Adaptive error compensation method and system for focused ion beam equipment

Abstract

The invention provides a self-adaptive error compensation method and system for focused ion beam equipment, and the method comprises the steps: obtaining a real-time error signal set generated by the focused ion beam equipment in a machining process, training a self-adaptive error compensation model according to the real-time error signal set, and generating an error compensation parameter set; inputting the error compensation parameter set into a control unit of the focused ion beam equipment, and generating a real-time compensation control instruction set; adjusting an ion source parameter and a deflection electrode voltage parameter of the focused ion beam equipment according to the real-time compensation control instruction set, and outputting an adjusted ion beam path parameter set; and performing ion beam processing on the target processing area according to the adjusted ion beam path parameter set, and monitoring morphology feature data of the processed area in real time to verify an error compensation effect. According to the invention, harmful effects caused by thermal effects, electric fields or magnetic fields can be eliminated, and the self-adaptability and long-term reliability of a compensation strategy are ensured.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular, to an adaptive error compensation method and system for a focused ion beam device. Background Art

[0002] Error compensation of a focused ion beam device is used to eliminate the ion beam path deviation caused by thermal effects, electromagnetic interference or mechanical vibration during the processing. In the prior art, threshold filtering or a static compensation model of an error signal (such as ion beam intensity or path deviation) is usually adopted, and the ion source current or deflection voltage is directly adjusted by preset compensation parameters to correct the path deviation. However, the compensation parameters based on the static model in the existing methods cannot dynamically respond to the coupling effect of the environmental interference signal (such as sudden change in electromagnetic noise frequency or thermal accumulation effect) and the internal error signal of the device (such as ion beam intensity fluctuation), resulting in problems of compensation lag or overcorrection, and lacking a cooperative processing mechanism for multi-source error signals, so that the compensation accuracy is limited by the linear superposition error of single-signal adjustment. At the same time, the traditional model training depends on historical data and cannot adapt to the dynamic interference scenario in the processing process in real time, resulting in weak anti-interference ability and insufficient long-term stability, and it is difficult to balance the processing accuracy and efficiency under complex working conditions. Summary of the Invention

[0003] The present invention provides an adaptive error compensation method and system for a focused ion beam device.

[0004] In a first aspect, an embodiment of the present invention provides an adaptive error compensation method for a focused ion beam device, the method including: acquiring a set of real-time error signals generated during the processing of the focused ion beam device, where the set of real-time error signals includes an ion beam path deviation signal, an ion beam intensity fluctuation signal and an environmental interference signal; training an adaptive error compensation model according to the set of real-time error signals to generate a set of error compensation parameters; the adaptive error compensation model includes a dynamic weight adjustment layer and a compensation path generation layer, where the dynamic weight adjustment layer is used to adjust the correlation weight between error signals according to the frequency characteristics of the environmental interference signal, and the compensation path generation layer is used to generate ion beam path correction parameters based on the adjusted correlation weight; inputting the set of error compensation parameters into a control unit of the focused ion beam device to generate a set of real-time compensation control instructions; adjusting the ion source parameters and deflection electrode voltage parameters of the focused ion beam device according to the set of real-time compensation control instructions, and outputting an adjusted set of ion beam path parameters; performing ion beam processing on a target processing area according to the adjusted set of ion beam path parameters, and monitoring the morphological feature data of the processed area in real time to verify the error compensation effect.

[0005] In a second aspect, an embodiment of the present invention provides an error compensation system, including: a memory in which a computer program is stored; and a processor configured to load the computer program to implement the adaptive error compensation method for a focused ion beam device as described above.

[0006] The adaptive error compensation method for a focused ion beam device provided by the present invention obtains a set of real-time error signals generated during the processing of the focused ion beam device (including ion beam path deviation signals, ion beam intensity fluctuation signals, and environmental interference signals), adaptively adjusts the correlation weights of the error signals based on the dynamic weight adjustment layer according to the frequency characteristics of the environmental interference signals, generates ion beam path correction parameters adapted to the weights through the compensation path generation layer, inputs the error compensation parameter set into the device control unit in real time to generate compensation instructions, synchronously adjusts the ion source parameters and the deflection electrode voltage parameters to output the corrected ion beam path parameters, and finally verifies the compensation effect in combination with the real-time topography feature data. It can effectively integrate the dynamic characteristics of multi-source error signals and the physical relevance of device control parameters, realize the collaborative optimization of external environmental interference suppression and internal path deviation correction, thereby significantly improving the ion beam processing accuracy and stability; at the same time, through the differential processing of the fast response of the dynamic weight adjustment layer to high-frequency interference signals and the gradual correction of low-frequency drift signals, the anti-interference ability under complex working conditions is enhanced, and the closed-loop feedback optimization of the compensation model based on the real-time topography feature data further ensures the self-adaptability and long-term reliability of the compensation strategy. Finally, while reducing the ion beam path deviation, the generation delay of the compensation instructions is shortened, taking into account both the processing efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 is a flowchart of an adaptive error compensation method for a focused ion beam device provided by an embodiment of the present invention.

[0008] Figure 2 is a schematic diagram of the composition of an error compensation system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0009] Please refer to Figure 1 , Figure 1 which is a flowchart of an adaptive error compensation method for a focused ion beam device provided by an embodiment of the present invention. The adaptive error compensation method for a focused ion beam device can be executed by an error compensation system. The adaptive error compensation method for a focused ion beam device may include the following steps:

[0010] Step S100: Obtain a set of real-time error signals generated during the processing of the focused ion beam device, where the set of real-time error signals includes ion beam path deviation signals, ion beam intensity fluctuation signals, and environmental interference signals.

[0011] During the processing of a focused ion beam device, the set of real-time error signals is an important basis for reflecting the operating state and processing accuracy of the device. The ion beam path deviation signal refers to the signal that the ion beam deviates from the preset path during transmission, which may be caused by various factors, such as the instability of the ion source, the interference of the electric or magnetic field, etc. The ion beam intensity fluctuation signal represents the unstable change of the ion beam intensity during processing, which may affect the processing depth and quality. The environmental interference signal refers to the signal generated by the interference factors from the external environment of the device, such as temperature change, electromagnetic interference, etc., and these interferences will have an adverse impact on the transmission and processing effect of the ion beam.

[0012] The acquisition of these sets of real-time error signals can be achieved through various sensors. For example, for the ion beam path deviation signal, a position sensor can be used to monitor the actual position of the ion beam and compare it with the preset path to obtain the deviation signal. For the ion beam intensity fluctuation signal, a beam current monitor can be used to measure the ion beam intensity in real time and record its fluctuation. For the environmental interference signal, devices such as temperature sensors and electromagnetic sensors can be used to detect the temperature change and electromagnetic interference in the environment and convert them into corresponding signals.

[0013] Step S200: Train an adaptive error compensation model according to the set of real-time error signals to generate a set of error compensation parameters; the adaptive error compensation model includes a dynamic weight adjustment layer and a compensation path generation layer, where the dynamic weight adjustment layer is used to adjust the correlation weight between error signals according to the frequency characteristics of the environmental interference signal, and the compensation path generation layer is used to generate ion beam path correction parameters based on the adjusted correlation weight.

[0014] The adaptive error compensation model can automatically adjust the error compensation parameters according to the set of real-time error signals to improve the processing accuracy of the focused ion beam device. The model consists of a dynamic weight adjustment layer and a compensation path generation layer.

[0015] The main function of the dynamic weight adjustment layer is to adjust the correlation weights between error signals according to the frequency characteristics of environmental interference signals. Environmental interference signals of different frequencies may have different degrees of influence on the ion beam path deviation signal and the ion beam intensity fluctuation signal. Therefore, it is necessary to dynamically adjust the correlation weights between them according to the frequency characteristics. For example, for high-frequency environmental interference signals, they may have a greater impact on the ion beam path deviation signal. At this time, the weight of the ion beam path deviation signal can be increased; for low-frequency environmental interference signals, they may have a greater impact on the ion beam intensity fluctuation signal. At this time, the weight of the ion beam intensity fluctuation signal can be increased. The compensation path generation layer generates ion beam path correction parameters based on the adjusted correlation weights. After obtaining the adjusted correlation weights, the compensation path generation layer will comprehensively analyze the error signals according to these weights, so as to generate parameters that can effectively correct the ion beam path deviation. These parameters will be used for subsequent error compensation operations to ensure that the ion beam can accurately reach the target processing area.

[0016] Machine learning algorithms, such as neural network algorithms like multi-layer perceptron (MLP), can be used to train the adaptive error compensation model. During the training process, the real-time error signal set is used as the input, and the error compensation parameter set is used as the output. By continuously adjusting the parameters of the model, the model can accurately generate appropriate compensation parameters according to the input error signals.

[0017] As an implementation manner, step S200, training the adaptive error compensation model according to the real-time error signal set to generate the error compensation parameter set, may specifically include the following steps:

[0018] Step S210: Obtain the error signal distribution pattern in the historical error signal set, and determine the dynamic compensation interval according to the error signal distribution pattern; the dynamic compensation interval includes the maximum fluctuation threshold of the environmental interference signal and the minimum correction step of the ion beam path deviation signal.

[0019] The historical error signal set is the error signal data recorded by the focused ion beam device during past processing. These data contain the distribution pattern information of the error signals. The error signal distribution pattern refers to the distribution law of the error signals in time and space. By analyzing the historical error signal set, these distribution laws can be obtained.

[0020] The dynamic compensation interval is an interval determined according to the error signal distribution pattern, which includes the maximum fluctuation threshold of the environmental interference signal and the minimum correction step of the ion beam path deviation signal. The maximum fluctuation threshold of the environmental interference signal refers to the maximum allowable fluctuation range of the environmental interference signal under normal circumstances. When the environmental interference signal exceeds this threshold, it may have a greater impact on ion beam processing and corresponding compensation is required. The minimum correction step of the ion beam path deviation signal refers to the minimum allowable adjustment amount each time when correcting the ion beam path deviation. The determination of this step can ensure the stability and accuracy of the correction process.

[0021] Obtaining the error signal distribution pattern in the historical error signal set can be achieved through data mining and analysis techniques. For example, statistical analysis methods can be used to process the historical error signal set to obtain statistical features such as the mean, variance, and frequency distribution of the error signals, thereby determining the distribution pattern of the error signals. Then, based on these distribution patterns and combined with the actual processing requirements and equipment performance, the specific parameters of the dynamic compensation interval are determined.

[0022] As an implementation manner, in step S210, obtaining the error signal distribution pattern in the historical error signal set and determining the dynamic compensation interval according to the error signal distribution pattern may specifically include the following steps:

[0023] Step S211: Collect the historical error signal set stored by the focused ion beam equipment in the past processing cycle, and perform timestamp alignment processing on the ion beam path deviation signal and the environmental interference signal in the historical error signal set to generate a time-synchronized error signal sequence.

[0024] The historical error signal set is the error signal data recorded by the focused ion beam equipment during past processing. These data contain information such as the ion beam path deviation signal and the environmental interference signal. Since these signals may be collected at different time points, timestamp alignment processing is required to ensure that they are synchronized in time.

[0025] Timestamp alignment processing refers to matching and aligning the ion beam path deviation signal and the environmental interference signal according to the timestamp information of the signals, so that they have corresponding values at the same time point. Through timestamp alignment processing, a time-synchronized error signal sequence can be generated, which is convenient for subsequent analysis and processing.

[0026] Collecting the historical error signal set can be achieved through the data storage system of the equipment. The equipment will record the error signal data during each processing and store it in a preset database. When the historical error signal set needs to be obtained, the corresponding data can be read from the database.

[0027] Step S212: Perform multi-scale decomposition on the time synchronization error signal sequence to separate out the high-frequency perturbation component, intermediate-frequency fluctuation component, and low-frequency drift component.

[0028] Multi-scale decomposition can decompose a complex signal into components of different scales. For the time synchronization error signal sequence, through multi-scale decomposition, the high-frequency perturbation component, intermediate-frequency fluctuation component, and low-frequency drift component can be separated out. The high-frequency perturbation component is usually caused by sudden interferences within a short period of time, with a relatively high frequency and rapid changes. The intermediate-frequency fluctuation component is caused by periodic interferences or fluctuations, with a moderate frequency and periodicity. The low-frequency drift component is caused by slow changes over a long period of time, with a relatively low frequency and relatively stable changes. These components can be separated using multi-scale decomposition methods such as wavelet decomposition. Wavelet decomposition is a commonly used signal processing method that can decompose a signal at different scales to obtain components of different frequencies.

[0029] Step S213: For the high-frequency perturbation component, extract its peak occurrence frequency and amplitude attenuation rate as the first distribution feature; for the intermediate-frequency fluctuation component, extract its periodic repetition interval and phase offset as the second distribution feature; for the low-frequency drift component, extract its trend change slope and cumulative deviation as the third distribution feature.

[0030] For the separated high-frequency perturbation component, intermediate-frequency fluctuation component, and low-frequency drift component, their characteristic information needs to be extracted for subsequent analysis and processing.

[0031] For the high-frequency perturbation component, the peak occurrence frequency refers to the number of times the peak appears within a set time, which reflects the frequency of high-frequency perturbations. The amplitude attenuation rate refers to the attenuation speed of the amplitude of the high-frequency perturbation signal over time, which reflects the attenuation characteristics of high-frequency perturbations. Extracting these two features can help us understand the variation law of high-frequency perturbations.

[0032] For the intermediate-frequency fluctuation component, the periodic repetition interval refers to the period length of the intermediate-frequency fluctuation signal, which reflects the periodicity of intermediate-frequency fluctuations. The phase offset refers to the phase difference between the intermediate-frequency fluctuation signal and the reference signal, which reflects the phase characteristics of intermediate-frequency fluctuations. Extracting these two features can help us understand the periodicity and phase relationship of intermediate-frequency fluctuations.

[0033] For the low-frequency drift component, the trend change slope refers to the slope of the change trend of the low-frequency drift signal, which reflects the change speed of low-frequency drift. The cumulative deviation refers to the cumulative deviation value of the low-frequency drift signal over a period of time, which reflects the cumulative effect of low-frequency drift. Extracting these two features can help us understand the change trend and cumulative impact of low-frequency drift.

[0034] Step S214: According to the first distribution feature, the second distribution feature, and the third distribution feature, fit and generate a composite waveform template in the error signal distribution pattern. The composite waveform template includes a high-frequency disturbance envelope, an intermediate-frequency fluctuation baseline, and a low-frequency drift trend line.

[0035] After extracting the characteristic information of the high-frequency disturbance component, the intermediate-frequency fluctuation component, and the low-frequency drift component, a composite waveform template in the error signal distribution pattern can be generated by fitting according to these characteristic information. The composite waveform template is a waveform that comprehensively reflects the error signal distribution pattern and is composed of a high-frequency disturbance envelope, an intermediate-frequency fluctuation baseline, and a low-frequency drift trend line.

[0036] The high-frequency disturbance envelope refers to the maximum amplitude change curve of the high-frequency disturbance component, which reflects the maximum influence range of the high-frequency disturbance. The intermediate-frequency fluctuation baseline refers to the average amplitude change curve of the intermediate-frequency fluctuation component, which reflects the average level of the intermediate-frequency fluctuation. The low-frequency drift trend line refers to the change trend curve of the low-frequency drift component, which reflects the long-term change trend of the low-frequency drift. Curve fitting algorithms such as the least squares method can be used to fit and generate the composite waveform template. Through curve fitting, the curve that best suits these characteristic information can be found, thereby generating the composite waveform template.

[0037] Step S215: Determine the maximum fluctuation threshold of the environmental interference signal based on the maximum coverage range of the high-frequency disturbance envelope, determine the minimum correction step of the ion beam path deviation signal based on the slope change rate of the low-frequency drift trend line, and mark the intersection area of the maximum fluctuation threshold and the minimum correction step as the dynamic compensation interval.

[0038] The maximum coverage range of the high-frequency disturbance envelope represents the maximum influence range of the high-frequency disturbance on the environmental interference signal. Based on this maximum coverage range, the maximum fluctuation threshold of the environmental interference signal can be determined. When the fluctuation of the environmental interference signal exceeds this threshold, it indicates that the influence of the environmental interference on the ion beam processing is relatively large and corresponding compensation is required.

[0039] The slope change rate of the low-frequency drift trend line reflects the influence speed of the low-frequency drift on the ion beam path deviation signal. Based on this slope change rate, the minimum correction step of the ion beam path deviation signal can be determined. The determination of the minimum correction step can ensure that when correcting the ion beam path deviation, the adjustment amplitude each time will not be too large or too small, thereby improving the accuracy and stability of the correction.

[0040] Mark the intersection area of the maximum fluctuation threshold and the minimum correction step as the dynamic compensation interval. The dynamic compensation interval is an interval that comprehensively considers the environmental interference signal and the ion beam path deviation signal. Performing error compensation within this interval can more effectively improve the processing accuracy of the focused ion beam device.

[0041] Step S216: Optimize the boundaries of the dynamic compensation interval using the validation dataset, and update the high-frequency disturbance envelope and low-frequency drift trend line in the error signal distribution pattern according to the optimized interval parameters to generate the effective time range and spatial scope of the final dynamic compensation interval.

[0042] The validation dataset is a dataset with known error conditions, which can be used to verify the effectiveness and accuracy of the dynamic compensation interval. By applying the dynamic compensation interval to the validation dataset and analyzing the verification results, the boundaries of the dynamic compensation interval can be optimized.

[0043] Boundary optimization refers to adjusting the upper and lower boundaries of the dynamic compensation interval according to the verification results, so that the dynamic compensation interval can better adapt to different error conditions. For example, if it is found that the upper boundary of the dynamic compensation interval is set too high, resulting in some error conditions that actually need to be compensated not being covered, then the upper boundary can be appropriately lowered; if it is found that the lower boundary is set too low, resulting in some situations that do not need to be compensated also being compensated, then the lower boundary can be appropriately raised.

[0044] Update the high-frequency disturbance envelope and low-frequency drift trend line in the error signal distribution pattern according to the optimized interval parameters. The high-frequency disturbance envelope and low-frequency drift trend line are important components of the error signal distribution pattern, and their update can more accurately reflect the distribution of the error signal.

[0045] Finally, generate the effective time range and spatial scope of the final dynamic compensation interval. The effective time range refers to the time period during which the dynamic compensation interval is effective, and the spatial scope refers to the spatial range within which the dynamic compensation interval is effective. Determining the effective time range and spatial scope can ensure the targeted and effective application of the dynamic compensation interval.

[0046] Step S220: Extract the frequency characteristics of the current environmental interference signal from the set of real-time error signals, and generate a weight adjustment coefficient according to the matching degree between the frequency characteristics and the dynamic compensation interval.

[0047] The current environmental interference signal in the set of real-time error signals contains rich frequency characteristic information, and these characteristics reflect the variation law of the environmental interference signal. By performing time-frequency conversion processing on the current environmental interference signal, a spectrogram containing instantaneous frequency components can be separated. In the spectrogram, the effective frequency bands higher than the preset background noise threshold can be identified, and these effective frequency bands can be divided into high-frequency pulse intervals, intermediate-frequency oscillation intervals, and low-frequency slow-varying intervals.

[0048] The matching degree between the frequency characteristics and the dynamic compensation interval refers to the similarity between the frequency characteristics of the current environmental interference signal and the preset frequency characteristics in the dynamic compensation interval. By calculating the first coincidence degree between the peak density of the high-frequency pulse interval and the coverage range of the high-frequency disturbance envelope in the dynamic compensation interval, the second coincidence degree between the periodic fluctuation amplitude of the intermediate-frequency oscillation interval and the phase tolerance of the intermediate-frequency fluctuation baseline in the dynamic compensation interval, and the third coincidence degree between the trend slope of the low-frequency slow-varying interval and the prediction deviation of the low-frequency drift trend line in the dynamic compensation interval, three coincidence degree indexes can be obtained.

[0049] According to these three coincidence degree indexes, a high-frequency matching factor, an intermediate-frequency matching factor, and a low-frequency matching factor are generated respectively. Multiply the high-frequency matching factor by the weight ratio of the maximum fluctuation threshold in the dynamic compensation interval, multiply the intermediate-frequency matching factor by the stability coefficient of the periodic repetition interval in the dynamic compensation interval, and multiply the low-frequency matching factor by the response priority of the minimum correction step in the dynamic compensation interval. Then, perform normalized weighted summation on the multiplied high-frequency matching factor, intermediate-frequency matching factor, and low-frequency matching factor to generate the dynamic environment adaptation value and the steady-state interference suppression value in the weight adjustment coefficient. Finally, based on the proportional relationship between the dynamic environment adaptation value and the steady-state interference suppression value, adjust the fusion priority of the weight adjustment coefficient for the ion beam path deviation signal and the ion beam intensity fluctuation signal to generate a real-time weight allocation strategy that matches the effective time range and spatial scope in the dynamic compensation interval.

[0050] As an implementation manner, in step S220, extracting the frequency characteristics of the current environmental interference signal from the real-time error signal set and generating a weight adjustment coefficient according to the matching degree between the frequency characteristics and the dynamic compensation interval may specifically include the following steps:

[0051] Step S221: Perform time-frequency conversion processing on the environmental interference signal in the real-time error signal set to separate the spectral distribution map containing instantaneous frequency components.

[0052] Time-frequency conversion processing is a signal processing technology that converts a signal from the time domain to the frequency domain. It can decompose the environmental interference signal into different frequency components and show the variation of these components over time. By performing time-frequency conversion processing on the environmental interference signal in the real-time error signal set, the spectral distribution map containing instantaneous frequency components can be separated.

[0053] The spectrum distribution diagram is a two-dimensional image. The horizontal axis represents time, and the vertical axis represents frequency. The colors or grayscales in the image represent the amplitudes of different frequency components. Through the spectrum distribution diagram, the frequency characteristics of environmental interference signals can be intuitively observed, such as the occurrence time, frequency range, and amplitude of different frequency components. Time-frequency conversion processing can use methods such as the short-time Fourier transform (STFT) and wavelet transform. The short-time Fourier transform is a commonly used time-frequency analysis method. It obtains the time-frequency distribution of the signal by windowing the signal and then performing Fourier transform on the signal within each window. The wavelet transform is a more flexible time-frequency analysis method. It can select different wavelet basis functions according to the characteristics of the signal to better capture the time-frequency characteristics of the signal.

[0054] Step S222: Identify the effective frequency bands higher than the preset background noise threshold from the spectrum distribution diagram, and divide the effective frequency bands into a high-frequency pulse interval, an intermediate-frequency oscillation interval, and a low-frequency slow-varying interval.

[0055] The preset background noise threshold is a preset threshold used to distinguish the effective frequency components and noise components in the signal. In the spectrum distribution diagram, the frequency bands higher than the preset background noise threshold are considered effective frequency bands, and these frequency bands contain the main information of the environmental interference signal.

[0056] Dividing the effective frequency bands into a high-frequency pulse interval, an intermediate-frequency oscillation interval, and a low-frequency slow-varying interval is to better analyze and process environmental interference signals of different frequencies. The high-frequency pulse interval usually contains burst interference signals within a short time, with a high frequency and a short duration. The intermediate-frequency oscillation interval contains periodic interference signals, with a moderate frequency and periodicity. The low-frequency slow-varying interval contains long-term slowly changing signals, with a low frequency and a relatively stable change.

[0057] Identifying the effective frequency bands and dividing the intervals can be achieved by setting appropriate thresholds and frequency ranges. For example, a threshold can be set according to the amplitude values in the spectrum distribution diagram, and the frequency bands higher than this threshold are regarded as effective frequency bands; then, the effective frequency bands are divided into a high-frequency pulse interval, an intermediate-frequency oscillation interval, and a low-frequency slow-varying interval according to the frequency range.

[0058] Step S223: Calculate the first coincidence degree between the peak density of the high-frequency pulse interval and the coverage range of the high-frequency disturbance envelope in the dynamic compensation interval, calculate the second coincidence degree between the periodic fluctuation amplitude of the intermediate-frequency oscillation interval and the phase tolerance of the intermediate-frequency fluctuation baseline in the dynamic compensation interval, and calculate the third coincidence degree between the trend slope of the low-frequency slow-varying interval and the prediction deviation of the low-frequency drift trend line in the dynamic compensation interval.

[0059] The first coincidence degree refers to the coincidence degree between the peak density in the high-frequency pulse interval and the coverage range of the high-frequency disturbance envelope in the dynamic compensation interval. The peak density in the high-frequency pulse interval refers to the frequency at which peaks appear within the high-frequency pulse interval, which reflects the frequency of high-frequency pulses. The coverage range of the high-frequency disturbance envelope in the dynamic compensation interval represents the maximum influence range of high-frequency disturbances. By calculating the coincidence degree between the two, the matching degree between the high-frequency pulse interval and the high-frequency disturbances in the dynamic compensation interval can be understood.

[0060] The second coincidence degree refers to the coincidence degree between the periodic fluctuation amplitude in the intermediate-frequency oscillation interval and the phase tolerance of the intermediate-frequency fluctuation baseline in the dynamic compensation interval. The periodic fluctuation amplitude in the intermediate-frequency oscillation interval refers to the maximum fluctuation amplitude of the intermediate-frequency oscillation signal, which reflects the intensity of intermediate-frequency oscillation. The phase tolerance of the intermediate-frequency fluctuation baseline in the dynamic compensation interval represents the allowable deviation range of the phase of intermediate-frequency fluctuations. By calculating the coincidence degree between the two, the matching degree between the intermediate-frequency oscillation interval and the intermediate-frequency fluctuations in the dynamic compensation interval can be understood.

[0061] The third coincidence degree refers to the coincidence degree between the trend slope in the low-frequency slow-varying interval and the prediction deviation of the low-frequency drift trend line in the dynamic compensation interval. The trend slope in the low-frequency slow-varying interval refers to the slope of the change trend of the low-frequency slow-varying signal, which reflects the change speed of low-frequency slow-varying. The prediction deviation of the low-frequency drift trend line in the dynamic compensation interval represents the prediction error range of the low-frequency drift trend line. By calculating the coincidence degree between the two, the matching degree between the low-frequency slow-varying interval and the low-frequency drift in the dynamic compensation interval can be understood.

[0062] Various similarity calculation methods can be used to calculate the coincidence degree, such as the ratio of intersection to union, correlation coefficient, etc. By calculating the coincidence degree, the matching degree between different frequency intervals and the dynamic compensation interval can be quantified, providing a basis for generating the weight adjustment coefficient subsequently.

[0063] Step S224: Generate a high-frequency matching factor, an intermediate-frequency matching factor, and a low-frequency matching factor according to the first coincidence degree, the second coincidence degree, and the third coincidence degree respectively, multiply the high-frequency matching factor by the weight ratio of the maximum fluctuation threshold in the dynamic compensation interval, multiply the intermediate-frequency matching factor by the stability coefficient of the periodic repetition interval in the dynamic compensation interval, and multiply the low-frequency matching factor by the response priority of the minimum correction step in the dynamic compensation interval.

[0064] The high-frequency matching factor, intermediate-frequency matching factor, and low-frequency matching factor respectively represent the matching degrees of the high-frequency pulse interval, intermediate-frequency oscillation interval, and low-frequency slow-varying interval with the dynamic compensation interval. Based on the first coincidence degree, second coincidence degree, and third coincidence degree, these three matching factors can be generated respectively. For example, the coincidence degree can be directly used as the matching factor, or the matching factor can be obtained after a preset transformation of the coincidence degree. The weight proportion of the maximum fluctuation threshold in the dynamic compensation interval represents the importance of the maximum fluctuation threshold in the entire dynamic compensation interval. Multiplying the high-frequency matching factor by the weight proportion of the maximum fluctuation threshold can obtain the weighted matching value of the high-frequency part. Similarly, multiplying the intermediate-frequency matching factor by the stability coefficient of the periodic repetition interval in the dynamic compensation interval can obtain the weighted matching value of the intermediate-frequency part; multiplying the low-frequency matching factor by the response priority of the minimum correction step in the dynamic compensation interval can obtain the weighted matching value of the low-frequency part. In this way, the matching degrees of different frequency intervals with the dynamic compensation interval and the importance of each parameter in the dynamic compensation interval can be comprehensively considered, providing a more accurate basis for generating the weight adjustment coefficient subsequently.

[0065] Step S225: Perform normalized weighted summation on the multiplied high-frequency matching factor, intermediate-frequency matching factor, and low-frequency matching factor to generate the dynamic environment adaptation value and steady-state interference suppression value in the weight adjustment coefficient.

[0066] Normalized weighted summation means performing weighted summation on the multiplied high-frequency matching factor, intermediate-frequency matching factor, and low-frequency matching factor, and normalizing the summation result so that the result is within a predefined range. Through normalized weighted summation, the matching information of different frequency intervals can be integrated to generate the dynamic environment adaptation value and steady-state interference suppression value in the weight adjustment coefficient.

[0067] The dynamic environment adaptation value represents the adaptation ability of the weight adjustment coefficient to dynamic environment changes. It reflects the ability of the weight adjustment coefficient to timely adjust the correlation weight between error signals under different environmental interference conditions. The steady-state interference suppression value represents the suppression ability of the weight adjustment coefficient to steady-state interference. It reflects the ability of the weight adjustment coefficient to effectively reduce the impact of interference on error compensation in the presence of steady-state interference. Various normalization methods can be used for normalization processing, such as dividing the summation result by the total sum, mapping the result to the [0,1] interval, etc. Through normalization processing, the dynamic environment adaptation value and steady-state interference suppression value can be made comparable and interpretable.

[0068] Step S226: Based on the proportional relationship between the dynamic environment adaptation value and the steady-state interference suppression value, adjust the fusion priority of the weight adjustment coefficient for the ion beam path deviation signal and the ion beam intensity fluctuation signal to generate a real-time weight allocation strategy that matches the effective time range and spatial scope of action in the dynamic compensation interval.

[0069] The proportional relationship between the dynamic environment adaptation value and the steady-state interference suppression value reflects the dynamic change degree of the current environmental interference and the intensity of the steady-state interference. Based on this proportional relationship, the weight adjustment coefficient can be adjusted to prioritize the fusion of the ion beam path deviation signal and the ion beam intensity fluctuation signal.

[0070] For example, if the dynamic environment adaptation value is large, it indicates that the dynamic change of the current environmental interference is large. At this time, the weight of the ion beam path deviation signal can be increased because the ion beam path deviation signal may be more easily affected by the dynamic environmental change; if the steady-state interference suppression value is large, it indicates that there is a strong steady-state interference currently. At this time, the weight of the ion beam intensity fluctuation signal can be increased because the ion beam intensity fluctuation signal may be more easily affected by the steady-state interference.

[0071] According to the adjusted fusion priority, a real-time weight allocation strategy matching the effective time range and spatial scope in the dynamic compensation interval is generated. The real-time weight allocation strategy means that in different time and space ranges, different weights are assigned to the ion beam path deviation signal and the ion beam intensity fluctuation signal to achieve more accurate error compensation. For example, within the effective time range of the dynamic compensation interval, the weight is dynamically adjusted according to the change of the environmental interference; within different spatial scopes, different weights are assigned according to the processing requirements and error conditions of the ion beam.

[0072] Step S230: Input the weight adjustment coefficient into the dynamic weight adjustment layer to perform weighted fusion on the ion beam path deviation signal and the ion beam intensity fluctuation signal, and generate a fused error signal sequence.

[0073] After obtaining the weight adjustment coefficient, input it into the dynamic weight adjustment layer. The dynamic weight adjustment layer will perform weighted fusion on the ion beam path deviation signal and the ion beam intensity fluctuation signal according to the weight adjustment coefficient. Weighted fusion means that different weights are assigned to the ion beam path deviation signal and the ion beam intensity fluctuation signal according to the weight adjustment coefficient, and then they are linearly combined to obtain a fused error signal sequence.

[0074] Step S240: Determine the path correction parameter generation strategy in the compensation path generation layer according to the mapping relationship between the fused error signal sequence and the preset error compensation rule set.

[0075] The preset error compensation rule set is a series of rules formulated in advance, and these rules stipulate the compensation methods and parameters corresponding to different error situations. The mapping relationship between the fused error signal sequence and the preset error compensation rule set means that according to the characteristics of the fused error signal sequence, the corresponding rules and parameters are found in the preset error compensation rule set.

[0076] Specifically, first, obtain the timestamp marking data in the fusion error signal sequence, and divide the signal processing window according to the timestamp marking data. Within each signal processing window, determine the initial compensation direction based on the cumulative offset of the ion beam path deviation signal, and determine the compensation intensity gradient based on the amplitude change rate of the ion beam intensity fluctuation signal.

[0077] Input the initial compensation direction and the compensation intensity gradient into the direction-intensity coupling module in the compensation path generation layer to generate a direction correction vector and an intensity correction vector. The direction correction vector represents the direction in which the ion beam path needs to be corrected, and the intensity correction vector represents the amplitude by which the ion beam intensity needs to be adjusted.

[0078] Generate a dynamic compensation factor according to the product of the direction correction vector and the intensity correction vector, and perform a superposition calculation on the dynamic compensation factor and the preset reference compensation parameter to generate a window-level compensation parameter. Finally, smooth the window-level compensation parameters of multiple consecutive signal processing windows to generate the time continuity constraint condition and the spatial distribution constraint condition in the path correction parameter generation strategy.

[0079] As an implementation manner, in step S240, determine the path correction parameter generation strategy in the compensation path generation layer according to the mapping relationship between the fusion error signal sequence and the preset error compensation rule set, which may specifically include the following steps:

[0080] Step S241: Obtain the timestamp marking data in the fusion error signal sequence, and divide the signal processing window according to the timestamp marking data.

[0081] The timestamp marking data in the fusion error signal sequence records the acquisition time of each error signal. According to these timestamp marking data, the fusion error signal sequence can be divided into multiple signal processing windows. The signal processing window is a time interval within which the error signals are centrally processed.

[0082] The purpose of dividing the signal processing window is to better analyze and process the error signals. In different time intervals, the characteristics of the error signals may be different. By dividing the signal processing window, the error signals within each window can be analyzed independently, so as to more accurately determine the path correction parameter generation strategy in the compensation path generation layer.

[0083] The length and overlap degree of the window can be set according to actual needs when dividing the signal processing window. For example, a window with a fixed length can be set, and there can be an overlap between each window to ensure the continuity and integrity of the signal.

[0084] Step S242: Within each signal processing window, determine the initial compensation direction based on the cumulative offset of the ion beam path deviation signal, and determine the compensation intensity gradient based on the rate of change of the amplitude of the ion beam intensity fluctuation signal.

[0085] Within each signal processing window, the cumulative offset of the ion beam path deviation signal represents the total distance by which the ion beam deviates from the preset path within that window. Based on this cumulative offset, the initial compensation direction can be determined. The initial compensation direction refers to the direction in which the ion beam needs to be adjusted in order to correct the ion beam path deviation.

[0086] The rate of change of the amplitude of the ion beam intensity fluctuation signal represents the rate at which the ion beam intensity changes within that window. Based on this rate of change of the amplitude, the compensation intensity gradient can be determined. The compensation intensity gradient refers to the rate of change of the amplitude by which the ion beam intensity needs to be adjusted over time in order to correct the ion beam intensity fluctuation.

[0087] Various analytical methods can be used to determine the initial compensation direction and the compensation intensity gradient, such as the least squares method, the gradient descent method, etc. By determining the initial compensation direction and the compensation intensity gradient, a basis can be provided for generating the direction correction vector and the intensity correction vector in the subsequent steps.

[0088] Step S243: Input the initial compensation direction and the compensation intensity gradient into the direction-intensity coupling module in the compensation path generation layer to generate a direction correction vector and an intensity correction vector.

[0089] The direction-intensity coupling module in the compensation path generation layer is a module specifically designed to handle the ion beam path deviation and the intensity fluctuation. After inputting the initial compensation direction and the compensation intensity gradient into this module, the module will generate a direction correction vector and an intensity correction vector according to the preset algorithms and models.

[0090] The direction correction vector represents the direction and amplitude by which the ion beam path needs to be corrected. It is a vector, the direction of which represents the correction direction, and the magnitude of which represents the correction amplitude. The intensity correction vector represents the amplitude and direction by which the ion beam intensity needs to be adjusted. It is also a vector, the direction of which represents the adjustment direction, and the magnitude of which represents the adjustment amplitude.

[0091] The direction-intensity coupling module can be implemented using algorithms such as neural networks and fuzzy logic. Through the direction-intensity coupling module, the effects of the ion beam path deviation and the intensity fluctuation can be comprehensively considered to generate more accurate direction correction vectors and intensity correction vectors.

[0092] Step S244: Generate a dynamic compensation factor based on the product of the direction correction vector and the intensity correction vector, and perform a superposition calculation of the dynamic compensation factor and the preset reference compensation parameter to generate a window-level compensation parameter.

[0093] The dynamic compensation factor is obtained by multiplying the direction correction vector and the intensity correction vector. The direction correction vector and the intensity correction vector respectively represent the correction information of the ion beam path and intensity. Their product can comprehensively consider the correction requirements of the path and intensity to obtain a more comprehensive dynamic compensation factor.

[0094] The preset reference compensation parameters are a set of compensation parameters set in advance. They are the compensation values required to correct errors under ideal conditions. By superimposing and calculating the dynamic compensation factor and the preset reference compensation parameters, the window-level compensation parameters can be obtained. The window-level compensation parameters are the compensation parameters required to correct errors within the current signal processing window.

[0095] The superimposing calculation can use simple addition operations. Through the superimposing calculation, the dynamic compensation factor and the preset reference compensation parameters can be combined to obtain more accurate window-level compensation parameters.

[0096] Step S245: Smooth the window-level compensation parameters of multiple consecutive signal processing windows to generate the time continuity constraint condition and the spatial distribution constraint condition in the path correction parameter generation strategy.

[0097] As an implementation manner, in step S245, smoothing the window-level compensation parameters of multiple consecutive signal processing windows to generate the time continuity constraint condition and the spatial distribution constraint condition in the path correction parameter generation strategy may specifically include the following steps:

[0098] Step S2451: Obtain the difference degree between the window-level compensation parameters of adjacent signal processing windows, and determine the parameter mutation threshold according to the difference degree.

[0099] The difference degree between the window-level compensation parameters of adjacent signal processing windows reflects the change of the compensation parameters in adjacent time intervals. Obtaining the difference degree can be achieved by calculating indexes such as the difference value and the change rate between adjacent window-level compensation parameters.

[0100] The parameter mutation threshold is a preset threshold used to judge whether the compensation parameters have mutated. Determining the parameter mutation threshold according to the difference degree can be adjusted according to the actual situation. For example, if the difference degree is large, it indicates that the change of the compensation parameters is relatively drastic. At this time, the parameter mutation threshold can be appropriately reduced to more sensitively detect parameter mutations; if the difference degree is small, it indicates that the change of the compensation parameters is relatively stable. At this time, the parameter mutation threshold can be appropriately increased to reduce the misjudgment of normal fluctuations.

[0101] Determining the parameter mutation threshold can use statistical analysis methods, such as calculating the mean value, standard deviation, etc. of the difference degree, and then determining the threshold according to these statistical indexes.

[0102] As an implementation manner, in order to obtain the difference degree between the window-level compensation parameters of adjacent signal processing windows and determine the parameter mutation threshold, it can be specifically divided into the following steps:

[0103] Step S24511: Extract the ion source current correction value sequence and the deflection electrode voltage correction value sequence in the window-level compensation parameters of adjacent signal processing windows, and perform time alignment processing on the sequences to generate synchronous compensation parameter pairs.

[0104] The window-level compensation parameters contain correction value information in multiple aspects. Among them, the ion source current correction value sequence and the deflection electrode voltage correction value sequence are parts closely related to the adjustment of key parameters in the ion beam processing process. The ion source current correction value sequence records the numerical sequence of the ion source current adjustment for correcting the ion beam intensity fluctuation within each signal processing window; the deflection electrode voltage correction value sequence records the numerical sequence of the deflection electrode voltage adjustment for correcting the ion beam path deviation. Since there may be differences in the acquisition times of adjacent signal processing windows, in order to accurately compare the differences between them, it is necessary to perform time alignment processing on these two sequences. The time alignment processing is achieved by matching the data at the corresponding time points in the two sequences, ensuring that the ion source current correction value and the deflection electrode voltage correction value at each time point can correspond one by one, thereby generating synchronous compensation parameter pairs. For example, if the acquisition time range of a signal processing window is from t1 to t2, and the acquisition time range of another adjacent window is from t2 to t3, through time alignment processing, the ion source current correction value and the deflection electrode voltage correction value corresponding to the same time point (such as t2) in the two windows can be combined into a synchronous compensation parameter pair. The purpose of this is to provide a basis for accurately calculating the difference degree between adjacent windows in the subsequent process.

[0105] Step S24512: Perform a rate-of-change analysis on the ion source current correction values in the synchronous compensation parameter pairs, and calculate the current change slope and the direction consistency index between adjacent windows; perform an amplitude difference analysis on the deflection electrode voltage correction values, and calculate the absolute value of the voltage difference and the polarity inversion frequency between adjacent windows.

[0106] For the ion source current correction values in the synchronous compensation parameter pairs, performing a rate-of-change analysis is to understand the change situation of the ion source current between adjacent signal processing windows. The current change slope reflects the change speed of the ion source current over time. By calculating the ratio of the difference between the ion source current correction values at the corresponding time points in adjacent windows to the time interval, the current change slope can be obtained. The direction consistency index is used to judge whether the change directions of the ion source current correction values in adjacent windows are consistent. For example, the direction consistency can be determined by comparing the positive and negative of the current changes in adjacent windows. If the current change directions in adjacent windows are the same, the direction consistency index is positive; otherwise, it is negative.

[0107] For the deflection electrode voltage correction value in the synchronous compensation parameter pair, the amplitude difference analysis is performed to evaluate the change amplitude and polarity change of the deflection electrode voltage between adjacent signal processing windows. Calculating the absolute value of the voltage difference between adjacent windows can intuitively reflect the magnitude of the voltage change. The larger the difference, the more drastic the voltage change. The polarity reversal frequency refers to the number of times the polarity of the deflection electrode voltage reverses in adjacent windows. Too high a polarity reversal frequency may cause instability in the ion beam path. By performing different types of analysis on the ion source current correction value and the deflection electrode voltage correction value, the changes in the window-level compensation parameters between adjacent signal processing windows can be fully understood from multiple angles.

[0108] Step S24513: input the current change slope, direction consistency index, voltage difference absolute value and polarity reversal frequency into the parameter difference evaluation model to generate a comprehensive difference index; the parameter difference evaluation model is trained based on the compensation parameter fluctuation range of historical adjacent windows, and is used to map the current change slope into a first difference component, the direction consistency index into a second difference component, the voltage difference absolute value into a third difference component, and the polarity reversal frequency into a fourth difference component.

[0109] The parameter difference evaluation model is a trained model. For example, the parameter difference evaluation model can adopt a linear regression model, a neural network model, etc. Its function is to combine multiple features (current change slope, direction consistency index, voltage difference absolute value and polarity reversal frequency) extracted from the synchronous compensation parameter pair to generate a comprehensive difference index that can fully reflect the degree of difference between window-level compensation parameters between adjacent signal processing windows. The model is trained based on the compensation parameter fluctuation range of historical adjacent windows, and learns the relationship between each feature and the comprehensive difference through a large amount of historical data. Inside the model, it maps the current change slope to the first difference component, which reflects the contribution of the ion source current change speed to the comprehensive difference; maps the direction consistency index to the second difference component, which reflects the influence of the ion source current change direction on the comprehensive difference; maps the voltage difference absolute value to the third difference component, which reflects the role of the deflection electrode voltage change amplitude; maps the polarity reversal frequency to the fourth difference component, which reflects the influence of the deflection electrode voltage polarity change on the comprehensive difference. Finally, the model comprehensively calculates these four difference components to generate a comprehensive difference index. For example, the model may adopt a weighted summation method to assign different weights to each difference component and then add them together to obtain a comprehensive difference index. The comprehensive difference index generated in this way can more accurately reflect the overall difference of the window-level compensation parameters between adjacent signal processing windows.

[0110] Step S24514: Determine the dynamic adjustment direction of the parameter mutation threshold according to the comparison result between the comprehensive difference index and the upper and lower limits of the preset difference reference interval; if the comprehensive difference index exceeds the upper limit of the difference reference interval, narrow the parameter mutation threshold to enhance the sensitivity to sudden anomalies; if the comprehensive difference index is lower than the lower limit of the difference reference interval, widen the parameter mutation threshold to reduce the false positive rate for normal fluctuations.

[0111] The preset difference reference interval is a pre-set range that represents the normal fluctuation range of the window-level compensation parameter between adjacent signal processing windows. The comprehensive difference index reflects the actual difference degree between the current adjacent windows. By comparing the comprehensive difference index with the upper and lower limits of the preset difference reference interval, it can be judged whether the current difference situation is within the normal range.

[0112] If the comprehensive difference index exceeds the upper limit of the difference reference interval, it indicates that the window-level compensation parameter between adjacent signal processing windows changes relatively violently, and there may be sudden anomaly situations. In order to detect these sudden anomalies more timely, it is necessary to narrow the parameter mutation threshold. The parameter mutation threshold is the standard for judging whether the compensation parameter has mutated. Narrowing this threshold means that it is easier to detect the change of the parameter, thereby enhancing the sensitivity to sudden anomalies. For example, originally the parameter mutation threshold is set to a relatively large value. When the comprehensive difference index exceeds the upper limit, the threshold is reduced, so that in subsequent processing, as long as the change of the compensation parameter is slightly larger, it will be determined that a mutation has occurred.

[0113] On the contrary, if the comprehensive difference index is lower than the lower limit of the difference reference interval, it indicates that the window-level compensation parameter between adjacent signal processing windows changes less and is within the normal fluctuation range. At this time, if the parameter mutation threshold is set too small, some normal fluctuations may be misjudged as parameter mutations, thereby increasing unnecessary processing and adjustments. Therefore, in order to reduce the false positive rate for normal fluctuations, it is necessary to widen the parameter mutation threshold. After widening the threshold, only when the change of the compensation parameter reaches the preset degree will it be determined that a mutation has occurred, reducing the overreaction to normal fluctuations.

[0114] Step S24515: Update the effective value of the parameter mutation threshold based on the dynamic adjustment direction, and associate and match the effective value with the change trend of the current window-level compensation parameter to generate a dynamic mutation detection condition adapted to the compensation timing sequence of the subsequent signal processing window.

[0115] After determining the dynamic adjustment direction of the parameter mutation threshold, it is necessary to update the effective value of the parameter mutation threshold according to this direction. If the parameter mutation threshold is to be reduced, the current parameter mutation threshold is decreased according to a preset ratio or a fixed value; if the parameter mutation threshold is to be increased, it is increased accordingly. The updated effective value will be used for subsequent mutation detection of the window-level compensation parameter.

[0116] To enable the parameter mutation threshold to better adapt to the actual compensation situation, it is also necessary to correlate and match the effective value with the change trend of the current window-level compensation parameter. The change trend of the current window-level compensation parameter can be determined by analyzing features such as the current change slope, direction consistency index, absolute value of voltage difference, and polarity inversion frequency obtained in the previous steps. For example, if the current window-level compensation parameter shows a gradually increasing trend, then when setting the parameter mutation threshold, its value can be appropriately adjusted to more accurately capture possible mutation situations.

[0117] Finally, based on the updated effective value of the parameter mutation threshold and the correlation matching result with the change trend of the current window-level compensation parameter, a dynamic mutation detection condition adapted to the compensation timing of the subsequent signal processing window is generated. The dynamic mutation detection condition is a set that includes the parameter mutation threshold and other relevant conditions, which will be used in the subsequent signal processing window to determine whether the window-level compensation parameter has mutated. The dynamically generated mutation detection condition can be adjusted according to the actual situation, better adapting to different compensation timings and parameter change situations, and improving the accuracy and effectiveness of mutation detection. In this way, during the subsequent error compensation process, mutations in the compensation parameter can be detected more timely and accurately, and corresponding measures can be taken for adjustment to ensure the processing accuracy and stability of the focused ion beam equipment.

[0118] Step S2452: If the difference degree exceeds the parameter mutation threshold, delay the current window-level compensation parameter and generate a temporary compensation parameter based on the moving average value of the previous window-level compensation parameter.

[0119] If the difference degree between the window-level compensation parameters of adjacent signal processing windows exceeds the parameter mutation threshold, it indicates that the compensation parameter has mutated. At this time, to avoid the impact of the mutation on error compensation, it is necessary to delay the current window-level compensation parameter.

[0120] The delay processing means postponing the usage time of the current window-level compensation parameter for a period of time to have more time to analyze and process the mutation situation. At the same time, a temporary compensation parameter is generated based on the moving average value of the previous window-level compensation parameter. The moving average value refers to the average value within a predefined time window, which can smooth the change of the compensation parameter and reduce the impact of mutations.

[0121] The generation of temporary compensation parameters can use a simple moving average algorithm. For example, for a moving window of length n, the temporary compensation parameter can be expressed as the average of the first n window-level compensation parameters.

[0122] Step S2453: Weightedly fuse the temporary compensation parameter with the current window-level compensation parameter to generate a sequence of transitional compensation parameters.

[0123] The weighted fusion of the temporary compensation parameter with the current window-level compensation parameter is to utilize the information of the current window-level compensation parameter as much as possible while ensuring the continuity of the compensation parameter. Weighted fusion means assigning different weights to the temporary compensation parameter and the current window-level compensation parameter, and then linearly combining them to obtain a sequence of transitional compensation parameters.

[0124] Step S2454: Determine the maximum allowable deviation rate and the minimum transitional step size in the time continuity constraint condition according to the parameter change trend in the sequence of transitional compensation parameters.

[0125] The parameter change trend in the sequence of transitional compensation parameters reflects the change of the compensation parameter over time. Based on this change trend, the maximum allowable deviation rate and the minimum transitional step size in the time continuity constraint condition can be determined.

[0126] The maximum allowable deviation rate refers to the maximum change rate that the compensation parameter is allowed within a preset time. It is used to limit the change speed of the compensation parameter to ensure the stability and continuity of the compensation process. The minimum transitional step size refers to the minimum adjustment amount that the compensation parameter is allowed between adjacent time points. It is used to ensure that the change of the compensation parameter is gradual and avoid sudden changes.

[0127] Statistical analysis methods can be used to determine the maximum allowable deviation rate and the minimum transitional step size, such as calculating the change rate, standard deviation, etc. of the sequence of transitional compensation parameters, and then determining the thresholds based on these statistical indicators.

[0128] Step S2455: Extract the regional compensation priority and the compensation direction consistency index in the spatial distribution constraint condition based on the distribution characteristics of the sequence of transitional compensation parameters in the spatial dimension.

[0129] Based on the distribution characteristics of the sequence of transitional compensation parameters in the spatial dimension, analyze the distribution of the compensation parameter in different processing regions. The spatial distribution characteristics can include the gradient, mean, variance, etc. of the parameter. Extract the regional compensation priority and the compensation direction consistency index in the spatial distribution constraint condition. The regional compensation priority refers to determining different compensation priorities in different processing regions according to the error situation and processing requirements. The compensation direction consistency index is used to measure the degree of consistency of the compensation direction in different regions.

[0130] As an implementation manner, in step S2455, based on the distribution characteristics of the transition compensation parameter sequence in the spatial dimension, extract the regional compensation priority and the compensation direction consistency index in the spatial distribution constraint conditions, which may specifically include the following steps:

[0131] Step S24551: Map the transition compensation parameter sequence to the spatial coordinate grid of the target machining area to generate a grid compensation parameter distribution map.

[0132] The spatial coordinate grid of the target machining area is a coordinate system that divides the target machining area into multiple small grids. Mapping the transition compensation parameter sequence to this spatial coordinate grid can correspond the compensation parameters to the specific positions of the target machining area and generate a grid compensation parameter distribution map.

[0133] The mapping process can be implemented according to the position information in the transition compensation parameter sequence and the definition of the spatial coordinate grid. For example, if the transition compensation parameter sequence contains the position coordinates corresponding to each compensation parameter, then these coordinates can be matched with the spatial coordinate grid to assign the compensation parameters to the corresponding grids. The grid compensation parameter distribution map can intuitively display the spatial distribution of the compensation parameters in the target machining area and provide a basis for subsequent extraction of the regional compensation priority and the compensation direction consistency index.

[0134] Step S24552: Determine the high error sensitive area and the low error sensitive area according to the parameter gradient difference of each grid node in the grid compensation parameter distribution map.

[0135] The parameter gradient difference of each grid node in the grid compensation parameter distribution map reflects the change of the compensation parameters in space. The area with a large parameter gradient difference indicates that the change of the compensation parameters is relatively drastic, and these areas may be high error sensitive areas; the area with a small parameter gradient difference indicates that the change of the compensation parameters is relatively stable, and these areas may be low error sensitive areas.

[0136] To determine the high error sensitive area and the low error sensitive area, a gradient calculation method can be used. For example, calculate the parameter gradient of each grid node, and then classify according to the magnitude of the gradient. A threshold can be set to divide the area with a gradient greater than the threshold into the high error sensitive area and the area with a gradient less than the threshold into the low error sensitive area.

[0137] Step S24553: Assign a first priority weight to the high error sensitive area and a second priority weight to the low error sensitive area.

[0138] In order to perform error compensation more effectively, it is necessary to assign different priority weights to high-error-sensitive regions and low-error-sensitive regions. The first priority weight refers to the weight assigned to the high-error-sensitive region, which indicates a relatively high importance of the high-error-sensitive region in error compensation; the second priority weight refers to the weight assigned to the low-error-sensitive region, which indicates a relatively low importance of the low-error-sensitive region in error compensation. The assignment of priority weights can be adjusted according to the actual situation. For example, the magnitude of the weight can be determined based on factors such as the area ratio of the high-error-sensitive region and the low-error-sensitive region, and the degree of error influence.

[0139] Step S24554: Calculate the consistency offset of the compensation direction based on the boundary overlap degree between the high-error-sensitive region and the low-error-sensitive region.

[0140] The boundary overlap degree between the high-error-sensitive region and the low-error-sensitive region reflects the spatial overlap situation of these two regions. Based on this overlap degree, the consistency offset of the compensation direction can be calculated.

[0141] The consistency offset of the compensation direction refers to the degree of inconsistency of the compensation direction at the boundary overlap of the high-error-sensitive region and the low-error-sensitive region. Calculating the consistency offset can be achieved by comparing the compensation directions of the high-error-sensitive region and the low-error-sensitive region at the boundary overlap. For example, the included angle of the compensation directions of the two regions at the boundary overlap can be calculated, and then the consistency offset can be determined based on the magnitude of the included angle.

[0142] Step S24555: Generate the dynamic adjustment coefficient and the direction locking condition in the compensation direction consistency index based on the comparison result between the consistency offset and the preset direction tolerance threshold.

[0143] Based on the comparison result between the consistency offset and the preset direction tolerance threshold, determine whether the consistency of the compensation direction meets the requirements. The preset direction tolerance threshold is a threshold set in advance for judging whether the consistency of the compensation direction is within an acceptable range.

[0144] Generate the dynamic adjustment coefficient and the direction locking condition in the compensation direction consistency index. The dynamic adjustment coefficient is used to adjust the magnitude of the compensation parameter according to the consistency of the compensation direction, and the direction locking condition is used to limit the change range of the compensation direction. For example, if the consistency offset exceeds the preset direction tolerance threshold, it indicates that the consistency of the compensation direction is poor. At this time, the dynamic adjustment coefficient can be increased to strengthen the adjustment of the compensation direction; at the same time, according to the situation of the consistency offset, the direction locking condition is determined, such as the allowable deviation direction range and the minimum locking maintenance duration.

[0145] As an implementation manner, in step S24555, based on the comparison result between the consistency offset and the preset direction tolerance threshold, generate the dynamic adjustment coefficient and the direction locking condition in the compensation direction consistency index, which may specifically include the following steps:

[0146] Step S245551: Extract the regional distribution vector of the consistency offset in the spatial coordinate grid, and obtain the allowable deviation angle range and the maximum allowable deviation distance corresponding to the preset direction tolerance threshold.

[0147] The regional distribution vector of the consistency offset in the spatial coordinate grid represents the spatial distribution of the consistency offset in the target machining area. Extracting this regional distribution vector can be achieved according to the values of the consistency offset at each grid node and the position information of the grid nodes.

[0148] The preset direction tolerance threshold is a preset threshold used to determine whether the consistency of the compensation direction meets the requirements. The allowable deviation angle range and the maximum allowable deviation distance corresponding to the preset direction tolerance threshold refer to the deviation angle and distance of the compensation direction that are allowed within this threshold range. Obtaining the allowable deviation angle range and the maximum allowable deviation distance can be set according to the actual situation.

[0149] Step S245552: Calculate the included angle between the offset direction of each grid node in the regional distribution vector and the preset target machining direction to generate a set of local direction deviation angles; at the same time, calculate the ratio of the grid node offset distance to the maximum allowable deviation distance to generate a set of local distance deviation ratios.

[0150] Calculating the included angle between the offset direction of each grid node in the regional distribution vector and the preset target machining direction can obtain the local direction deviation angle of each grid node. These local direction deviation angles form a set of local direction deviation angles. Calculating the ratio of the grid node offset distance to the maximum allowable deviation distance can obtain the local distance deviation ratio of each grid node. These local distance deviation ratios form a set of local distance deviation ratios.

[0151] Step S245553: Determine the direction consistency compliance area and the direction mismatch area according to the comparison result between each angle value in the set of local direction deviation angles and the allowable deviation angle range; determine the distance compliance area and the distance over-standard area according to the comparison result between each ratio value in the set of local distance deviation ratios and the preset ratio threshold.

[0152] Compare each angle value in the set of local direction deviation angles with the allowable deviation angle range. If the angle value is within the allowable deviation angle range, the area where the grid node is located is an area where the direction consistency meets the standard; if the angle value exceeds the allowable deviation angle range, the area where the grid node is located is an area with direction mismatch. Compare each ratio value in the set of local distance deviation ratios with the preset ratio threshold. If the ratio value is less than the preset ratio threshold, the area where the grid node is located is an area where the distance complies; if the ratio value is greater than the preset ratio threshold, the area where the grid node is located is an area where the distance exceeds the standard. By determining the area where the direction consistency meets the standard, the area with direction mismatch, the area where the distance complies, and the area where the distance exceeds the standard, the deviation of the compensation direction in the target processing area can be understood more clearly.

[0153] Step S245554: Statistically calculate the overlapping area of the area with direction mismatch and the area where the distance exceeds the standard, and generate the direction sensitivity factor in the dynamic adjustment coefficient according to the ratio of the overlapping area to the total area of the target processing area; the direction sensitivity factor is positively correlated with the ratio and is used to increase the adjustment range of the compensation direction proportionally.

[0154] Statistically calculating the overlapping area of the area with direction mismatch and the area where the distance exceeds the standard can be achieved by calculating the number of overlapping grid nodes of these two areas in the spatial coordinate grid and then multiplying by the area of each grid node. Generate the direction sensitivity factor in the dynamic adjustment coefficient according to the ratio of the overlapping area to the total area of the target processing area. The direction sensitivity factor is positively correlated with the ratio, that is, the larger the ratio, the larger the direction sensitivity factor. The direction sensitivity factor is used to increase the adjustment range of the compensation direction proportionally to improve the consistency of the compensation direction.

[0155] Step S245555: According to the boundary characteristics of the largest continuous area in the area where the direction consistency meets the standard, extract the allowable deviation direction range and the minimum locking maintenance duration in the direction locking condition; the allowable deviation direction range is determined according to the boundary extension direction of the largest continuous area, and the minimum locking maintenance duration is calculated and generated according to the historical stable period of the largest continuous area.

[0156] The boundary characteristics of the largest continuous area in the area where the direction consistency meets the standard reflect the shape and range of this area. According to these boundary characteristics, the allowable deviation direction range and the minimum locking maintenance duration in the direction locking condition can be extracted.

[0157] The allowable deviation direction range refers to the range of allowable deviation of the compensation direction within this area. It is determined according to the boundary extension direction of the largest continuous area. For example, the allowable deviation direction range can be determined by calculating the tangent direction of the boundary.

[0158] The minimum locking maintenance duration refers to the shortest time during which the compensation direction needs to remain stable within this area. It is calculated based on the historical stable period of the maximum continuous area. For example, the stable time of this area over a certain period in the past can be statistically analyzed, and then the average value is taken as the minimum locking maintenance duration.

[0159] Extracting the allowable deviation direction range and the minimum locking maintenance duration can help us better control the stability and consistency of the compensation direction during the error compensation process.

[0160] Step S245556: Correlate and map the direction sensitivity factor in the dynamic adjustment coefficient with the allowable deviation direction range in the direction locking condition to generate a dynamic direction constraint rule linked to the real-time compensation path generation strategy, ensuring that the direction correction vector is adjusted within the allowable deviation direction range according to the adjustment amplitude of the direction sensitivity factor during subsequent compensation.

[0161] Correlating and mapping the direction sensitivity factor in the dynamic adjustment coefficient with the allowable deviation direction range in the direction locking condition can establish a corresponding relationship between the two. Through this correlation mapping, a dynamic direction constraint rule linked to the real-time compensation path generation strategy is generated.

[0162] The dynamic direction constraint rule refers to the rule that constrains the adjustment amplitude and direction of the direction correction vector during the error compensation process. It ensures that the direction correction vector is adjusted within the allowable deviation direction range according to the adjustment amplitude of the direction sensitivity factor during subsequent compensation.

[0163] For example, if the direction sensitivity factor is large, it indicates that the deviation of the compensation direction is relatively serious. At this time, according to the dynamic direction constraint rule, the adjustment amplitude of the direction correction vector can be increased, but at the same time, it is necessary to ensure that the adjusted direction is within the allowable deviation direction range. By generating the dynamic direction constraint rule, the adjustment of the compensation direction can be more effectively controlled, improving the effect of error compensation.

[0164] Step S250: Based on the path correction parameter generation strategy, perform hierarchical compensation calculations on the fused error signal sequence to generate the ion source current correction value, deflection electrode voltage correction value, and beam focusing intensity correction value in the error compensation parameter set.

[0165] The path correction parameter generation strategy provides guidance for the compensation calculation of the fused error signal sequence. Based on this strategy, hierarchical compensation calculations are performed on the fused error signal sequence. Hierarchical compensation calculation refers to analyzing and processing the fused error signal sequence according to different levels, and respectively calculating the ion source current correction value, deflection electrode voltage correction value, and beam focusing intensity correction value.

[0166] The ion source current correction value refers to the value for adjusting the ion source current to correct the intensity fluctuation of the ion beam. The deflection electrode voltage correction value refers to the value for adjusting the voltage of the deflection electrode to correct the path deviation of the ion beam. The beam current focusing intensity correction value refers to the value for adjusting the beam current focusing intensity to ensure the focusing effect of the ion beam.

[0167] Through hierarchical compensation calculation, precise compensation can be carried out for different error factors, thereby improving the processing accuracy of the focused ion beam equipment.

[0168] Step S300: Input the error compensation parameter set into the control unit of the focused ion beam equipment to generate a real-time compensation control instruction set.

[0169] Input the error compensation parameter set into the control unit of the focused ion beam equipment. The control unit generates a real-time compensation control instruction set according to the error compensation parameter set. The real-time compensation control instruction set contains a series of instructions for adjusting the parameters of the focused ion beam equipment to achieve error compensation for the ion beam path and intensity.

[0170] For example, for parameters such as the ion source current correction value, deflection electrode voltage correction value, and beam current focusing intensity correction value in the error compensation parameter set, the control unit converts these parameters into corresponding control instructions, such as instructions for adjusting the pulse frequency and duty cycle of the ion source emission current, adjusting the voltage gradient distribution and polarity switching timing of the deflection electrode, and adjusting the magnetic field intensity and electric field uniformity parameters of the focusing lens.

[0171] As an implementation manner, in step S300, inputting the error compensation parameter set into the control unit of the focused ion beam equipment to generate a real-time compensation control instruction set may specifically include the following steps:

[0172] Step S310: Analyze the ion source current correction value in the error compensation parameter set to generate a first control instruction to adjust the pulse frequency and duty cycle of the ion source emission current.

[0173] The ion source current correction value in the error compensation parameter set is the value for adjusting the ion source current to correct the intensity fluctuation of the ion beam. Analyzing this correction value can obtain specific adjustment information, such as the current value to be increased or decreased, the adjustment time, etc.

[0174] According to the information obtained by the analysis, generate a first control instruction to adjust the pulse frequency and duty cycle of the ion source emission current. The pulse frequency refers to the frequency at which the pulses of the ion source emission current appear, and the duty cycle refers to the ratio of the pulse duration to the pulse period. By adjusting the pulse frequency and duty cycle, the intensity and time distribution of the ion source emission current can be precisely controlled, thereby achieving the correction of the ion beam intensity.

[0175] Step S320: Analyze the deflection electrode voltage correction value in the error compensation parameter set, and generate a second control instruction to adjust the voltage gradient distribution and polarity switching timing of the deflection electrode.

[0176] The deflection electrode voltage correction value in the error compensation parameter set is the value required to adjust the deflection electrode voltage to correct the path deviation of the ion beam. Analyzing this correction value can obtain specific adjustment information, such as the voltage value to be increased or decreased, the adjustment position, etc.

[0177] According to the information obtained from the analysis, generate a second control instruction to adjust the voltage gradient distribution and polarity switching timing of the deflection electrode. The voltage gradient distribution refers to the voltage change situation at different positions on the deflection electrode, and the polarity switching timing refers to the switching order of the polarity of the deflection electrode at different times. By adjusting the voltage gradient distribution and polarity switching timing, the deflection direction and angle of the ion beam can be precisely controlled, thereby realizing the correction of the ion beam path. Generating the second control instruction can be achieved according to the control interface and protocol of the deflection electrode.

[0178] Step S330: Analyze the beam current focusing intensity correction value in the error compensation parameter set, and generate a third control instruction to adjust the magnetic field intensity and electric field uniformity parameter of the focusing lens.

[0179] The beam current focusing intensity correction value in the error compensation parameter set is the value required to adjust the beam current focusing intensity to ensure the focusing effect of the ion beam. Analyzing this correction value can obtain specific adjustment information, such as the magnetic field intensity to be increased or decreased, the adjustment value of the electric field uniformity parameter, etc. According to the information obtained from the analysis, generate a third control instruction to adjust the magnetic field intensity and electric field uniformity parameter of the focusing lens. The magnetic field intensity refers to the intensity of the magnetic field generated by the focusing lens, and the electric field uniformity parameter refers to the uniformity degree of the electric field generated by the focusing lens. By adjusting the magnetic field intensity and electric field uniformity parameter, the focusing effect of the ion beam can be precisely controlled, thereby improving the processing accuracy of the ion beam. Generating the third control instruction can be achieved according to the control interface and protocol of the focusing lens.

[0180] Step S340: Align the first control instruction, the second control instruction, and the third control instruction according to the time synchronization signal to generate a parallel execution queue in the real-time compensation control instruction set.

[0181] The time synchronization signal is a signal used to synchronize the execution time of each control instruction. Aligning the first control instruction, the second control instruction, and the third control instruction according to the time synchronization signal can ensure that these control instructions are executed at the correct time points, avoiding time conflicts and interference.

[0182] Generating a parallel execution queue in the real-time compensation control instruction set means arranging the aligned first control instruction, second control instruction, and third control instruction in a predefined order to form a queue that can be executed in parallel. The parallel execution queue can improve the execution efficiency of control instructions and reduce the time for error compensation. The generation of timing alignment and the parallel execution queue can be implemented using hardware circuits or software algorithms.

[0183] Step S350: According to the execution delay times of the instructions in the parallel execution queue, dynamically adjust the trigger time points of the control instructions to match the ion beam scanning cycle.

[0184] According to the execution delay times of the instructions in the parallel execution queue, analyze the time required for each control instruction to be completed from triggering to execution. The ion beam scanning cycle refers to the time required for the ion beam to perform a complete scan in the target processing area.

[0185] Dynamically adjust the trigger time points of the control instructions to match the ion beam scanning cycle. To ensure that the control instructions can be executed at the appropriate time during the ion beam scan, it is necessary to adjust the trigger time points of the control instructions according to the execution delay times of the instructions and the remaining time window of the ion beam scanning cycle.

[0186] For example, if the execution delay time of a certain control instruction is long and the remaining time window of the ion beam scanning cycle is short, this instruction needs to be triggered in advance; if the execution delay time is short and the remaining time window is long, the triggering of this instruction can be appropriately delayed.

[0187] As an implementation, step S350, according to the execution delay times of the instructions in the parallel execution queue, dynamically adjust the trigger time points of the control instructions to match the ion beam scanning cycle, can specifically include the following steps:

[0188] Step S351: Obtain the start time point and end time point of the ion beam scanning cycle, and calculate the theoretical execution durations of the instructions in the parallel execution queue.

[0189] The ion beam scanning cycle refers to the time required for the ion beam to complete one scan. Obtaining the start time point and end time point of the ion beam scanning cycle can be achieved through the control interface or monitoring system of the ion beam scanning device. The theoretical execution durations of the instructions in the parallel execution queue refer to the time required for each control instruction to be completed from triggering to execution. Calculating the theoretical execution durations can be determined according to factors such as the type of instruction and the performance of the device.

[0190] Step S352: Determine the instruction compression ratio or expansion ratio according to the matching degree between the theoretical execution duration and the remaining time window of the ion beam scanning cycle.

[0191] According to the matching degree between the theoretical execution duration and the remaining time window of the ion beam scanning period, analyze whether the theoretical execution duration can be completed within the remaining time window. If the theoretical execution duration is greater than the remaining time window, it indicates that the instruction needs to be compressed; if the theoretical execution duration is less than the remaining time window, it indicates that the instruction can be extended. Determine the instruction compression ratio or extension ratio. The instruction compression ratio is used to reduce the execution time of the instruction, and the instruction extension ratio is used to extend the execution time of the instruction. By adjusting the execution time of the instruction, make it match the ion beam scanning period. For example, calculate the ratio of the theoretical execution duration to the remaining time window, and determine the instruction compression ratio or extension ratio according to the comparison result of this ratio with the preset threshold. If the ratio is greater than 1, it indicates that the instruction needs to be compressed, and determine the compression ratio according to the difference between the ratio and 1; if the ratio is less than 1, it indicates that the instruction can be extended, and determine the extension ratio according to the difference between 1 and the ratio.

[0192] As an implementation manner, in step S352, according to the matching degree between the theoretical execution duration and the remaining time window of the ion beam scanning period, determine the instruction compression ratio or extension ratio, which may specifically include the following steps:

[0193] Step S3521: Obtain the total time span of the instruction queue corresponding to the theoretical execution duration, and extract the start time point and end time point of the remaining time window of the ion beam scanning period.

[0194] The total time span of the instruction queue corresponding to the theoretical execution duration refers to the sum of the theoretical execution durations of all control instructions in the parallel execution queue. Obtaining this total time span can be achieved by adding the theoretical execution durations of each control instruction.

[0195] The remaining time window of the ion beam scanning period refers to the time interval from the current time to the end time of the ion beam scanning period. Extracting the start time point and end time point of the remaining time window can be achieved through the control interface or monitoring system of the ion beam scanning device.

[0196] Obtaining the total time span of the instruction queue and the start time point and end time point of the remaining time window can provide data for subsequent matching degree calculation.

[0197] Step S3522: Calculate the ratio between the total time span of the instruction queue and the time length of the remaining time window to generate an initial matching degree coefficient; if the initial matching degree coefficient is greater than the preset upper threshold, it is determined that instruction compression is required; if the initial matching degree coefficient is less than the preset lower threshold, it is determined that instruction extension is required.

[0198] The ratio between the total time span of the instruction queue and the time length of the remaining time window can reflect the matching degree between the time required for instruction execution and the remaining time window. Generating the initial matching degree coefficient can use this ratio as the initial matching degree coefficient.

[0199] The preset upper threshold and the preset lower threshold are preset thresholds used to determine whether instruction compression or expansion is required. If the initial matching degree coefficient is greater than the preset upper threshold, it indicates that the time required for instruction execution exceeds the remaining time window. At this time, instruction compression is required to ensure that the instruction can be executed within the remaining time window. If the initial matching degree coefficient is less than the preset lower threshold, it indicates that the time required for instruction execution is less than the remaining time window. At this time, instruction expansion is required to make full use of the remaining time window.

[0200] Step S3523: Determine the basic adjustment amount of the instruction compression ratio or the expansion ratio according to the difference between the initial matching degree coefficient and the preset upper threshold or the lower threshold; the basic adjustment amount has a linear relationship with the difference, and the greater the difference, the greater the adjustment amount increases according to the preset gradient.

[0201] The difference between the initial matching degree coefficient and the preset upper threshold or the lower threshold reflects the deviation degree between the time required for instruction execution and the remaining time window. According to this difference, determine the basic adjustment amount of the instruction compression ratio or the expansion ratio.

[0202] The basic adjustment amount has a linear relationship with the difference, that is, the greater the difference, the greater the basic adjustment amount. The preset gradient is a preset ratio used to control the growth rate of the basic adjustment amount with the difference. For example, if the preset gradient is 0.1, then when the difference increases by 1 each time, the basic adjustment amount increases by 0.1.

[0203] Step S3524: Analyze the priority tags of each instruction in the parallel execution queue, mark the non-critical instructions as compressible instruction segments, and mark the critical instructions as expandable instruction segments; reduce the time duty cycle of the compressible instruction segments in equal proportion according to the basic adjustment amount, or extend the execution interval of the expandable instruction segments in segments.

[0204] The priority tags of each instruction in the parallel execution queue reflect the importance of each instruction in the error compensation process. Analyze these priority tags, mark the non-critical instructions as compressible instruction segments, and mark the critical instructions as expandable instruction segments.

[0205] A compressible instruction segment refers to an instruction segment that can be compressed in time, such as some control instructions with low time requirements. An expandable instruction segment refers to an instruction segment that can be expanded in time, such as some critical control instructions that require more time to complete.

[0206] The duty cycle of the compressible instruction segment is reduced in geometric proportion according to the basic adjustment amount, or the execution interval of the extensible instruction segment is extended in segments. Geometric reduction means reducing the execution time of the compressible instruction segment by the same proportion; segment extension means dividing the execution time of the extensible instruction segment into multiple small segments and then adding an interval time between each small segment.

[0207] In this way, on the premise of ensuring the execution effect of critical instructions, the execution time of instructions can be reasonably adjusted to match the ion beam scanning cycle.

[0208] Step S3525: Based on the time distribution of the adjusted compressible instruction segment or extensible instruction segment, recalculate the matching degree between the total time span of the instruction queue and the remaining time window. If the matching degree coefficient falls within the preset tolerance interval, lock the current instruction compression ratio or extension ratio as the final effective value;

[0209] Recalculating the matching degree between the total time span of the instruction queue and the remaining time window can be achieved according to the time distribution of the adjusted compressible instruction segment or extensible instruction segment. If the matching degree coefficient falls within the preset tolerance interval, it means that the matching degree between the adjusted instruction execution time and the remaining time window has met the requirements. At this time, the current instruction compression ratio or extension ratio can be locked as the final effective value.

[0210] The preset tolerance interval is a pre-set interval used to judge whether the matching degree coefficient meets the requirements. If the matching degree coefficient is within the preset tolerance interval, it means that the deviation between the instruction execution time and the remaining time window is within an acceptable range; if the matching degree coefficient exceeds the preset tolerance interval, it means that the instruction compression ratio or extension ratio needs to be further adjusted.

[0211] Step S3526: Apply the final effective value to the instruction timing arrangement in the parallel execution queue to generate a control instruction stream with timestamp marks synchronized with the ion beam scanning trajectory coordinate order.

[0212] Applying the final effective value to the instruction timing arrangement in the parallel execution queue means readjusting the execution time and order of each instruction in the parallel execution queue according to the compression ratio or expansion ratio of the finally effective instructions. Generating a control instruction stream with timestamp marks synchronized with the coordinate order of the ion beam scanning trajectory means adding timestamp marks to the arranged instruction sequence and arranging them in the coordinate order of the ion beam scanning trajectory to form a control instruction stream. The timestamp marks can ensure that the control instructions are executed at the correct time points, and synchronizing with the coordinate order of the ion beam scanning trajectory can ensure that the control instructions are consistent with the scanning position of the ion beam, achieving precise error compensation. By generating such a control instruction stream, it can be ensured that during the scanning process of the focused ion beam device, each control instruction can be accurately executed according to the predetermined time and order, thereby effectively compensating the parameters such as the path, intensity, and focus of the ion beam in real time.

[0213] Step S353: If the remaining time window is less than the theoretical execution duration, compress or partially omit the non-critical instructions in the parallel execution queue.

[0214] When the remaining time window is less than the theoretical execution duration, it means that all the instructions in the parallel execution queue cannot be completed on time within the current ion beam scanning cycle. To ensure that the critical instructions can be executed within the specified time, the non-critical instructions need to be processed. Non-critical instructions are those instructions that have relatively little impact on the error compensation effect, such as some auxiliary adjustment instructions. Compressing the non-critical instructions, that is, shortening their execution time, can be achieved by increasing the execution speed of the instructions or reducing the execution steps of the instructions. Partial omission means directly skipping some operations in the non-critical instructions that have little impact on the error compensation. For example, if a non-critical instruction is some fine-tuning operations on the ion source and this operation contributes little to the error compensation effect of the current scanning area, then the execution time of this instruction can be compressed, or some minor adjustment steps can be omitted. In this way, within the limited time, the execution of critical instructions can be guaranteed first, thereby minimizing the impact on the error compensation effect due to insufficient time.

[0215] Step S354: If the remaining time window is greater than the theoretical execution duration, perform segmented extended execution on the critical instructions in the parallel execution queue.

[0216] If the remaining time window is greater than the theoretical execution time, it means that there is enough time to execute the instructions in the parallel execution queue within the current ion beam scanning cycle, and there is even time margin. At this time, the key instructions in the parallel execution queue can be extended and executed in segments. Key instructions are those that play a key role in the error compensation effect, such as instructions for adjusting the ion beam path deviation, instructions for controlling the ion beam intensity, etc. Segmented extended execution means dividing the execution process of key instructions into multiple small segments and adding a preset time interval between each small segment. The advantage of doing this is that the execution process of instructions can be controlled more finely and the accuracy of error compensation can be improved. For example, for key instructions for adjusting the ion beam path deviation, its execution process can be divided into several stages, each stage corresponding to a different adjustment range, and the time between each stage can be appropriately extended to better observe the adjustment effect of the ion beam path and make further fine-tuning according to actual conditions. By executing key instructions in segments, the remaining time window can be fully utilized to improve the quality of error compensation.

[0217] Step S355: Generate a control instruction stream marked with a timestamp according to the adjusted instruction execution sequence, and cache and distribute the control instruction stream in the coordinate order of the ion beam scanning trajectory.

[0218] After the instructions in the parallel execution queue are compressed, expanded or omitted, the adjusted instruction execution sequence is obtained. A control instruction stream with a timestamp mark is generated according to this sequence. The timestamp mark can accurately record the execution time point of each instruction to ensure that the instruction can be accurately executed in a predetermined time sequence. Then, the control instruction stream is cached and distributed according to the coordinate order of the ion beam scanning trajectory. The purpose of caching is to quickly obtain the required control instructions during the ion beam scanning process and improve the response speed of the system. Distribution is to accurately send the control instructions to the corresponding control units, such as the ion source control unit, the deflection electrode control unit and the focusing lens control unit, so that these units can adjust the parameters of the ion beam in real time according to the instructions. For example, when the ion beam scans to a preset coordinate position, the corresponding control instruction can be quickly retrieved from the cache and distributed to the corresponding control unit to achieve error compensation for the ion beam at this position.

[0219] Step S400: adjusting the ion source parameters and the deflection electrode voltage parameters of the focused ion beam device according to the real-time compensation control instruction set, and outputting the adjusted ion beam path parameter set.

[0220] The real-time compensation control instruction set is generated after a series of processing and analysis, and is used to precisely control the focused ion beam equipment to achieve error compensation. It contains the adjustment information for ion source parameters and deflection electrode voltage parameters, which are determined based on the error signals obtained in the previous steps and the trained error compensation parameters.

[0221] Ion source parameters mainly include the pulse frequency and duty cycle of the ion source emission current, etc. The pulse frequency of the ion source emission current determines the emission frequency of the ion beam, while the duty cycle determines the duration of each pulse. These two parameters directly affect the intensity and stability of the ion beam. When the real-time compensation control instruction set contains instructions for adjusting ion source parameters, the pulse frequency and duty cycle of the ion source emission current are adjusted according to the instructions. For example, if the instruction requires increasing the pulse frequency of the ion source emission current, the emission frequency of the ion beam is correspondingly increased, thereby increasing the number of ions emitted per unit time and further increasing the intensity of the ion beam; if the instruction requires adjusting the duty cycle, the duration of each pulse is changed, making the emission of the ion beam more stable and controllable.

[0222] Deflection electrode voltage parameters include the voltage gradient distribution and polarity switching timing of the deflection electrode, etc. The voltage gradient distribution of the deflection electrode determines the deflection direction and angle of the ion beam in the electric field, while the polarity switching timing affects the scanning trajectory and accuracy of the ion beam. According to the relevant instructions in the real-time compensation control instruction set, the voltage gradient distribution and polarity switching timing of the deflection electrode are adjusted. For example, when the instruction requires changing the voltage gradient distribution, the voltage values at different positions on the deflection electrode are adjusted, thereby changing the deflection direction and angle of the ion beam to make it reach the target processing area more accurately; when the instruction requires adjusting the polarity switching timing, the polarity of the deflection electrode is changed according to the new timing to achieve more precise ion beam scanning.

[0223] After adjusting the ion source parameters and deflection electrode voltage parameters, the path of the ion beam will change accordingly. Through precise control of these adjustments, the adjusted ion beam path parameter set is finally output. The adjusted ion beam path parameter set contains the path information of the ion beam under the new parameter settings, such as the trajectory, position accuracy, focusing degree, etc. of the ion beam. These parameter sets are obtained after error compensation adjustment and can more accurately reflect the actual path of the ion beam, providing a basis for subsequent precise ion beam processing of the target processing area.

[0224] Step S500: Perform ion beam processing on the target processing area according to the adjusted ion beam path parameter set, and monitor the morphological feature data of the processed area in real time to verify the error compensation effect.

[0225] The set of adjusted ion beam path parameters includes parameters such as the corrected value of the ion source current, the corrected value of the deflection electrode voltage, and the corrected value of the beam focusing intensity after error compensation. These parameters can enable the ion beam to process the target processing area more accurately along the predetermined path. The topographic feature data of the processed area is monitored in real time to verify the error compensation effect. The topographic feature data reflects information such as the surface morphology and dimensional accuracy of the target processing area after ion beam processing. By monitoring these data in real time, it can be determined whether the error compensation is effective and whether the expected processing accuracy is achieved.

[0226] As an implementation manner, in step S500, ion beam processing is performed on the target processing area according to the set of adjusted ion beam path parameters, and the topographic feature data of the processed area is monitored in real time to verify the error compensation effect, which may specifically include the following steps:

[0227] Step S510: During the ion beam processing, a sequence of topographic images of the target processing area is obtained through an in-situ electron microscope.

[0228] An in-situ electron microscope is a device that can observe the surface morphology of a sample in real time during the sample processing. During the ion beam processing, the in-situ electron microscope is used to continuously observe and photograph the target processing area to obtain a series of topographic images. These images are arranged in chronological order to form a sequence of topographic images. The in-situ electron microscope has the characteristics of high resolution and real-time imaging, and can clearly capture the minute changes in the target processing area during the ion beam processing. For example, when performing ion beam etching on a semiconductor chip, the in-situ electron microscope can record in real time the surface morphology changes in the etching area, including information such as the etching depth and edge roughness. By obtaining the sequence of topographic images, a detailed data basis can be provided for subsequent topographic feature analysis.

[0229] Step S520: Edge contour extraction and surface roughness analysis are performed on the sequence of topographic images to generate a dataset of topographic features.

[0230] Process the acquired sequence of topography images. First, perform edge contour extraction. Edge contour extraction refers to identifying the boundary contour of the target machining area from the topography image, which can be achieved through image processing algorithms such as edge detection algorithms. Possible edge detection algorithms include the Sobel operator, Canny operator, etc. These algorithms can detect edge information based on the change in pixel gray values in the image. Through edge contour extraction, the shape and size information of the target machining area can be obtained. Then, perform surface roughness analysis. Surface roughness refers to the microscopic geometric shape error of the surface of the target machining area, which has an important impact on machining quality. Surface roughness analysis can be achieved by calculating parameters such as the change rate of pixel gray values and height fluctuations in the image. For example, the power spectral density analysis method can be used to evaluate surface roughness. This method can transform the surface height data into the frequency domain to analyze the surface fluctuations of different frequency components. Organize and summarize the results obtained from edge contour extraction and surface roughness analysis to generate a topography feature dataset. This dataset contains important information such as the shape, size, and surface roughness of the target machining area, providing a quantitative basis for verifying the error compensation effect in the subsequent steps.

[0231] Step S530: Compare the topography feature dataset with the preset target machining accuracy standards to determine the topography error distribution map and the local error exceeding standard areas.

[0232] The preset target machining accuracy standards are a series of parameters preset according to the requirements of the machining task, such as the shape accuracy of the target machining area, dimensional tolerance, surface roughness requirements, etc. Compare the generated topography feature dataset with these standards and evaluate the error situation by calculating the difference between the two. For example, for the size of the target machining area, compare the measured size in the topography feature dataset with the preset size standard and calculate the size deviation; for surface roughness, compare the measured roughness parameters with the preset roughness standard to determine whether they meet the requirements. According to the comparison results, draw a topography error distribution map, which can visually show the error distribution of each position in the target machining area. At the same time, by setting an error threshold, determine the local error exceeding standard areas, that is, the areas where the error exceeds this threshold. For example, in the topography error distribution map, mark the areas where the error value is greater than the preset threshold as local error exceeding standard areas. Determining the topography error distribution map and the local error exceeding standard areas can help us quickly locate the areas with larger errors, providing a clear target for subsequent feedback compensation.

[0233] Step S540: Generate feedback compensation parameters according to the error accumulation trend in the topography error distribution map to correct the parameters of the dynamic weight adjustment layer in the adaptive error compensation model.

[0234] The error accumulation trend in the topography error distribution map reflects the change of errors in the target machining area during the whole machining process. By analyzing the error accumulation trend, it is possible to understand how errors accumulate with the change of machining time and position. For example, if the error shows a gradually increasing trend in a certain area, it indicates that there may be some continuous interference factors or insufficient error compensation in this area. According to the error accumulation trend, feedback compensation parameters are generated. The feedback compensation parameters are a set of values used to correct the parameters of the dynamic weight adjustment layer in the adaptive error compensation model. The dynamic weight adjustment layer is responsible for adjusting the correlation weights between error signals according to the frequency characteristics of the environmental interference signal. By correcting its parameters, the adaptive error compensation model can better adapt to different error situations. For example, if the error accumulation trend shows that the environmental interference signal has a greater impact on the ion beam path deviation signal, then by adjusting the parameters of the dynamic weight adjustment layer, the weight of the ion beam path deviation signal can be increased, thereby improving the effect of error compensation. The process of generating feedback compensation parameters can be based on machine learning algorithms or empirical formulas, and specific parameter values are determined according to the characteristics of the error accumulation trend and the characteristics of the adaptive error compensation model.

[0235] Step S550: If it is detected that the area of the local error exceeding standard region exceeds the preset tolerance threshold, trigger the compensation model retraining process to update the ion source current correction value and the deflection electrode voltage correction value in the error compensation parameter set.

[0236] If it is detected that the area of the local error exceeding standard region exceeds the preset tolerance threshold, it indicates that the current error compensation effect is not ideal and the adaptive error compensation model needs to be adjusted. The preset tolerance threshold is a pre-set area threshold used to judge whether the area of the local error exceeding standard region is too large.

[0237] Trigger the compensation model retraining process to update the ion source current correction value and the deflection electrode voltage correction value in the error compensation parameter set. The compensation model retraining process is a process of retraining the adaptive error compensation model. By collecting more error data and machining result data, the parameters of the model are adjusted so that the model can better adapt to the actual machining situation. Updating the ion source current correction value and the deflection electrode voltage correction value in the error compensation parameter set can more accurately adjust the path and intensity of the ion beam and improve the machining accuracy.

[0238] As an implementation manner, in step S550, triggering the compensation model retraining process to update the ion source current correction value and the deflection electrode voltage correction value in the error compensation parameter set may specifically include the following steps:

[0239] Step S551: Pause the current ion beam machining process and collect the topography feature data and the real-time error signal set of the most recent N machining cycles, where N≥1.

[0240] When it is detected that the area of the local error exceeding the standard region exceeds the preset tolerance threshold, it indicates that the current error compensation effect is not ideal, and the adaptive error compensation model needs to be retrained. First, pause the current ion beam processing process to avoid further processing errors. Then collect the topography feature data and the set of real-time error signals for the most recent N processing cycles. The topography feature data contains information such as the shape, size, and surface roughness of the target processing area, and the set of real-time error signals includes ion beam path deviation signals, ion beam intensity fluctuation signals, and environmental interference signals, etc. The purpose of collecting these data is to obtain more comprehensive error information for more accurate retraining of the adaptive error compensation model. For example, in the ion beam processing of semiconductor chips, collecting data for the most recent several processing cycles can help us analyze the variation law of errors and possible influencing factors. The value of N can be adjusted according to the actual situation. Generally speaking, the larger N is, the richer the collected data is, but at the same time, it will also increase the complexity and time cost of data processing.

[0241] Step S552: According to the coordinates of the error exceeding the standard region in the topography feature data, trace back the abnormal signal segments in the corresponding set of real-time error signals in reverse.

[0242] After collecting the topography feature data and the set of real-time error signals, according to the coordinates of the error exceeding the standard region in the topography feature data, trace back the abnormal signal segments in the corresponding set of real-time error signals in reverse. The coordinates of the error exceeding the standard region indicate the specific positions in the target processing area where the error exceeds the preset threshold. Through these coordinates, the corresponding real-time error signals during processing at these positions can be located. The abnormal signal segments refer to the signal parts that are significantly different from the normal signals when processing in the error exceeding the standard region. These abnormal signals may be the reasons for the error exceeding the standard. For example, at a preset coordinate position, the topography feature data shows that the surface roughness exceeds the standard. By tracing back the set of real-time error signals in reverse, it can be found that the ion beam path deviation signal corresponding to this position has a large fluctuation, and this fluctuation signal is the abnormal signal segment. By tracing back the abnormal signal segments, the reasons for the error generation can be analyzed more pertinently, providing a basis for subsequent model adjustment.

[0243] Step S553: Conduct spectral analysis and time-domain feature extraction on the abnormal signal segments to determine the signal mutation points and persistent interference patterns.

[0244] Perform spectral analysis and time-domain feature extraction on the abnormal signal segments obtained by reverse tracing. Spectral analysis is to transform the abnormal signal from the time domain to the frequency domain and analyze its frequency components and distribution. Through spectral analysis, the energy distribution of different frequency components in the abnormal signal can be understood, and whether there is an interference signal of a predefined frequency can be judged. For example, if the spectral analysis result shows a large energy peak in a certain high-frequency band, it indicates that there may be a high-frequency interference signal. Time-domain feature extraction is to analyze the characteristics of the abnormal signal in the time domain, such as the amplitude change of the signal, the slope of the rising edge and falling edge, and the duration. Through time-domain feature extraction, signal mutation points and persistent interference patterns can be determined. Signal mutation points refer to the points where the amplitude of the abnormal signal suddenly changes, and these mutation points may be caused by sudden interference events. Persistent interference patterns refer to the regular and persistent interference characteristics in the abnormal signal, such as periodic fluctuations and slow drifts. Determining signal mutation points and persistent interference patterns can help us understand the mechanism of error generation more deeply and provide key information for adjusting the adaptive error compensation model.

[0245] Step S554: Based on the signal mutation points and persistent interference patterns, adjust the frequency response range of the dynamic weight adjustment layer and the parameter generation granularity of the compensation path generation layer in the adaptive error compensation model.

[0246] Based on the signal mutation points and persistent interference patterns, adjust the frequency response range of the dynamic weight adjustment layer in the adaptive error compensation model. The frequency response range of the dynamic weight adjustment layer determines the sensitivity of the model to error signals of different frequencies. If the signal mutation points correspond to high-frequency interference, then the processing bandwidth of the dynamic weight adjustment layer for high-frequency signals can be expanded, enabling the model to respond to high-frequency interference more promptly.

[0247] At the same time, adjust the parameter generation granularity of the compensation path generation layer. The parameter generation granularity refers to the fineness of the compensation parameters generated by the compensation path generation layer. For persistent interference patterns, if the change of the interference is relatively slow, the parameter generation granularity can be appropriately increased to reduce unnecessary adjustments; if the interference changes rapidly, the parameter generation granularity needs to be reduced to improve the compensation accuracy.

[0248] As an implementation, step S554, based on the signal mutation points and persistent interference patterns, adjust the frequency response range of the dynamic weight adjustment layer and the parameter generation granularity of the compensation path generation layer in the adaptive error compensation model, which can specifically include the following steps:

[0249] Step S5541: Expand the high-frequency signal processing bandwidth of the dynamic weight adjustment layer according to the time distribution density of the signal mutation points.

[0250] The time distribution density of signal mutation points reflects the frequency of signal mutations. If the time distribution density of signal mutation points is large, it indicates that signal mutations occur frequently in a short period of time, and these mutant signals usually contain higher frequency components. To better process these high-frequency signals, it is necessary to expand the high-frequency signal processing bandwidth of the dynamic weight adjustment layer. The high-frequency signal processing bandwidth of the dynamic weight adjustment layer determines the frequency range of high-frequency signals that it can process. Expanding the bandwidth can enable this layer to capture more high-frequency signal information, thereby more accurately adjusting the correlation weights between error signals. For example, when it is detected that the time distribution density of signal mutation points increases, the filter parameters of the dynamic weight adjustment layer can be adjusted to improve its response ability to high-frequency signals and expand the high-frequency signal processing bandwidth. This can enable the adaptive error compensation model to better cope with sudden high-frequency interference and improve the timeliness and accuracy of error compensation.

[0251] Step S5542: According to the energy distribution characteristics of the continuous interference pattern, reduce the low-frequency signal attenuation coefficient of the dynamic weight adjustment layer.

[0252] The energy distribution characteristics of the continuous interference pattern reflect the energy distribution of continuous interference signals in different frequency bands. If the continuous interference pattern has a large amount of energy in the low-frequency band, it indicates that the low-frequency interference signal has a greater impact on the error. To better process these low-frequency interference signals, it is necessary to reduce the low-frequency signal attenuation coefficient of the dynamic weight adjustment layer. The low-frequency signal attenuation coefficient determines the attenuation degree of the dynamic weight adjustment layer for low-frequency signals. Reducing this coefficient can make the low-frequency signals attenuate less when passing through this layer, thereby retaining more low-frequency signal information. For example, when analyzing the energy distribution characteristics of the continuous interference pattern and finding that the energy in the low-frequency band is high, the filter parameters of the dynamic weight adjustment layer can be adjusted to reduce the attenuation coefficient of the low-frequency signals. This can enable the adaptive error compensation model to more effectively cope with low-frequency interference and improve the stability of error compensation.

[0253] Step S5543: Add short-term compensation path generation channels and long-term compensation path generation channels in the compensation path generation layer to process signal mutation points and continuous interference patterns respectively.

[0254] To more specifically process signal mutation points and persistent interference patterns, a short-term compensation path generation channel and a long-term compensation path generation channel are added in the compensation path generation layer. The short-term compensation path generation channel is specifically used to process signal mutation points. Since signal mutation points are usually sudden and short-term interferences, the short-term compensation path generation channel needs to have the ability to respond quickly. It can generate corresponding compensation paths according to the characteristics of signal mutation points to quickly eliminate the impact of mutation signals on ion beam processing. For example, when a signal mutation point is detected, the short-term compensation path generation channel can immediately adjust the path and intensity of the ion beam to avoid further expansion of processing errors. The long-term compensation path generation channel is used to process persistent interference patterns. Persistent interference patterns usually have regularity and long-term nature. The long-term compensation path generation channel can generate stable compensation paths through the analysis and modeling of persistent interference patterns to continuously correct errors. For example, for periodic interference signals, the long-term compensation path generation channel can generate corresponding periodic compensation paths according to their period and amplitude characteristics to ensure the stability of the accuracy of ion beam processing over a long period of time.

[0255] Step S5544: Set a fast response time window for the short-term compensation path generation channel and set a smooth transition constraint condition for the long-term compensation path generation channel.

[0256] Set a fast response time window for the short-term compensation path generation channel. The fast response time window refers to the maximum allowable time for the short-term compensation path generation channel to generate an effective compensation path from detecting a signal mutation point. Setting the fast response time window can ensure that the short-term compensation path generation channel can respond to signal mutation points within a short time and eliminate the impact of mutation signals in a timely manner. For example, the fast response time window can be set to the millisecond level. In this way, after detecting a signal mutation point, the short-term compensation path generation channel can generate a compensation path and send it to the control unit within an extremely short time to achieve rapid adjustment of the ion beam. Set a smooth transition constraint condition for the long-term compensation path generation channel. The smooth transition constraint condition means that during the generation of the long-term compensation path, it is required that the change of the compensation path cannot be too drastic and should achieve a smooth transition. This is because persistent interference patterns usually change slowly, and too drastic changes in the compensation path may lead to the generation of new errors. For example, the smooth transition constraint condition can be achieved by restricting the change rate of the compensation path, setting a filtering link, etc., to ensure that the compensation path generated by the long-term compensation path generation channel can stably and smoothly adjust the parameters of the ion beam.

[0257] Step S5545: Synchronize the parameters of the adjusted dynamic weight adjustment layer and the compensation path generation layer to all online processing devices to achieve consistency in the error compensation strategy among multiple devices.

[0258] After adjusting the parameters of the dynamic weight adjustment layer and the compensation path generation layer in the adaptive error compensation model, synchronize these adjusted parameters to all online processing devices. The purpose of doing this is to achieve the consistency of the error compensation strategy among multiple devices, ensure that all devices can adopt the same compensation strategy when facing the same error situation, and improve the overall processing quality and efficiency. Synchronizing parameters can be achieved through network communication and other means, sending the adjusted parameters to the adaptive error compensation model of each online processing device to update its corresponding parameters. For example, in a large-scale semiconductor manufacturing factory, there are multiple focused ion beam devices processing simultaneously. By synchronizing the adjusted parameters to all devices, these devices can maintain a consistent strategy when dealing with errors, avoiding problems of inconsistent processing quality caused by differences between devices. At the same time, achieving the consistency of the error compensation strategy among multiple devices can also facilitate the monitoring and management of the entire production process, improving the stability and reliability of production.

[0259] Step S555: Retrain the real-time error signal set using the adjusted adaptive error compensation model to generate an updated error compensation parameter set and resume ion beam processing.

[0260] After completing the parameter adjustment of the dynamic weight adjustment layer and the compensation path generation layer in the adaptive error compensation model and synchronizing the adjusted parameters to all online processing devices, use the adjusted adaptive error compensation model to retrain the real-time error signal set. The retraining process takes the real-time error signal set as input, calculates and learns through the adjusted model, continuously optimizes the model's parameters, so that it can more accurately generate appropriate compensation parameters according to the input error signals. After retraining, an updated error compensation parameter set is generated, and this set contains parameters such as updated ion source current correction values, deflection electrode voltage correction values, and beam current focusing intensity correction values. Finally, resume the ion beam processing process, input the updated error compensation parameter set into the control unit of the focused ion beam device, and adjust the parameters of the ion beam in real time according to these parameters to achieve more accurate error compensation. After resuming processing, continue to monitor the morphological feature data of the processed area in real time to verify the updated error compensation effect. If there are still error problems, the above steps can be repeated for further adjustment and optimization.

[0261] Please refer to Figure 2 , Figure 2Schematic structural diagram of an error compensation system provided by an embodiment of the present invention. The error compensation system is, for example, a computer system embedded in or connected to a focused ion beam device, and at least includes a processor 101, a communication interface 102, and a memory 103. Among them, the processor 101, the communication interface 102, and the memory 103 can be connected through a bus or other means. Among them, the processor 101 (or central processing unit (CPU)) is the computing core and control core of the error compensation system, which can parse various instructions in the error compensation system and process various data of the error compensation system. The communication interface 102 may optionally include a standard wired interface, a wireless interface (such as WI-FI, a mobile communication interface, etc.), and under the control of the processor 101, it can be used to send and receive data; the communication interface 102 can also be used for the transmission and interaction of internal data of the error compensation system. The memory 103 (Memory) is a memory device in the error compensation system, used to store programs and data. It can be understood that the memory 103 here can include both the built-in memory of the error compensation system and, of course, the extended memory supported by the error compensation system. The memory 103 provides a storage space, and this storage space stores the operating system of the error compensation system, which may include but is not limited to the Android system, the iOS system, the WindowsPhone system, etc., and the present invention does not make any limitations in this regard. In one embodiment, the processor 101 executes the adaptive error compensation method for the focused ion beam device provided above in the embodiments of the present invention by running the computer program in the memory 103.

Claims

1. An adaptive error compensation method for a focused ion beam device, characterized in that The method includes: Obtaining a set of real-time error signals generated during the processing of a focused ion beam device, where the set of real-time error signals includes an ion beam path deviation signal, an ion beam intensity fluctuation signal, and an environmental interference signal; Training an adaptive error compensation model based on the set of real-time error signals to generate a set of error compensation parameters; the adaptive error compensation model includes a dynamic weight adjustment layer and a compensation path generation layer, where the dynamic weight adjustment layer is used to adjust the correlation weights between error signals according to the frequency characteristics of the environmental interference signal, and the compensation path generation layer is used to generate ion beam path correction parameters based on the adjusted correlation weights; Inputting the set of error compensation parameters into the control unit of the focused ion beam device to generate a set of real-time compensation control instructions; Adjusting the ion source parameters and deflection electrode voltage parameters of the focused ion beam device according to the set of real-time compensation control instructions, and outputting a set of adjusted ion beam path parameters; Performing ion beam processing on a target processing area according to the set of adjusted ion beam path parameters, and real-time monitoring the morphological feature data of the processed area to verify the error compensation effect.

2. The method according to claim 1, characterized in that, The training of the adaptive error compensation model based on the set of real-time error signals to generate a set of error compensation parameters includes: Obtaining the error signal distribution pattern in the set of historical error signals, and determining a dynamic compensation interval according to the error signal distribution pattern; the dynamic compensation interval includes the maximum fluctuation threshold of the environmental interference signal and the minimum correction step of the ion beam path deviation signal; Extracting the frequency characteristics of the current environmental interference signal from the set of real-time error signals, and generating a weight adjustment coefficient according to the matching degree between the frequency characteristics and the dynamic compensation interval; Inputting the weight adjustment coefficient into the dynamic weight adjustment layer to perform weighted fusion on the ion beam path deviation signal and the ion beam intensity fluctuation signal, and generating a sequence of fused error signals; Determining a path correction parameter generation strategy in the compensation path generation layer according to the mapping relationship between the sequence of fused error signals and a preset set of error compensation rules; Performing hierarchical compensation calculation on the sequence of fused error signals based on the path correction parameter generation strategy to generate the ion source current correction value, the deflection electrode voltage correction value, and the beam current focusing intensity correction value in the set of error compensation parameters.

3. The method according to claim 2, wherein The determining of the path correction parameter generation strategy in the compensation path generation layer according to the mapping relationship between the sequence of fused error signals and a preset set of error compensation rules includes: Obtaining the timestamp marking data in the sequence of fused error signals, and dividing a signal processing window according to the timestamp marking data; Within each signal processing window, determining an initial compensation direction according to the cumulative offset of the ion beam path deviation signal, and determining a compensation intensity gradient according to the amplitude change rate of the ion beam intensity fluctuation signal; Inputting the initial compensation direction and the compensation intensity gradient into a direction-intensity coupling module in the compensation path generation layer to generate a direction correction vector and an intensity correction vector; Generate a dynamic compensation factor based on the product of the direction correction vector and the intensity correction vector, and superimpose and calculate the dynamic compensation factor with a preset reference compensation parameter to generate a window-level compensation parameter; Smooth the window-level compensation parameters of multiple consecutive signal processing windows to generate the time continuity constraint condition and the space distribution constraint condition in the path correction parameter generation strategy.

4. The method according to claim 3, characterized in that, The smoothing the window-level compensation parameters of multiple consecutive signal processing windows to generate the time continuity constraint condition and the space distribution constraint condition in the path correction parameter generation strategy includes: Obtain the difference degree between the window-level compensation parameters of adjacent signal processing windows, and determine the parameter mutation threshold according to the difference degree; If the difference degree exceeds the parameter mutation threshold, perform a delay process on the current window-level compensation parameter, and generate a temporary compensation parameter based on the moving average value of the previous window-level compensation parameter; Perform weighted fusion on the temporary compensation parameter and the current window-level compensation parameter to generate a sequence of transition compensation parameters; Determine the maximum allowable deviation rate and the minimum transition step size in the time continuity constraint condition according to the parameter change trend in the sequence of transition compensation parameters; Extract the regional compensation priority and the compensation direction consistency index in the space distribution constraint condition based on the distribution characteristics of the sequence of transition compensation parameters in the spatial dimension.

5. The method according to claim 4, characterized in that, The extracting the regional compensation priority and the compensation direction consistency index in the space distribution constraint condition based on the distribution characteristics of the sequence of transition compensation parameters in the spatial dimension includes: Map the sequence of transition compensation parameters to the spatial coordinate grid of the target processing area to generate a grid compensation parameter distribution map; Determine the high error-sensitive area and the low error-sensitive area according to the parameter gradient difference of each grid node in the grid compensation parameter distribution map; Assign a first priority weight to the high error-sensitive area and a second priority weight to the low error-sensitive area; Calculate the consistency offset of the compensation direction according to the boundary overlap degree of the high error-sensitive area and the low error-sensitive area; Generate the dynamic adjustment coefficient and the direction locking condition in the compensation direction consistency index based on the comparison result of the consistency offset and the preset direction tolerance threshold.

6. The method according to claim 2, wherein The inputting the error compensation parameter set into the control unit of the focused ion beam device to generate a set of real-time compensation control instructions includes: Analyze the ion source current correction value in the error compensation parameter set to generate a first control instruction to adjust the pulse frequency and duty cycle of the ion source emission current; Analyze the deflection electrode voltage correction value in the error compensation parameter set to generate a second control instruction to adjust the voltage gradient distribution and polarity switching timing of the deflection electrode; Analyze the beam focusing intensity correction value in the error compensation parameter set to generate a third control instruction to adjust the magnetic field intensity and electric field uniformity parameters of the focusing lens; Align the first control instruction, the second control instruction, and the third control instruction according to the time synchronization signal to generate a parallel execution queue in the set of real-time compensation control instructions; Dynamically adjust the triggering time point of the control instruction according to the execution delay time of each instruction in the parallel execution queue to match the ion beam scanning period.

7. The method according to claim 6, characterized in that, The step of dynamically adjusting the triggering time point of the control instruction according to the execution delay time of each instruction in the parallel execution queue to match the ion beam scanning period includes: Obtain the start time point and end time point of the ion beam scanning period, and calculate the theoretical execution duration of each instruction in the parallel execution queue; Determine the instruction compression ratio or expansion ratio according to the matching degree between the theoretical execution duration and the remaining time window of the ion beam scanning period; If the remaining time window is less than the theoretical execution duration, compress or partially omit non-critical instructions in the parallel execution queue; If the remaining time window is greater than the theoretical execution duration, perform segmented expansion execution on critical instructions in the parallel execution queue; Generate a control instruction stream marked with time stamps according to the adjusted instruction execution sequence, and cache and distribute the control instruction stream in the coordinate order of the ion beam scanning trajectory.

8. The method according to claim 1, characterized in that, The step of performing ion beam machining on the target machining area according to the adjusted ion beam path parameter set and real-time monitoring the morphological feature data of the machined area to verify the error compensation effect includes: During the ion beam machining process, obtain a sequence of morphological images of the target machining area through an in-situ electron microscope; Perform edge contour extraction and surface roughness analysis on the sequence of morphological images to generate a morphological feature data set; Compare the morphological feature data set with a preset target machining accuracy standard to determine the morphological error distribution map and local error exceeding standard areas; Generate feedback compensation parameters according to the error accumulation trend in the morphological error distribution map to correct the parameters of the dynamic weight adjustment layer in the adaptive error compensation model; If it is detected that the area of the local error exceeding standard area exceeds the preset tolerance threshold, trigger a compensation model retraining process to update the ion source current correction value and deflection electrode voltage correction value in the error compensation parameter set.

9. The method according to claim 8, wherein The step of triggering a compensation model retraining process to update the ion source current correction value and deflection electrode voltage correction value in the error compensation parameter set includes: Pause the current ion beam machining process, and collect the morphological feature data and real-time error signal set of the most recent N machining cycles, where N≥1; According to the coordinates of the error exceeding standard area in the morphological feature data, trace back the abnormal signal segments in the corresponding real-time error signal set; Perform frequency spectrum analysis and time domain feature extraction on the abnormal signal segments to determine signal mutation points and persistent interference patterns; Based on the signal mutation points and persistent interference patterns, adjust the frequency response range of the dynamic weight adjustment layer and the parameter generation granularity of the compensation path generation layer in the adaptive error compensation model; Use the adjusted adaptive error compensation model to retrain the real-time error signal set, generate an updated error compensation parameter set, and resume ion beam machining. Among them, adjusting the frequency response range of the dynamic weight adjustment layer and the parameter generation granularity of the compensation path generation layer in the adaptive error compensation model based on the signal mutation point and the persistent interference pattern includes: Expanding the high-frequency signal processing bandwidth of the dynamic weight adjustment layer according to the time distribution density of the signal mutation point; Reducing the low-frequency signal attenuation coefficient of the dynamic weight adjustment layer according to the energy distribution characteristics of the persistent interference pattern; Adding a short-term compensation path generation channel and a long-term compensation path generation channel in the compensation path generation layer to process the signal mutation point and the persistent interference pattern respectively; Setting a fast response time window for the short-term compensation path generation channel and setting a smooth transition constraint condition for the long-term compensation path generation channel; Synchronizing the parameters of the adjusted dynamic weight adjustment layer and the compensation path generation layer to all online processing devices to achieve the consistency of the error compensation strategy among multiple devices.

10. An error compensation system, characterized in that, Including: A memory in which a computer program is stored; A processor for loading the computer program to implement the adaptive error compensation method for a focused ion beam device according to any one of claims 1-9.

Images