CN-122023551-A - Electron beam parameter optimization method and system based on real-time image quality

CN122023551ACN 122023551 ACN122023551 ACN 122023551ACN-122023551-A

Abstract

The invention provides an electron beam parameter optimization method and system based on real-time image quality, and relates to the technical field of electron beam microscopic imaging. The method comprises the steps of synchronously collecting and time aligning characteristic X-ray signals and cathode luminescence signals excited when an electron beam scans a sample, generating time aligned original spectrum data, extracting spectral line characteristics according to the time aligned original spectrum data and combining the spectral line characteristics into dynamic spectrum fingerprints, generating an image quality predicted value of a current scanning point through a pre-trained neural network model, calculating and generating an adjustment quantity instruction of electron beam parameters through a controller algorithm based on deviation of the predicted value and a preset target value, controlling an executing component according to the instruction, adjusting the electron beam scanning parameters, finally driving the scanning and collecting signals by utilizing the adjusted parameters, generating an optimized scanning image, realizing real-time, automatic and closed-loop optimization of the electron beam parameters, and effectively improving stability, consistency and system automation level of imaging quality.

Inventors

LIU YUNPENG
FAN HAIFENG
MENG NING
QIN XIAODI

Assignees

天津市佳通兄弟科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (9)

1. The electron beam parameter optimization method based on the real-time image quality is characterized by comprising the following steps of: s1, synchronously acquiring and time-aligning at least two types of signals including a characteristic X-ray signal and a cathodoluminescence signal based on physical signals excited in an interaction area when an electron beam scans a sample, and generating time-aligned original spectrum data; s2, extracting spectral line characteristics based on time-aligned original spectral data and combining the spectral line characteristics into characteristic vectors to generate dynamic spectral fingerprints, wherein S2 comprises: S21, respectively carrying out background subtraction and peak position identification on a characteristic X-ray spectrum and a cathode luminescence spectrum based on time-aligned original spectrum data to generate net spectrum peak data; S22, carrying out normalization processing on the intensity of the selected characteristic peak based on the net spectrum peak data to generate a normalized spectrum peak intensity value; S23, based on the standardized spectrum peak intensity values, sequentially splicing the characteristic intensity values of different spectrums into a multidimensional vector to generate a dynamic spectrum fingerprint; s3, calculating through a pre-trained neural network model based on the dynamic spectrum fingerprint to generate an image quality predicted value corresponding to the current scanning point; S4, calculating through a controller algorithm based on the image quality predicted value and a preset image quality target value, and generating an adjustment quantity instruction of the electron beam parameters; s5, based on an adjustment quantity instruction of the electron beam parameters, sending a control signal to an execution component in the electron optical system to generate adjusted electron beam scanning parameters; and S6, driving the electron beam to scan and collect sample signals based on the adjusted electron beam scanning parameters, and generating an optimized scanning image.
2. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, wherein S23 comprises: s231, organizing the standardized spectrum peak intensity values into an array arranged according to the spectrum line type and the energy sequence; S232, inputting the array into a feature fusion module, wherein the feature fusion module comprises a full connection layer; S233, performing linear transformation and dimension reduction on the groups through the full connection layer, outputting a multi-dimension fusion feature vector, and generating a dynamic spectrum fingerprint.
3. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, wherein S3 comprises: S31, forming a time sequence input sequence by the dynamic spectrum fingerprint of the current scanning point and the dynamic spectrum fingerprints of the previous continuous N scanning points together, wherein N is a positive integer; s32, inputting a time sequence input sequence into a time sequence neural network model comprising a convolution layer and a pooling layer; S33, extracting a local mode of a time sequence input sequence through a convolution layer in the time sequence neural network model, and performing feature selection through a pooling layer to generate high-level time sequence features; S34, carrying out regression calculation through an output layer of the time sequence neural network model based on the high-layer time sequence characteristics to generate an image quality predicted value.
4. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, wherein S4 comprises: S41, calculating a numerical deviation between the image quality predicted value and the image quality target value; S42, calculating through a proportional-integral control law based on the numerical deviation and the historical accumulated value thereof to generate a continuous control output quantity; s43, quantizing the control output quantity into an adjustment value and direction of the specific electron beam parameters, and generating an adjustment quantity instruction of the electron beam parameters.
5. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, wherein in S5, the control signal sent to the execution unit is specifically: s51, analyzing a target change value of electron beam intensity based on an adjustment quantity instruction of electron beam parameters; s52, inquiring a beam control voltage lookup table based on the target change value to obtain a corresponding grid voltage value; And S53, generating an analog voltage signal based on the grid voltage value and applying the analog voltage signal to a grid of the electron gun to generate a changed electron beam current.
6. The electron beam parameter optimizing method based on real-time image quality as claimed in claim 5, after S53, the method further comprises: s54, inputting the changed electron beam current to a beam spot stabilizing controller; s55, calculating a compensation value of the current of the condenser lens according to a preset beam current-lens current relation through a beam spot stabilizing controller; and S56, adjusting the driving current of the condenser coil based on the compensation value of the condenser current to generate an electron beam condition with stable beam spot size.
7. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, wherein S5 comprises: S5a, analyzing the adjustment quantity of the scanning residence time based on the adjustment quantity instruction of the electron beam parameters; s5b, converting into an adjustment command for the pixel clock frequency of the scanning generator based on the adjustment amount for the scanning residence time; And S5c, changing the pulse interval of the pixel clock based on the adjustment command, and generating new scanning residence time, namely the adjusted electron beam scanning parameters.
8. The method for optimizing electron beam parameters based on real-time image quality according to claim 1, further comprising, prior to S1: S0, planning a scanning track point sequence of the electron beam based on a region to be imaged of the sample; wherein S1 to S6 are loop processes sequentially performed for each point in the scan trajectory point sequence.
9. Electron beam parameter optimizing system based on real-time image quality, characterized in that the system employs the electron beam parameter optimizing method based on real-time image quality according to any one of claims 1 to 8, the system comprising: The signal acquisition module is used for synchronously acquiring and time-aligning at least two types of signals comprising characteristic X-ray signals and cathodoluminescence signals based on physical signals excited in an interaction area when the electron beam scans a sample, and generating time-aligned original spectrum data; The dynamic spectrum fingerprint generation module is used for extracting spectral line characteristics based on time-aligned original spectral data and combining the spectral line characteristics into characteristic vectors to generate dynamic spectrum fingerprints; The image quality prediction module is used for calculating through a pre-trained neural network model based on the dynamic spectrum fingerprint to generate an image quality prediction value corresponding to the current scanning point; The parameter adjustment instruction generation module is used for calculating through a controller algorithm based on the image quality predicted value and a preset image quality target value to generate an adjustment quantity instruction of the electron beam parameters; the electron beam parameter adjusting module is used for sending a control signal to an executing component in the electron optical system based on an adjusting quantity instruction of the electron beam parameter to generate an adjusted electron beam scanning parameter; and the scanning image generation module is used for driving the electron beam to scan and collect the sample signal based on the adjusted electron beam scanning parameters, and generating an optimized scanning image.

Description

Electron beam parameter optimization method and system based on real-time image quality Technical Field The invention relates to the technical field of electron beam microscopic imaging, in particular to an electron beam parameter optimization method and system based on real-time image quality. Background In the field of electron beam microscopy imaging technology, such as Scanning Electron Microscopy (SEM) or Electron Probe Microscopy Analyzer (EPMA), obtaining high quality images is the basis for performing accurate observations and analyses. Image quality (e.g., resolution, signal-to-noise ratio, contrast) is highly dependent on the settings of the electron beam parameters (e.g., beam current, acceleration voltage, scan speed, etc.). Currently, the setting and optimization of electron beam parameters is mainly dependent on the experience of the operator. In the imaging process, an operator needs to estimate a group of parameters according to the characteristics of a sample in advance, scan and image, then manually evaluate the image quality, manually adjust the parameters according to the image quality, and approach the optimal imaging condition through repeated attempts. The method has obvious disadvantages that firstly, the repeatability of adjustment results of different personnel or the same personnel in different states is poor and the efficiency is low seriously depending on the skills and experience of operators, secondly, the manual adjustment and verification period is long, the method cannot adapt to the situation that local differences exist in the surface components, the morphology or the electrical characteristics of a sample, the image quality of different areas in the scanning process is unstable, and finally, the whole process is open-loop and non-self-adaptive, and an intelligent method capable of automatically, quickly and in a closed-loop optimizing the electron beam parameters according to the physical signals of the sample acquired in real time so as to stably maintain or dynamically approximate the quality of a target image in the whole scanning process is lacked. Therefore, a technical solution capable of realizing real-time, automatic and closed-loop optimization of electron beam parameters is needed to improve the automation level of the imaging system, the stability and consistency of image quality, and the overall imaging efficiency. Disclosure of Invention In order to solve the above problems in the prior art, a first aspect of the present invention provides an electron beam parameter optimization method based on real-time image quality, including: s1, synchronously acquiring and time-aligning at least two types of signals including a characteristic X-ray signal and a cathodoluminescence signal based on physical signals excited in an interaction area when an electron beam scans a sample, and generating time-aligned original spectrum data; s2, extracting spectral line characteristics based on time-aligned original spectral data and combining the spectral line characteristics into characteristic vectors to generate dynamic spectral fingerprints, wherein S2 comprises: S21, respectively carrying out background subtraction and peak position identification on a characteristic X-ray spectrum and a cathode luminescence spectrum based on time-aligned original spectrum data to generate net spectrum peak data; S22, carrying out normalization processing on the intensity of the selected characteristic peak based on the net spectrum peak data to generate a normalized spectrum peak intensity value; S23, based on the standardized spectrum peak intensity values, sequentially splicing the characteristic intensity values of different spectrums into a multidimensional vector to generate a dynamic spectrum fingerprint; s3, calculating through a pre-trained neural network model based on the dynamic spectrum fingerprint to generate an image quality predicted value corresponding to the current scanning point; S4, calculating through a controller algorithm based on the image quality predicted value and a preset image quality target value, and generating an adjustment quantity instruction of the electron beam parameters; s5, based on an adjustment quantity instruction of the electron beam parameters, sending a control signal to an execution component in the electron optical system to generate adjusted electron beam scanning parameters; and S6, driving the electron beam to scan and collect sample signals based on the adjusted electron beam scanning parameters, and generating an optimized scanning image. Compared with the prior art, the invention has the beneficial effects that: Through close coordination of six steps, a complete real-time closed-loop optimization system is constructed, and the problems that the background technology depends on manpower, has low efficiency, cannot respond in real time and lacks closed-loop optimization are effectively solved. First, step S1 generates time-aligned raw spe