employ filamentary scanning detection. This method collects an approximate number of photons through the photoelectric effect by placing a filamentary detector within the optical path. However, this method cuts the beam shape, affecting the spot shape and hindering subsequent experiments.

[0007] The Beijing Synchrotron Radiation Facility (BSRF) employs a gas ionization chamber detection method. This method uses the interaction between X-rays and the target gas in the ionization chamber to ionize the target gas and create a potential difference within the chamber, thereby calculating the X-ray intensity. However, this method absorbs most of the soft X-rays and cannot be used for intensity measurement at the front end of the experimental device.

[0008] The Pohang Light Source (PLS) in South Korea uses a diamond blade detector for intensity acquisition, while the National Synchrotron Radiation Facility (NSLS) in the United States uses a diamond surface detector. Both methods utilize the physical properties of diamond and the photoelectric effect to accurately measure the intensity of soft X-rays. However, the diamond blade detector requires a diamond blade to cut into the beam, causing beam deformation. The diamond surface detector absorbs some X-rays, resulting in a decrease in beam intensity.

[0009] The semiconductor detection method uses a silicon edge detector to form a through-hole at the light-emitting slit, and monitors the light intensity by monitoring the intensity of the interaction between the X-ray and the detector. This method reduces the intensity of the beamline and also affects the size of the light spot to some extent.

[0010] The PMT photon counter primarily utilizes single-photon detection technology using photomultiplier tubes. The photomultiplier tubes record a single photon pulse, and the data is acquired through developed hardware and software. Finally, the electronic counter filters the data to measure the number of photons per unit time. The PMT photon counter's ability to detect single photons increases the accuracy and real-time performance of light intensity monitoring.

[0011] However, if the PMT photon counter is placed in the optical path in front of the experiment, it will directly block the transmission of light and will not be able to measure the high-speed changes in light intensity in real time during the experiment. Therefore, it cannot be placed in the optical path at the front of the experimental device for real-time high-speed monitoring.

[0012] One existing online X-ray intensity real-time high-speed monitoring system (such as patent document CN202410129592.0) addresses the issue of the photon counter (PMT) not blocking light by placing a YAG light-emitting device at the front end of the light-blocking slit. This converts the X-rays blocked at the slit edge into visible light, which is then collected by the photon counter of the reflected light PMT as a reference signal. Simultaneously, a trigger acquisition board enables nanosecond-level synchronous triggering of two PMTs. A mean filtering module is used to locate the electron beam injection moment and perform mean filtering on the downstream light intensity. Finally, the transmitted light signal is linearly normalized using the reference light signal. However, existing online X-ray intensity real-time high-speed monitoring systems still suffer from the following inherent and insurmountable technical challenges:

[0013] (1) The fundamental correction error caused by the difference in energy between the reference light and the experimental light: The reference light of the existing online X-ray intensity real-time high-speed monitoring system is taken from the part of X-rays discarded at the edge of the slit. This part of the X-rays is stray light of non-experimental target energy generated after the beam is split by the grating monochromator, and its propagation direction is different from that of the experimental light. The experimental light is the target energy X-ray that propagates along the optical axis after passing through the slit opening. Since the response characteristics of X-rays of different energies in the storage ring, such as orbital drift and beam jitter, are fundamentally different, the intensity change trend of the non-target energy reference light signal is not strictly consistent with that of the experimental light, and this correlation is extremely unstable—sometimes correlated and sometimes uncorrelated. The root cause is that the two are light of different energies. This "non-identical energy" reference makes simple linear normalization unable to achieve effective correction, and the residual correction error is large and unpredictable. The physical root cause is that X-rays of different energies emitted from the grating monochromator have different orbital response functions in the storage ring. When the beam drifts, the jitter amplitude and phase of different energy lights in space are not the same. Therefore, in existing online X-ray intensity real-time high-speed monitoring systems, there is no definite mapping relationship between "edge non-target energy light" and "center target energy light". Correction based on this relationship is essentially normalizing an unrelated signal and cannot effectively eliminate the intensity fluctuations of the experimental light itself.

[0014] (2) Inability to detect and separate stripe noise: Existing online X-ray intensity real-time high-speed monitoring systems can only identify and mean-filter short-term peak attenuation (injection noise) caused by electron beam injection using a fixed threshold method. For low-frequency long-term stripe drift (stripe noise) caused by slow beam trajectory drift, since it manifests as a slowly varying baseline in the time domain, existing online X-ray intensity real-time high-speed monitoring systems lack effective baseline extraction and separation methods, resulting in large areas of bright and dark stripe artifacts in the imaging results.

[0015] (3) Linear normalization ignores nonlinear response: Existing online X-ray intensity real-time high-speed monitoring systems use simple linear division normalization. However, experiments show that the response of the reference PMT channel and the transmission PMT channel to changes in light intensity is not a strictly linear 1:1 relationship due to factors such as the difference in YAG crystal thickness and photomultiplier tube gain. Simple linear normalization introduces new correction distortion when the light intensity fluctuates significantly. Summary of the Invention

[0016] The present invention aims to provide a dynamic synchronous correction system and method for incident light monitoring based on a ring scintillation crystal, in order to solve the problems of correction failure caused by the difference in energy between the reference light and the experimental light, the inability to detect stripe noise, and linear correction distortion in the prior art.

[0017] To achieve the above objectives, the present invention provides an incident light monitoring dynamic synchronous correction system based on a ring scintillation crystal, comprising a ring scintillation crystal coaxially arranged with the light source, a sample experimental device and a transmitted light photon counter, an off-axis photon counter offset from the optical axis and aligned with the side of the ring scintillation crystal facing the sample experimental device, a trigger acquisition board connected to both the transmitted light photon counter and the off-axis photon counter, and a host computer connected to the trigger acquisition board.

[0018] The host computer is equipped with a reading module, a display module, a dual-scale median filtering module, a MAD threshold detection module, and a hierarchical multiplication correction module. The dual-scale median filtering module is configured to perform small-window median filtering and large-window median filtering on the reference light signal acquired by the off-axis photon counter to obtain the filtering results. The MAD threshold detection module is configured to process the filtering results of the small-window and large-window median filtering to obtain a relative residual sequence, detect short-time injection events based on the relative residual sequence to generate an injection noise mask, and detect baseline offset regions based on the filtering results of the large-window median filtering to obtain a stripe noise mask. The hierarchical multiplication correction module is configured to obtain injection region correction factors and stripe region correction factors based on the injection noise mask and the stripe noise mask, and perform hierarchical power-law correction on the light intensity of the transmitted light signal acquired by the transmitted light photon counter based on the injection region correction factors and the stripe region correction factors to obtain the correction result of the transmitted light signal.

[0019] The small window has a length of 3 points, and the large window has a length of 51 points. The MAD threshold detection module divides the filtering results of the small window mid-range filtering and the large window mid-range filtering to obtain the relative residual sequence, and calculates the median m and MAD value σ of the relative residual sequence. The injected noise mask is detected with m+4σ+4%·|m| as the entry threshold, m+σ+1.5%·|m| as the exit threshold, and the number of sustained points is between 2 and 20. After replacing or deleting the mask area of ​​the banner noise, the large window mid-range filtering is performed again to obtain the updated slowly changing baseline. To update the slowly changing baseline The criteria for detecting banner noise are: an offset of ≥4% relative to the median of the entire segment and a duration that meets the condition.

[0020] In the layered multiplication correction module, the injection region correction factor is used to recover the mask for injected noise.

[0021] Wherein, the injection region correction factor f_event=η^(-γ), where η is the ratio of the mean of the reference optical signal within the event to the baseline mean of the reference optical signal before the event, and γ is the experimentally calibrated injection correction power law exponent.

[0022] For masking banner noise, a continuously varying banner correction factor is used for correction;

[0023] Wherein, the stripe correction factor f_stripe[n]=(B_global / I_ref_large[n])^α, B_global is the baseline mean of the reference optical signal during the noise-free period, I_ref_large[n] is the baseline of the large window mid-range filter, n is the sampling point number in the time series, and α is the experimentally calibrated stripe correction power law exponent;

[0024] The total correction factor is the product of the injection region correction factor f_event and the stripe correction factor f_stripe[n].

[0025] The sample experimental apparatus includes a zone plate, a graded aperture, and a sample arranged sequentially along the X-ray propagation direction; the ring scintillation crystal is located on the optical axis about 50 mm upstream of the sample experimental apparatus, and receives X-rays with the same energy as those received by the sample experimental apparatus and converts them into visible light; the inner wall diameter of the central channel of the ring scintillation crystal is larger than the diameter of the zone plate in the sample experimental apparatus, so that the central region of the X-rays passes through the central channel of the ring scintillation crystal without loss and continues to propagate along the optical axis.

[0026] The ring scintillation crystal, the sample experimental apparatus, and the transmitted light photon counter are installed in the same experimental chamber, which has a vacuum environment. The transmitted light photon counter is used to collect X-rays that pass through the sample, convert them into visible light, and count them as the transmitted light signal passing through the sample. The off-axis photon counter is set at a position off the optical axis, with its photocathode aligned with the side of the ring scintillation crystal facing the sample experimental apparatus, to collect and count the visible light converted by the ring scintillation crystal as a reference light signal.

[0027] The trigger acquisition board is configured to provide trigger signals indicating the start and end of counting, so as to synchronously trigger the transmitted light photon counter and the off-axis photon counter to perform synchronous integration counting.

[0028] The material of the ring scintillation crystal 1 is YAG, LuAG:Ce or GAGG.

[0029] On the other hand, the present invention provides a dynamic synchronization correction method for incident light monitoring based on a common-path ring scintillation crystal, comprising:

[0030] S0: Build the incident light monitoring dynamic synchronous correction system based on the ring scintillation crystal described above, and set the system parameters in the host computer;

[0031] S1: The trigger acquisition board synchronously triggers the transmitted light photon counter and the off-axis photon counter to acquire the transmitted light signal and the reference light signal;

[0032] S2: Use the dual-scale median filtering module of the host computer to perform small window median filtering and large window median filtering on the reference optical signal to obtain the filtering result;

[0033] S3: Using the MAD threshold detection module of the host computer, the relative residual sequence is obtained by processing the filtering results of small window mid-range filtering and large window mid-range filtering; short-term injection events are detected based on the relative residual sequence to generate a mask for injected noise; and the baseline offset region is detected based on the filtering results of large window mid-range filtering to obtain a mask for banner noise.

[0034] S4: For the transmitted light signal, a layered multiplication correction module is used to obtain the injection region correction factor and the stripe region correction factor based on the injection noise mask and the stripe noise mask. Then, the light intensity of the transmitted light signal is subjected to layered power-law correction based on the injection region correction factor and the stripe region correction factor to obtain the correction result of the transmitted light signal.

[0035] S5: Determine whether the data collection is complete. If yes, proceed to step S6. If no, return to step S1 to continue data collection.

[0036] S6: Use the display module to display the calibration results.

[0037] In step S3, the filtering results of small window mid-range filtering and large window mid-range filtering are divided to obtain the relative residual sequence. The median m and MAD value σ of the relative residual sequence are calculated. The injected noise mask is detected with m+4σ+4%·|m| as the entry threshold, m+σ+1.5%·|m| as the exit threshold, and the number of persistence points is between 2 and 20. After replacing or deleting the mask region of the banner noise, large window mid-range filtering is performed again to obtain the updated slowly changing baseline. To update the slowly changing baseline The criteria for detecting banner noise are: an offset of ≥4% relative to the median of the entire segment and a duration that meets the condition.

[0038] The incident light monitoring dynamic synchronization correction system and method based on a ring scintillation crystal of the present invention has the following technical advantages:

[0039] (1) Solving the problem of correction failure caused by the difference in energy between the reference light and the experimental light, and eliminating the principle error by monitoring the same energy along the same path: The present invention places the ring scintillation crystal on the main optical axis inside the experimental cavity. The ring directly receives X-rays with the same energy and the same path as the experimental light. The experimental light passes through the central channel of the ring scintillation crystal, and the outer ring collects signals with the same energy and the same path. Compared with the existing technical solution that uses non-target energy stray light blocked at the edge of the slit as a reference, the present invention fundamentally ensures the energy consistency between the reference light and the experimental light, so that the reference light can truly reflect the intensity fluctuation of the experimental light, overcomes the problem of unstable correlation of light drift at different energies, reduces the normalized residual fluctuation range from 0.1 to within 0.075, and improves the correction accuracy by more than 25%.

[0040] (2) This invention solves the problem of detecting and separating stripe noise, and realizes the detection and correction of stripe noise: Through large window mid-range filtering and MAD baseline offset determination, this invention achieves effective mask generation and power-law correction of long-term stripe drift noise in soft X-ray dynamic correction for the first time. This overcomes the limitation of existing technologies that can only identify short-term peak attenuation (injection noise) caused by electron beam injection through a fixed threshold method, but cannot extract stripe noise. Experimental verification shows that in the STXM image corrected by this invention, large areas of bright and dark stripes completely disappear, and the image uniformity is significantly better than existing solutions.

[0041] (3) Improve the accuracy of correction by using a nonlinear hierarchical model and solve the problem of ignoring nonlinear response by linear normalization: This invention abandons the simple linear division of the prior art and adopts a power law model with a calibration exponent α=1.3 to accurately compensate for the nonlinear gain difference between the reference PMT channel and the transmission PMT channel. The correction effect is better under high attenuation amplitude conditions, and avoids the new correction distortion introduced by linear normalization in the prior art.

[0042] (4) High-speed synchronous triggering and algorithm robustness: The present invention uses a trigger acquisition board to provide 20ns synchronization accuracy and 1MHz sampling rate to meet the needs of sub-millisecond dynamic process monitoring; and the dual-scale filtering and MAD threshold are based on robust statistics, which are not sensitive to outliers, and the algorithm is robust. It can automatically process long-term series data without manual intervention, and is suitable for high-throughput and long-term automated experimental scenarios.

[0043] (5) No light blockage and no loss: The central channel structure of the ring scintillation crystal ensures that the experimental beam passes through without obstruction or absorption, and the off-axis photon counter is not placed on the main optical path, so it will not cause any attenuation or spot deformation to the experimental beam, ensuring that the sample throughput is not lost. Attached Figure Description

[0044] Figure 1 This is a structural block diagram of the incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to the present invention;

[0045] Figure 2 yes Figure 1 The diagram shows the structure of the transmission light photon counter.

[0046] Figure 3 This is a graph showing the results of the count values ​​synchronously acquired from the transmitted light photon counter and the off-axis photon counter, read by the host computer under sample conditions.

[0047] Figure 4 yes Figure 3 A magnified view of region A in the image;

[0048] Figure 5 Yes Figure 4 A magnified image showing the result of further magnification of a single injection point;

[0049] Figure 6 This is a magnified view of a localized area where injection point noise exists under sample conditions.

[0050] Figure 7 This is an STXM imaging result image acquired under sample conditions.

[0051] Figure 8 The hierarchical multiplication correction module in the incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal of this invention is used to... Figure 6 The processed result image;

[0052] Figure 9 This is a flowchart of data acquisition and transmission for a dynamic synchronous correction method for incident light monitoring based on a common-path ring scintillation crystal according to the present invention;

[0053] Figure 10 This is a flowchart of data processing for a dynamic synchronous correction method for incident light monitoring based on a common-path ring scintillation crystal according to the present invention. Detailed Implementation

[0054] The preferred embodiments of the present invention are given below with reference to the accompanying drawings and described in detail to enable a better understanding of the functions and features of the present invention.

[0055] like Figure 1The incident light monitoring dynamic synchronous correction system based on a ring scintillation crystal of the present invention includes a ring scintillation crystal 1, a sample experimental device 2, and a transmitted light photon counter 3, all coaxially arranged with the light source 10; an off-axis photon counter 4, offset from the optical axis and aligned with the side of the ring scintillation crystal facing the sample experimental device 2; a trigger acquisition board 5 connected to both the transmitted light photon counter 3 and the off-axis photon counter 4; and a host computer 6 connected to the trigger acquisition board 5. Thus, the light beam propagates along the optical axis, passing sequentially through the ring scintillation crystal 1, the sample experimental device 2, and the transmitted light photon counter 3.

[0056] The light source 10 receives white light from a synchrotron radiation source and emits monochromatic X-rays. The light source 10 is preferably a grating monochromator.

[0057] The sample experimental apparatus 2 includes a zone plate 21, a sequentially selected aperture 22, and a sample 23 arranged sequentially along the X-ray propagation direction. The zone plate 21 focuses the X-rays, which serve as the experimental light, into a micrometer-sized spot to illuminate the sample 23 for STXM scanning imaging; the sequentially selected aperture 22 is used to block higher-order diffraction light.

[0058] In this embodiment, the material of the ring scintillation crystal 1 is YAG.

[0059] In other embodiments, the ring scintillation crystal 1 can be replaced with other fast scintillation crystal materials, such as LuAG:Ce, GAGG, etc., as long as its light emission decay time is less than 1 microsecond and it can be processed into a ring structure.

[0060] The ring scintillation crystal 1, the sample experimental device 2, and the transmitted light photon counter 3 are installed within the same experimental cavity 100, which provides a vacuum environment. The ring scintillation crystal 1 is located approximately 50 mm upstream of the sample experimental device 2 on the optical axis. It receives X-rays with the same energy and complete path as those received by the sample experimental device 2 and converts them into visible light. The inner wall diameter of its central channel is larger than the diameter of the zone plate 21 in the sample experimental device 2, allowing the central region of the X-rays to pass through the central channel of the ring scintillation crystal 1 without loss and continue propagating along the optical axis. Thus, the ring scintillation crystal 1 acquires a reference light signal with the exact same path and energy as the experimental light without obstructing or attenuating it. In this embodiment, the inner wall diameter of the ring scintillation crystal 1 is 400 μm, and the diameter of the zone plate 21 in the sample experimental device 2 is 300 μm. The spot size of the light propagating forward from the ring scintillation crystal 1 to the zone plate 21 is approximately 500 μm. Therefore, this invention utilizes the portion of light beyond the 300µm diameter of the zone plate and leaves a 100µm tolerance to prevent the annular scintillation crystal 1 from blocking the X-rays required for the experiment, ensuring that the X-rays pass through the central channel without obstruction.

[0061] The transmitted light photon counter 3 is used to collect X-rays transmitted through sample 23, convert them into visible light, and count them as the transmitted light signal passing through sample 23. The off-axis photon counter 4 is positioned off the optical axis, with its photocathode aligned with the surface of the ring scintillation crystal 1 facing the sample experimental device 2 (i.e., the downstream side of the beam) to collect and count the visible light converted by the ring scintillation crystal 1 as a reference light signal. Both the transmitted light photon counter 3 and the off-axis photon counter 4 are photomultiplier tube (PMT) photon counters. The arrangement of the ring scintillation crystal 1, the transmitted light photon counter 3, and the off-axis photon counter 4 solves the problem of correction failure caused by the difference in energy between the reference light and the experimental light in the prior art, as well as the problem of light intensity loss caused by traditional on-axis monitoring.

[0062] Figure 2 for Figure 1 The diagram shows the structure of the transmission light photon counter 3. The transmission light photon counter 3 is located behind the sample, and its specific structure is existing, as shown below. Figure 2 As shown, the transmission light photon counter 3 includes a planar scintillation crystal 31 (preferably a YAG crystal), a light guide tube 32, and a photomultiplier tube 33 arranged sequentially along the incident direction of the X-rays. The photomultiplier tube 33 is connected to a high-voltage power supply 34 and is connected to the FPGA trigger acquisition board 5 through a counting element 35. The counting element 35 includes a discriminator and a counter.

[0063] Therefore, as Figure 2 As shown, X-rays are converted into visible light after irradiating the planar scintillation crystal 31 at the front end of the transmission light photon counter 3. The off-axis photon counter 4 converts the X-rays into visible light through the ring scintillation crystal 1. Both the transmission light photon counter 3 and the off-axis photon counter 4 operate as integrating types, and their working principles can be summarized as follows: Visible light is focused within the light guide tube 32 of the PMT photon counter and then irradiates the photomultiplier tube 33. The electron beam output from the photomultiplier tube 33 and its internal amplifier strikes the discriminator of the counting element 35. The photomultiplier tube 33 is powered by a high-voltage power supply 34, which accelerates the electrons. The voltage of the high-voltage power supply 34 is set to -1000V. The discriminator of the counting element 35 shapes the electron beam to improve the signal-to-noise ratio, making it easier for the counter to distinguish individual signals and reducing the influence of noise. The signal distinguished by the discriminator of the counting element 35 is recorded as a valid signal, and the count value is incremented by 1, representing that the PMT photon counter has received a single photon signal.

[0064] The trigger acquisition board 5 is connected to both the transmitted light photon counter 3 and the off-axis photon counter 4. It is configured to provide trigger signals indicating the start and end of counting to synchronize the acquisition of data between the transmitted light photon counter 3 and the off-axis photon counter 4. The synchronization trigger accuracy is 20 ns, and the maximum acquisition rate is 1 MHz. The trigger acquisition board 5 is preferably an FPGA; in this embodiment, it is a ZYNQXA7020 manufactured by Xilinx.

[0065] The trigger acquisition board 5 is connected to the host computer 6 via a network port transmission protocol to transmit the transmitted light signal and reference light signal synchronously acquired by the synchronous transmitted light photon counter 3 and the off-axis photon counter 4 to the host computer 6.

[0066] The host computer 6 is equipped with a reading module, a display module, a dual-scale median filtering module, a MAD threshold detection module, and a hierarchical multiplication correction module for data reading, display, processing, and correction. Preferably, each module of the host computer 6 is implemented using Python. In other embodiments, the processing modules in the host computer 6 can be implemented using other programming languages ​​such as C++ and MATLAB.

[0067] The dual-scale median filtering module is configured to perform small-window median filtering (preserving short-term abrupt changes such as injection spikes) and large-window median filtering (extracting the stripe baseline) on the reference light signal acquired by the off-axis photon counter 4 to obtain the filtering results.

[0068] The MAD threshold detection module is configured to process the filtering results of small window mid-range filtering and large window mid-range filtering to obtain a relative residual sequence, detect short-term injection events based on the relative residual sequence to generate a mask for injected noise, and detect baseline offset regions based on the filtering results of large window mid-range filtering to obtain a mask for banner noise.

[0069] Therefore, the reference light signal pmt_ref (data synchronously acquired in the upstream off-axis region) collected by the off-axis photon counter 4 is identified, and the stripe region and injection point region are marked in the form of a mask, serving as a mask for injection noise and a mask for stripe noise. Since the reference light signal pmt_ref and the transmitted light signal pmt passing through the sample are acquired synchronously, the transmitted light signal pmt can be processed within the same mask region. This processing refers to restoring the injection noise mask using a single-event injection region correction factor, and correcting the stripe noise mask using a continuously varying stripe correction factor. The specific processing method will be detailed below.

[0070] Injection masks and stripe masks are binary labels used to separate two types of noise in a time series: the injection mask precisely locates the start and end times of each electron beam injection event, ensuring that the injection region correction factor only applies to the time period in which the injection occurs, avoiding point-by-point contamination of the sample signal; the stripe mask marks slowly varying drift regions with a baseline offset ≥4% and lasting for ≥20 sampling points, ensuring that the stripe correction factor is continuously applied only within the stripe region without interfering with normal data. The two masks do not overlap on the time axis, allowing injection noise and stripe noise to be independently corrected by their respective power-law models. This masking achieves a temporal "divide and conquer" strategy, providing precise spatial localization for the hierarchical multiplicative correction model.

[0071] The MAD threshold detection module divides the filtering results of small-window and large-window mid-range filtering to obtain the relative residual sequence. It then calculates the median *m* and the MAD value *σ* of the relative residual sequence. Using *m+4σ+4%·|m|* as the entry threshold and *m+σ+1.5%·|m|* as the exit threshold, and with a duration of 2-20 points, it detects the injected noise mask. After replacing or deleting the mask region containing the banner noise, it performs large-window mid-range filtering again to obtain a slowly changing baseline that is unaffected by short-term events. When a region updates slowly, the baseline changes. A continuous deviation of ≥4% from the median over the entire duration, with a duration that meets certain conditions (e.g., ≥20 sampling points), is used as the criterion for determining banner noise, and a mask for detecting banner noise is obtained.

[0072] Specifically, the synchronization time series of the transmitted light signal acquired after the sample and the reference light signal detected by the ring yag are respectively the transmitted light signal sequence. and reference optical signal sequence First, the reference optical signal sequence... Small window mid-value filtering is performed to obtain a reference sequence for event detection. (i.e., the filtering result of small window mid-range filtering) is used to suppress short-term injection noise caused by electron beam injection; simultaneously, large window mid-range filtering is used to obtain the initial slowly changing baseline (i.e., the filtering result of large window mid-range filtering). The small window mid-range filtering is used to mark the decrease in the PMT value of the injection point noise, while the large window mid-range filtering is used to find the baseline. Dividing the two yields a relative residual sequence; those exceeding a threshold are marked as injection point events. Based on this principle, the MAD threshold detection module is configured to use the reference sequence for event detection. The relative residual sequence obtained by dividing the relative residual sequence by the initial slowly changing baseline is used as the event detection metric, and an adaptive threshold is constructed based on the median *m* and median absolute deviation *σ* of the relative residual sequence. Median absolute deviation (MAD) is a robust statistic of dispersion, defined as: σ = 1.4826 multiplied by the median (|rm|). Compared to standard deviation, MAD is insensitive to outliers and is suitable for signals with a large amount of injected spike noise. A higher seed threshold of 4σ is used to identify the core of the anomalous event and to provide a higher threshold for injection detection to avoid false triggering. A lower release threshold of 1σ is used to expand the event boundary and to ensure timely exit after the event ends to avoid mask extension. The minimum relative changes set in this paper are 4% and 1.5%, respectively. Therefore, the MAD threshold detection module uses *m + 4σ + 4%·|m|* as the entry threshold, *m + σ + 1.5%·|m|* as the exit threshold, and a duration of 2–20 points to detect the mask of injected noise.

[0073] Based on the mask of injected noise obtained from the relative residual sequence, further processing can be performed to obtain the baseline offset of each acquisition point relative to the median of the entire segment. Specifically, after deleting or replacing the masked area of ​​the identified events (i.e., banner noise) with the corresponding initial slowly changing baseline, large window mid-range filtering is performed again to obtain an updated slowly changing baseline that is unaffected by short-term events. When a region updates its baseline slowly... When a region consistently deviates more than 4% from its median over its entire length, and the duration meets the specified criteria, the region is marked as a mask for stripe noise. Update the slowly changing baseline. The offset relative to the median over its entire duration refers to the difference between the median of all sequence data and the slowly changing baseline. Perform subtraction, then divide by the updated slow-changing baseline. The resulting value is (Mid-base) / base, where base is the updated slowly changing baseline and Mid is the median of all sequence data.

[0074] In this embodiment, the preferred length of the small window is 3 points. The selection of a 3-point window is based on the injection point noise duration being 2 to 20 sampling points. A 3-point window can suppress single-point random fluctuations while preserving the complete injection spike. The preferred length of the large window is 51 points. The selection of a 51-point window is based on the stripe noise being a low-frequency drift lasting more than 20 sampling points. A 51-point window can smooth short-term fluctuations and reflect the baseline trend. The window size of the dual-scale median filter can be adjusted according to the sampling frequency and noise characteristics of the specific experiment.

[0075] The hierarchical multiplication correction module is configured to obtain injection region correction factors and stripe region correction factors based on the injection noise mask and the stripe noise mask, and then perform hierarchical power-law correction on the intensity of the transmitted light signal collected by the transmitted light photon counter 3 based on the injection region correction factors and the stripe region correction factors to obtain the correction result of the transmitted light signal; the display module is used to display the correction result. The injection region correction factor and the stripe region correction factor are both obtained through calibration experiments. The total correction factor is the product of the injection region correction factor and the stripe correction factor.

[0076] The hierarchical multiplicative correction module employs independent power-law correction models for injection noise and stripe noise: for the injection noise mask, a single-event injection region correction factor f_event=η^(-γ) is used for recovery, where η is the ratio of the mean of the reference light signal within the event to the baseline mean of the reference light signal before the event, and γ is the experimentally calibrated injection correction power-law exponent. A uniform injection region correction factor γ is used for each injection event to avoid point-by-point noise contamination of the sample structure in the reference channel.

[0077] For the stripe noise mask, a continuously varying stripe correction factor f_stripe[n]=(B_global / I_ref_large[n])^α is used for correction, where B_global is the baseline mean of the reference optical signal during the noise-free period, I_ref_large[n] is the baseline of the large window mid-range filter (i.e., the baseline of the reference optical signal), n is the sampling point number in the time series, i.e., the count value obtained by the nth synchronous trigger acquisition of the FPGA trigger acquisition board; α is the experimentally calibrated stripe correction power law exponent.

[0078] The total correction factor is the product of the injection region correction factor f_event and the stripe correction factor f_stripe[n]. The stripe correction power-law exponent α = 1.3 was obtained through experimental calibration and reflects the non-1:1 power-law response relationship between the reference light signal and the transmitted light signal.

[0079] n is the sampling point number in the time series, and the value of n ranges from 0, 1, 2, ..., N-1 (N is the total number of sampling points in a single experiment). The corresponding actual physical time is t = n × ΔT, where ΔT is the sampling interval set by the FPGA trigger acquisition board (5) (typical value 1ms). The calculation method of the baseline of the large window mid-range filtering is as follows: For each sampling point n, take the count value of the n-25th to n+25th points (a total of 51 points) in the reference optical signal sequence, sort them according to their numerical size, and take the middle value (i.e., the 26th value). This middle value is the baseline estimate of the point. The filtering result of the large window mid-range filtering is the baseline of the large window mid-range filtering. Edge points (n<25 or n>N-26) are processed by symmetrical filling. The physical meaning of this baseline is the typical level of the reference light intensity within the time window of the nth sampling point and the 25 sampling points before and after the point, reflecting the slow change trend after removing short-term injection spikes and single-point random fluctuations within the time window.

[0080] The calibration of α reflects the power-law relationship of the slow-point amplitude response of the two channels: Data collected under sample-free conditions, after removing injection points, is used to perform linear regression on the logarithms of the transmitted light signal pmt and the reference light signal pmt_ref collected after the sample. Currently, α is manually adjusted; for some data from BL08U1A, α is set to 1.3. The algorithm will be improved for automatic calibration in the future. The calibration of γ reflects the instantaneous injection point event response relationship. Injection point events are extracted, each is calculated, and the mean is taken. The value of γ depends on the collected data.

[0081] The corrected downstream light intensity I_PMT_corr[n] is obtained by directly multiplying the total correction factor f_total[n] by the original downstream light intensity I_PMT[n] collected by the transmitted light photon counter 3. Specifically, for each sampling point n, the correction formula is:

[0082] I_PMT_corr[n] = I_PMT[n] × f_total[n],

[0083] The total correction factor f_total[n] is calculated as follows: when the point is marked as an injection mask (event_mask=1), f_event[n] takes the power-law factor η^(-γ) of the corresponding injection event; otherwise, it takes 1. When the point is marked as a stripe mask (stripe_mask=1), f_stripe[n] takes (B_global / I_ref_large[n])^α; otherwise, it takes 1. The total correction factor f_total[n] is the product of the two:

[0084] f_total[n] = f_event[n] × f_stripe[n].

[0085] The total correction factor is then multiplied by the original light intensity to complete the correction at that point. The correction process is performed independently point by point, without involving spatial domain operations, ensuring that the spatial resolution of the STXM image is not affected.

[0086] like Figure 3 As shown, under sample conditions, the host computer 6 reads the count values ​​synchronously acquired from the transmitted light photon counter 3 and the off-axis photon counter 4 via the trigger acquisition board 5. Figure 3 In the graph, the horizontal axis represents time (ms), and the vertical axis represents the number of photons. The lower curve represents the reference light intensity collected by off-axis photon counter 4 (reflecting only light source fluctuations), and the upper curve represents the downstream light intensity collected by transmitted light photon counter 3 (modulated by sample absorption). From Figure 3 As can be seen, both signals exhibit periodic electron beam injection-induced light intensity spikes (injection point noise), while the signal baseline shows slow fluctuations (strip noise). This demonstrates that the injection point noise and strip noise are independent of the sample and originate from fluctuations in the light source itself.

[0087] like Figure 4 As shown, for Figure 3 The noise regions at the injection points and along the stripes are locally magnified, where the circular data points represent the transmitted light signal (sample signal), and the cross-shaped data points represent the reference light signal. From Figure 4 As shown in the magnified image, the stripe noise exhibits a consistent slow variation trend in both the transmitted light signal and the reference light signal, demonstrating good correlation. At the injection point, both the transmitted light signal and the reference light signal show a sharp attenuation, and this peak feature can be accurately identified and located in the background baseline. This proves that the common-path light extraction scheme implemented by the ring scintillation crystal 1 in this invention enables the reference light signal to truly reflect the fluctuations of the experimental light, providing a reliable basis for the accurate generation of the subsequent injection mask.

[0088] like Figure 5 As shown, further magnification of a single injection point reveals that the attenuation waveforms of the transmitted light signal and the reference light signal at the injection moment almost completely overlap, proving that the ring scintillation crystal 1 used in this invention enables the reference light and the experimental light to have excellent time correlation, providing a reliable reference light signal for subsequent accurate correction.

[0089] like Figure 6 As shown, under sample conditions, the local area with injection point noise is displayed. Circled data points represent transmitted light signals (sample signals), and cross-shaped data points represent reference light signals. From... Figure 6As can be seen, the spike feature at the injection point is clearly distinguishable in both signals, indicating that the method of the present invention can accurately identify and precisely locate the injection point, unaffected by changes in signal amplitude caused by sample absorption.

[0090] like Figure 7 As shown, in the STXM imaging results acquired under sample conditions, due to the presence of injection point noise, horizontal black stripes (outlined areas) appear in the image. At the same time, due to the presence of stripe noise, the background brightness of the image is uneven (showing large-area changes in brightness). The details of the sample structure are obscured by noise, which seriously affects the imaging quality.

[0091] like Figure 8 As shown, the hierarchical multiplication correction module of the present invention is used to correct the multiplication error. Figure 7 After processing, both injection point noise and stripe noise were effectively eliminated, resulting in uniform image contrast and clear identification of sample structural details.

[0092] like Figure 9 and Figure 10 As shown, the present invention also provides a dynamic synchronization correction method for incident light monitoring based on a common-path ring scintillation crystal, comprising the following steps:

[0093] Step S0: Build the incident light monitoring dynamic synchronization correction system based on the ring scintillation crystal described above, and set the system parameters in the host computer 6;

[0094] The parameters set in the host computer 6 include the entry threshold, exit threshold and banner noise judgment conditions of the MAD threshold detection module, as well as the trigger acquisition time interval and number of trigger acquisition points of the trigger acquisition board 5.

[0095] Step S1: Trigger acquisition board 5 synchronously triggers transmission light photon counter 3 and off-axis photon counter 4 to acquire transmitted light signal and reference light signal;

[0096] Step S2: Use the dual-scale median filtering module of the host computer 6 to perform small window median filtering and large window median filtering on the reference optical signal to obtain the filtering result;

[0097] Specifically, the time series of the reference optical signal is subjected to a 3-point small window mid-range filter to obtain a time series that retains the injection spike characteristics; the time series of the reference optical signal is subjected to a 51-point large window mid-range filter to obtain a time series that reflects the slowly varying baseline of the stripe.

[0098] Step S3: Using the MAD threshold detection module of the host computer 6, the relative residual sequence is obtained by processing the filtering results of small window mid-range filtering and large window mid-range filtering; short-term injection events are detected based on the relative residual sequence to generate a mask for injected noise, and the baseline offset region is detected based on the filtering results of large window mid-range filtering to obtain a mask for banner noise.

[0099] The relative residual sequence is obtained by dividing the filtering results of the small window mid-range filter and the large window mid-range filter, and the median m and MAD value σ of the relative residual sequence are calculated.

[0100] The mask for injected noise is obtained by using m+4σ+4%·|m| as the entry threshold, m+σ+1.5%·|m| as the exit threshold, and the number of continuous points is between 2 and 20.

[0101] After replacing or removing the mask region containing banner noise, large window mid-range filtering is performed again to obtain the updated slowly changing baseline. To update the slowly changing baseline The criteria for detecting banner noise are: an offset of ≥4% relative to the median of the entire segment and a duration that meets the condition.

[0102] Step S4: For the transmitted light signal, a layered multiplication correction module is used to obtain the injection region correction factor and the stripe region correction factor based on the injection noise mask and the stripe noise mask. Then, the light intensity of the transmitted light signal is subjected to layered power-law correction based on the injection region correction factor and the stripe region correction factor to obtain the correction result of the transmitted light signal.

[0103] The power law exponent α = 1.3 for the banner correction was obtained through experimental calibration.

[0104] Step S5: Determine whether the data collection is complete. If yes, proceed to step S6; otherwise, return to step S1 to continue data collection.

[0105] Step S6: Display the calibration results using the display module.

[0106] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. Any simple equivalent changes and modifications made based on the claims and description of this application fall within the protection scope of the claims of this patent. All aspects not described in detail in this invention are conventional technical content.

Claims

1. A dynamic synchronization correction system for incident light monitoring based on a ring scintillation crystal, characterized in that, The device includes a ring scintillation crystal (1) coaxially arranged with the light source (10), a sample experimental device (2) and a transmitted light photon counter (3), an off-axis photon counter (4) offset from the optical axis and aligned with the side of the ring scintillation crystal facing the sample experimental device (2), a trigger acquisition board (5) connected to both the transmitted light photon counter (3) and the off-axis photon counter (4), and a host computer (6) connected to the trigger acquisition board (5).

2. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 1, characterized in that, The host computer (6) is equipped with a reading module, a display module, a dual-scale median filtering module, a MAD threshold detection module, and a hierarchical multiplication correction module; the dual-scale median filtering module is configured to perform small window median filtering and large window median filtering on the reference light signal collected by the off-axis photon counter (4) to obtain the filtering result; the MAD threshold detection module is configured to process the filtering results of small window median filtering and large window median filtering to obtain a relative residual sequence, detect short-time injection events based on the relative residual sequence to generate a mask for injection noise, and detect the baseline offset region based on the filtering result of large window median filtering to obtain a mask for stripe noise; The hierarchical multiplication correction module is configured to obtain the injection region correction factor and the stripe region correction factor based on the injection noise mask and the stripe noise mask, and perform hierarchical power-law correction on the light intensity of the transmitted light signal collected by the transmitted light photon counter (3) based on the injection region correction factor and the stripe region correction factor to obtain the correction result of the transmitted light signal.

3. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 2, characterized in that, The small window has a length of 3 points, and the large window has a length of 51 points. The MAD threshold detection module divides the filtering results of the small window mid-range filtering and the large window mid-range filtering to obtain the relative residual sequence, and calculates the median m and MAD value σ of the relative residual sequence. The injected noise mask is detected by using m+4σ+4%·|m| as the entry threshold, m+σ+1.5%·|m| as the exit threshold, and the number of continuous points is between 2 and 20. After replacing or removing the mask region containing banner noise, large window mid-range filtering is performed again to obtain the updated slowly changing baseline. To update the slowly changing baseline The criteria for detecting banner noise are: an offset of ≥4% relative to the median of the entire segment and a duration that meets the condition.

4. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 2, characterized in that, In the layered multiplication correction module, the injection region correction factor is used to recover the mask for injected noise. Wherein, the injection region correction factor f_event=η^(-γ), where η is the ratio of the mean of the reference optical signal within the event to the baseline mean of the reference optical signal before the event, and γ is the experimentally calibrated injection correction power law exponent. For masking banner noise, a continuously varying banner correction factor is used for correction; Wherein, the stripe correction factor f_stripe[n]=(B_global / I_ref_large[n])^α, B_global is the baseline mean of the reference optical signal during the noise-free period, I_ref_large[n] is the baseline of the large window mid-range filter, n represents the sampling point number in the time series, and α is the experimentally calibrated stripe correction power law exponent; The total correction factor is the product of the injection region correction factor f_event and the stripe correction factor f_stripe[n].

5. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 1, characterized in that, The sample experimental device (2) includes a zone plate (21), a graded aperture (22), and a sample (23) arranged sequentially along the X-ray propagation direction. The ring scintillation crystal (1) is located on the optical axis about 50 mm upstream of the sample experimental device (2). It receives X-rays that are completely co-pathed with and have the same energy as the X-rays received by the sample experimental device (2) and converts them into visible light. The inner wall diameter of the central channel of the ring scintillation crystal (1) is larger than the diameter of the zone plate (21) in the sample experimental device (2), so that the central region of the X-rays passes through the central channel of the ring scintillation crystal (1) without loss and continues to propagate along the optical axis.

6. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 5, characterized in that, The ring scintillation crystal (1), the sample experimental device (2), and the transmitted light photon counter (3) are installed in the same experimental cavity (100), which has a vacuum environment. The transmitted light photon counter (3) is used to collect X-rays that pass through the sample (23), convert them into visible light and count them as the transmitted light signal passing through the sample (23); the off-axis photon counter (4) is set at a position off the optical axis, and its photocathode is aligned with the side surface of the ring scintillation crystal (1) facing the sample experimental device (2) to collect the visible light converted by the ring scintillation crystal (1) and count it as a reference light signal.

7. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 1, characterized in that, The trigger acquisition board (5) is configured to provide trigger signals indicating the start and end of counting, so as to synchronously trigger the transmitted light photon counter (3) and the off-axis photon counter (4) to perform synchronous integration counting.

8. The incident light monitoring dynamic synchronization correction system based on a ring scintillation crystal according to claim 1, characterized in that, The material of the ring scintillation crystal (1) is YAG, LuAG:Ce or GAGG.

9. A dynamic synchronization correction method for incident light monitoring based on a common-path ring scintillation crystal, characterized in that, include: Step S0: Build the incident light monitoring dynamic synchronization correction system based on the ring scintillation crystal according to any one of claims 1-8, and set the system parameters in the host computer (6); Step S1: Trigger the acquisition board (5) to synchronously trigger the transmitted light photon counter (3) and the off-axis photon counter (4) to acquire the transmitted light signal and the reference light signal; Step S2: Use the dual-scale median filtering module of the host computer (6) to perform small window median filtering and large window median filtering on the reference optical signal to obtain the filtering result; Step S3: Using the MAD threshold detection module of the host computer (6), the relative residual sequence is obtained by processing the filtering results of the small window mid-range filtering and the large window mid-range filtering; the short-term injection event is detected according to the relative residual sequence to generate the mask of injection noise, and the baseline offset region is detected according to the filtering result of the large window mid-range filtering to obtain the mask of banner noise. Step S4: For the transmitted light signal, a layered multiplication correction module is used to obtain the injection region correction factor and the stripe region correction factor based on the injection noise mask and the stripe noise mask. Then, the light intensity of the transmitted light signal is subjected to layered power-law correction based on the injection region correction factor and the stripe region correction factor to obtain the correction result of the transmitted light signal. Step S5: Determine whether the data collection is complete. If yes, proceed to step S6; otherwise, return to step S1 to continue data collection. Step S6: Display the calibration results using the display module.

10. The method according to claim 9, characterized in that, In step S3, the filtering results of small window mid-range filtering and large window mid-range filtering are divided to obtain the relative residual sequence, and the median m and MAD value σ of the relative residual sequence are calculated. The mask for injected noise is obtained by using m+4σ+4%·|m| as the entry threshold, m+σ+1.5%·|m| as the exit threshold, and the number of continuous points is between 2 and 20. After replacing or removing the mask region containing banner noise, large window mid-range filtering is performed again to obtain the updated slowly changing baseline. To update the slowly changing baseline The criteria for detecting banner noise are: an offset of ≥4% relative to the median of the entire segment and a duration that meets the condition.

Images