Method and device for diagnosing abnormal discharge of multispectral fusion plasma etching table

Abstract

The invention discloses a multispectral fusion plasma etching table abnormal discharge diagnosis method and device, and relates to the technical field of plasma etching, and the method comprises the steps: carrying out the spectrum detection of a key position through employing a miniature spectrum probe along with a plasma etching process, obtaining multispectral data, entropy field conversion and entropy change analysis are carried out through an entropy state detection assembly, and a first detection result is determined; and when the entropy change is close to a threshold value, triggering the plasma detection assembly to carry out further verification to obtain a second detection result, judging abnormal discharge according to the second detection result, carrying out diagnosis, and carrying out feedback control on the plasma etching process. The technical problems that abnormal discharge detection lags behind and is insufficient in accuracy in an existing plasma etching process, and the stability and the production efficiency of the etching process are affected are solved, early diagnosis and high-precision recognition of abnormal discharge are achieved through multispectral fusion and deep analysis, and the accuracy of abnormal discharge detection is improved. And the stability of the etching process and the production efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of plasma etching technology, specifically to a method and apparatus for diagnosing abnormal discharges in a plasma etching stage using multispectral fusion. Background Technology

[0002] Plasma etching technology is widely used in semiconductor manufacturing, thin film processing, and other fields, achieving a precise etching process through the reaction of high-energy plasma with the material surface. However, abnormal discharge phenomena may occur during plasma etching, which can not only affect etching accuracy and quality but also lead to equipment damage and production delays. Therefore, early diagnosis and control of abnormal discharges has become a significant challenge in optimizing etching processes. Traditional anomaly detection methods typically rely on a single sensor or simple threshold judgment, making it difficult to identify early signs of abnormal discharges in real time and accurately, resulting in slow response and low accuracy. Summary of the Invention

[0003] This application provides a multispectral fusion plasma etching stage abnormal discharge diagnosis method and device to solve the technical problems of lagging and inaccurate abnormal discharge detection in existing plasma etching processes, which affect the stability and production efficiency of the etching process.

[0004] The first aspect of this application provides a multispectral fusion-based method for diagnosing abnormal discharges on a plasma etching stage. The method includes: performing spectral detection at key locations using a micro-spectral probe as the plasma etching process progresses to determine multispectral data; uploading the multispectral data; performing spectral information entropy field conversion and entropy change analysis on the spectral pixels using an entropy state detection component to determine a first detection result; if the first detection result approaches an entropy change threshold, triggering the plasma detection component to perform a depth-based time-frequency characteristic-based verification to determine a second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder; and determining an abnormal discharge diagnosis result based on the second detection result to provide feedback control over the plasma etching process.

[0005] A second aspect of this application provides a multispectral fusion-based plasma etching stage abnormal discharge diagnostic device. The device includes: a spectral detection module for performing spectral detection at key locations using a micro-spectral probe as the plasma etching process progresses, determining multispectral data; an early diagnosis module for uploading the multispectral data and performing spectral information entropy field conversion and entropy change analysis on the spectral pixels based on an entropy state detection component, determining a first detection result; a depth verification module for triggering the plasma detection component to perform depth-based time-frequency characteristic verification if the first detection result approaches an entropy change threshold, determining a second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder; and a feedback control module for determining the abnormal discharge diagnostic result based on the second detection result and performing feedback control on the plasma etching process.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The multispectral fusion plasma etching stage abnormal discharge diagnosis method and device provided in this application relates to the field of plasma etching technology. By acquiring real-time multispectral data and detecting entropy state, combined with depth time-frequency feature analysis, it achieves early diagnosis of abnormal discharge in the plasma etching process. Entropy change analysis is used to determine whether the critical point is approaching, and the plasma detection component is triggered to perform depth verification. Finally, the etching process is controlled based on the detection results. This solves the technical problems of lagging and inaccurate abnormal discharge detection in existing plasma etching processes, which affect the stability and production efficiency of the etching process. It achieves the technical effect of early diagnosis and high-precision identification of abnormal discharge through multispectral fusion and depth analysis, thereby improving the stability and production efficiency of the etching process. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic flowchart of the multispectral fusion plasma etching stage abnormal discharge diagnosis method provided in the embodiments of this application; Figure 2 A schematic diagram of the structure of the multispectral fusion plasma etching stage abnormal discharge diagnostic device provided in the embodiments of this application.

[0009] Figure labeling: 11 Spectral detection module, 12 Early diagnosis module, 13 Depth verification module, 14 Feedback control module. Detailed Implementation

[0010] This application provides a multispectral fusion plasma etching stage abnormal discharge diagnosis method and device to solve the technical problems of lagging and inaccurate abnormal discharge detection in existing plasma etching processes, which affect the stability and production efficiency of the etching process.

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0012] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0013] Example 1, as Figure 1 As shown, this application provides a multispectral fusion-based method for diagnosing abnormal discharges in a plasma etching stage. The method includes: P10: As the plasma etching process progresses, a micro-spectral probe is used to perform spectral detection at key locations to determine multispectral data.

[0014] Specifically, in order to achieve real-time monitoring and precise control of the process status, this application employs a miniature spectral probe to perform spectral detection at key locations as the plasma etching process progresses, thereby determining multispectral data. The miniature spectral probe is a high-precision optical detection device; it is small in size, highly sensitive, and adaptable to the complex environment within the plasma etching chamber. It can capture minute changes occurring during the process in real time, thus providing first-hand data for timely detection of abnormal discharges or plasma instability.

[0015] In practical applications, miniature spectral probes are positioned at critical locations on the plasma etching stage, specifically in the areas most closely related to the etching process. For example, the probe can be mounted at a specific angle near the reaction chamber to ensure the capture of the most representative spectral data from the etched area. The probe's mounting angle, distance, and alignment method need to be adjusted according to the actual requirements of the etching process to ensure that the probe can acquire spectral signals to the maximum extent possible and avoid interference from environmental noise or reflected light.

[0016] During the etching process, the physicochemical state of the plasma is revealed through its emission spectrum. Miniature spectroscopic probes collect spectral signals from the plasma at different wavelengths, acquiring full-band spectral data from ultraviolet to infrared. This data is known as multispectral data, which refers to spectral information acquired across multiple specific wavelength ranges. Compared to single-band spectral data, it provides richer and more comprehensive information about the plasma state. For example, in the ultraviolet band, the spectral signal may reflect the excitation process of high-energy electrons in the plasma; while in the infrared band, the spectral signal may be related to molecular vibrations and rotations within the plasma. Analysis of this multispectral data allows for in-depth understanding of key parameters of the plasma, such as electron density, temperature, and chemical composition.

[0017] Furthermore, the detection process of the miniature spectral probe is performed in real time and can be synchronized with the plasma etching process, thereby enabling dynamic monitoring of the process progress. Once an abnormal signal is detected in the multispectral data, such as a sudden change in spectral intensity or a change in spectral shape, subsequent detection procedures can be quickly triggered for further analysis and processing.

[0018] P20: Upload the multispectral data, and perform spectral information entropy field conversion and entropy change analysis on the spectral pixels according to the entropy state detection component to determine the first detection result.

[0019] Optionally, after acquiring multispectral data at key locations in the plasma etching process, the acquired multispectral data is uploaded to the entropy state detection component. The entropy state detection component is a key device in this application used for preliminary judgment of the plasma etching process state. Its core function is to perform spectral information entropy field conversion and entropy change analysis on spectral pixels. Entropy is an important indicator in information theory, representing the degree of disorder in a system. In spectral data processing, changes in entropy can reflect the dynamic changes in plasma. When the state of plasma fluctuates drastically, the entropy value also changes significantly; therefore, entropy change analysis can determine whether there are any anomalies in the plasma.

[0020] Specifically, spectral information entropy field transformation is the process of converting multispectral data from a traditional spectral intensity distribution to an information entropy distribution. By calculating the information entropy of each spectral pixel, the spectral information entropy field can be obtained. Information entropy is an important indicator of data complexity and orderliness; a higher value indicates greater complexity and disorder in the spectral data, while a lower value indicates greater orderliness. Entropy change analysis is the process of monitoring and analyzing the changing trends of the spectral information entropy field. In plasma etching processes, the spectral information entropy field is relatively stable under normal process conditions, but it changes significantly when process anomalies occur (such as abnormal discharge). By monitoring these changes, these anomalies can be detected promptly.

[0021] The key to entropy change analysis lies in setting an entropy change threshold, which is the critical point where the entropy state transitions from order to disorder. When the change in spectral information entropy approaches or exceeds this threshold, it indicates a potential anomaly in the plasma etching process. For example, a sudden increase in entropy change may mean that the stability of the plasma has been disturbed, resulting in partial discharge or other abnormal phenomena. The entropy state detection component is highly efficient and fast, capable of making preliminary judgments on large amounts of spectral data in a short time. The main advantages of this detection method are its efficiency and real-time performance, allowing it to be performed synchronously with the plasma etching process and promptly detect potential anomalies. However, the detection results of the entropy state detection component are based solely on changes in information entropy and do not consider specific spectral characteristics. Therefore, it is mainly used for preliminary judgment of the process status. When the detected entropy change approaches the threshold, it triggers subsequent depth detection components to perform more detailed detection processing on that node.

[0022] Based on the analysis results of the entropy state detection component, a first detection result is determined. This first detection result is a preliminary judgment indicating whether the current plasma etching process is normal. If the first detection result indicates that the entropy change has not approached the threshold, the process is considered normal, and routine monitoring continues. If the first detection result indicates that the entropy change is approaching the threshold, an anomaly is considered to exist, triggering subsequent in-depth detection procedures to further confirm the specific type and extent of the anomaly. Through this preliminary judgment method based on entropy state detection, this invention can quickly and efficiently detect potential anomalies during the plasma etching process, providing crucial information for subsequent in-depth detection and process control.

[0023] Furthermore, in the construction of the entropy state detection component, step P20 of this embodiment further includes: P21: Define the intensity change of a unit spectral pixel at different bands and times as the local information entropy value; P22: Construct a first reconstruction node based on the spectral information entropy field reconstructed from the local information entropy value, wherein the reconstruction method is the reconstruction of a one-dimensional spectral sequence into the three-dimensional space of the reaction cavity; P23: Construct a second detection node based on the extraction of entropy dynamic behavior features and the determination of entropy change, wherein the entropy dynamic behavior features include at least the propagation speed of the entropy wave and the growth rate of the abnormally high entropy region; P24: Cascade the first reconstruction node and the second detection node as the entropy state detection component.

[0024] It should be understood that the construction process of the entropy state detection component can be further refined. First, the intensity change of a unit spectral pixel at different wavelengths and times is defined as the local information entropy value. This definition is based on the concept of entropy in information theory, assessing the plasma state by measuring the changes of each spectral pixel in time and wavelength. The essence of abnormal discharge is the runaway process of local entropy increase in the information field. Normal plasma is usually in a low-entropy ordered state, while abnormal discharge exhibits a high-entropy disordered state. By monitoring the dynamic parameters of the entropy field, it can be determined whether the critical point of phase transition from order to disorder is approaching. Specifically, when the correlation length of the entropy field increases sharply, even if no abnormality is detected by visual observation and traditional models, a pre-instability warning can be quickly triggered, thereby achieving early diagnosis of abnormal discharge.

[0025] Next, the first reconstruction node is established by reconstructing the spectral entropy field based on local information entropy values. This reconstruction process extends the one-dimensional spectral sequence to the three-dimensional space of the reaction chamber, mapping the spectral data from a single spectral dimension to the three-dimensional space of the plasma etching reaction chamber. This reconstruction process combines spectral information with the actual physical space of the etching process, thus more intuitively reflecting the distribution and changes of plasma within the etching chamber. Through the first reconstruction node, a three-dimensional spectral entropy field can be generated, providing spatialized data support for subsequent detection.

[0026] Subsequently, a second detection node was established based on the extraction of entropy dynamics characteristics and the determination of entropy changes. The extraction of entropy dynamics characteristics was achieved through dynamic analysis of the spectral entropy field, including key parameters such as the propagation speed of entropy waves and the growth rate of abnormally high-entropy regions. The propagation speed of entropy waves reflects the propagation of plasma state changes in space, while the growth rate of abnormally high-entropy regions indicates the area and speed of increased plasma disorder. Through the extraction and analysis of these characteristics, entropy changes can be more accurately determined, thereby enabling timely detection of anomalies in the plasma etching process.

[0027] Finally, the first reconstruction node and the second detection node are cascaded to form a complete entropy state detection component. This cascaded structure enables a complete detection process from spectral data acquisition to the final determination of the entropy state. The first reconstruction node is responsible for reconstructing the spectral data into an entropy field in three-dimensional space, while the second detection node extracts entropy dynamics characteristics and determines entropy changes based on the reconstructed entropy field. Through this cascaded approach, the entropy state detection component can efficiently complete the initial judgment of the plasma etching process state and trigger subsequent in-depth detection processes when an anomaly is detected.

[0028] Furthermore, based on the entropy state detection component, spectral information entropy field conversion and entropy change analysis are performed on the spectral pixels to determine the first detection result. Step P20 in this embodiment of the application also includes: P25: Import the multispectral data into the entropy state detection component, extract and reconstruct the entropy value for each spectral pixel, and use it as the real-time spectral information entropy field; P26: Extract entropy dynamic behavior features based on the second detection node, determine whether it approaches the entropy change threshold, and generate the first detection result.

[0029] Optionally, the first detection result is determined by performing spectral information entropy field transformation and entropy change analysis on the spectral pixels using the entropy state detection component. First, the acquired multispectral data is imported into the entropy state detection component. Within this component, entropy values ​​are extracted and reconstructed for each spectral pixel. This process is based on previously defined local information entropy values. By quantifying the intensity changes of each spectral pixel at different wavelengths and times, the information entropy value of each pixel is obtained. Subsequently, using these local information entropy values, a real-time spectral information entropy field is generated according to the reconstruction method from a one-dimensional spectral sequence to the three-dimensional space of the reaction cavity. The entropy value of each spectral pixel reflects the plasma state at that location, including its stability, disorder, and other key characteristics. This real-time spectral information entropy field can reflect the dynamic changes of spectral information during the plasma etching process, providing a basis for subsequent entropy change analysis.

[0030] Next, entropy dynamics characteristics are extracted based on the second detection node. For example, the real-time spectral entropy field is dynamically analyzed to extract key entropy dynamics characteristics, such as the propagation speed of entropy waves and the growth rate of abnormally high-entropy regions. By analyzing these characteristics, it can be determined whether the current state of the plasma etching process is approaching a set entropy change threshold. The entropy change threshold is the critical point where the entropy state changes from order to disorder. When the entropy dynamics characteristics indicate that the entropy change is approaching or exceeds this threshold, it means that the process may be abnormal. Based on this, a first detection result is generated, which will serve as the basis for whether to trigger the subsequent deep detection process.

[0031] Furthermore, step P20 in this embodiment of the application also includes: P27: If the entropy change threshold is not approached, the first detection result is a normal discharge diagnosis result and no diagnostic response is made.

[0032] In one possible embodiment of this application, if, after extracting entropy dynamics characteristics, it is determined that the entropy change of the spectral information entropy field does not approach the set entropy change threshold, the first detection result will be judged as a normal discharge diagnosis result, indicating that the current plasma etching process is in normal operation and no abnormal discharge has occurred. Therefore, there is no need to trigger the subsequent depth detection process, nor to perform any diagnostic response. In other words, the system confirms the stability of the process in the preliminary detection stage and can continue to operate according to the predetermined etching process parameters, thereby ensuring the efficient and stable operation of the plasma etching process while avoiding unnecessary consumption of detection resources and process interruptions.

[0033] P30: If the first detection result is close to the entropy change threshold, the plasma detection component is triggered to perform a verification based on depth time-frequency characteristics to determine the second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder.

[0034] Specifically, when the entropy state detection component determines that the first detection result is close to the entropy change threshold, it indicates that there may be an anomaly in the plasma etching process. At this time, the plasma detection component will be triggered to perform a depth-based time-frequency feature-based verification to further determine the second detection result.

[0035] Specifically, the entropy change threshold is the critical point where the entropy state transitions from order to disorder. When the entropy change of the spectral information entropy field approaches or exceeds this threshold, it indicates that the stability of the plasma may be disturbed, resulting in abnormal discharge or other unstable phenomena. To more accurately diagnose the specific type and extent of the anomaly, the plasma detection component will perform in-depth analysis of the spectral data in the plasma etching process, extracting the temporal intensity and frequency domain characteristics of the spectral data. The temporal intensity reflects the changes in the plasma over time, such as power fluctuations and pulse width variations; the frequency domain characteristics reflect the distribution of the plasma in frequency, such as harmonics and noise. Through comprehensive analysis of these time-frequency characteristics, a more comprehensive understanding of the plasma state changes can be obtained. Furthermore, based on the verification results of the in-depth time-frequency characteristics, a second detection result is determined. The second detection result will clearly indicate whether abnormal discharge exists in the plasma etching process, as well as the specific type and extent of the anomaly. This result will provide an important basis for subsequent process feedback control, enabling timely measures to correct the anomaly and ensure the stability and reliability of the plasma etching process.

[0036] Furthermore, prior to triggering the plasma detection component, the construction of the plasma detection component, in step P30 of this embodiment, further includes: P31: A plasma detection component is built by embedding a plasma spectral library and supervising training to convergence; wherein, the embedded plasma spectral library includes: P31-1: defining deep time-frequency features, wherein the deep time-frequency features include at least the plasma excited-state particle population density, electron temperature spatial gradient, and reactive group concentration ratio; P31-2: retrieving historical plasma etching data and mining the plasma spectral library based on the deep time-frequency features, wherein the abnormal discharge causes are used to mine normal feature sequences and abnormal mutation sites, and the abnormal discharge causes include at least process parameter deviation, component aging, and foreign matter intrusion.

[0037] P32: The plasma detection component is connected to the entropy state detection component, wherein the plasma detection component is adaptively activated based on the output state of the entropy state detection component; wherein the entropy state detection component and the plasma detection component are central control embedded plugs of the plasma etching stage.

[0038] It should be understood that before triggering the plasma detection component, it is necessary to first construct the plasma detection component to ensure that it can perform effective in-depth analysis and anomaly diagnosis after receiving the signal output by the entropy state detection component.

[0039] Specifically, a plasma detection component is built by embedding a plasma spectral library and supervising its training to convergence. The construction of the embedded plasma spectral library first requires defining deep time-frequency features, which include at least the plasma excited-state particle population density, electron temperature spatial gradient, and reactive group concentration ratio. The plasma excited-state particle population density reflects the number of excited-state particles in the plasma and can be used to determine the electron density and temperature of the plasma. The electron temperature spatial gradient represents the degree of spatial variation in electron temperature within the plasma and can help assess the thermodynamic properties of the plasma. The reactive group concentration ratio helps to understand the consumption and generation relationship of reactants during the etching process, further improving the sensitivity to process variations.

[0040] Subsequently, historical plasma etching data was retrieved, and a plasma spectral library based on the aforementioned depth and time-frequency characteristics was mined. The mining process was based on the causes of anomalous discharges, identifying normal characteristic sequences and anomalous abrupt change points. By analyzing historical data, with particular focus on the causes of anomalous discharges, normal characteristic sequences and anomalous abrupt change points were identified. Causes of anomalous discharges may include deviations in process parameters, such as abnormal fluctuations in gas flow rate, pressure, and power; performance degradation due to aging of equipment components; or the intrusion of external foreign objects, all of which can affect plasma stability and etching quality. Through this process, a plasma spectral library containing both normal and anomalous states can be established, providing a data foundation for subsequent supervised training.

[0041] After the spectral library is constructed, the plasma detection component and the entropy state detection component are integrated. The plasma detection component is connected after the entropy state detection component and is adaptively activated based on the output state of the entropy state detection component. This means that the plasma detection component will only be activated for depth analysis when the entropy change value is close to or exceeds the set entropy change threshold. Specifically, the output state of the entropy state detection component, i.e., the first detection result, will determine whether to trigger the plasma detection component. When the entropy state detection component detects that the entropy change is close to the threshold, it indicates that there may be an abnormal discharge. At this time, the plasma detection component will be activated to perform depth time-frequency feature verification. Both the entropy state detection component and the plasma detection component are embedded plug-ins in the central control of the plasma etching stage. This design allows the two components to be seamlessly integrated into the control system of the plasma etching stage, realizing an automated process from preliminary detection to depth diagnosis, thereby improving detection efficiency and enhancing the stability and reliability of the system.

[0042] Furthermore, a detection based on deep time-frequency features is performed to determine the second detection result. Step P30 in this embodiment of the application also includes: P33: If the first detection result is close to the entropy change threshold, trigger the plasma detection component and import the multispectral data into the plasma detection component; P34: Perform deep time-frequency feature extraction on the multispectral data, and locate abnormal mutation points by comparing with normal feature sequences; P35: If the abnormal mutation points are not an empty set, integrate them to generate the second detection result.

[0043] Specifically, if the first detection result is close to the entropy change threshold, this indicates that the preliminary detection has identified a possible anomaly in the plasma etching process. At this point, the plasma detection component is triggered, and the multispectral data is imported into the component for in-depth analysis.

[0044] Subsequently, deep time-frequency feature extraction is performed on the imported multispectral data. This involves dual analysis of the multispectral data in both the time and frequency domains to identify the dynamic behavior of the plasma and potential abnormal discharge characteristics. Time-frequency analysis allows for the extraction of important features from the data, such as signal frequency fluctuations and amplitude variations. These features typically exhibit significant differences between abnormal discharges and normal process conditions. Therefore, the extracted time-frequency features are compared with normal feature sequences to further determine whether any abrupt changes occur.

[0045] Finally, by comparing with normal feature sequences, abnormal mutation points in the data are located. If the comparison results show obvious abnormal mutations, and these mutation points are not empty sets, the system will integrate this information to generate a second detection result. The second detection result will comprehensively consider the time-frequency characteristics of the multispectral data, the deviation from the normal feature sequence, and the distribution of abnormal mutation points, providing a reliable basis for subsequent abnormal discharge diagnosis. If abnormal mutation points exist, the second detection result will indicate that there may be abnormal discharge or other unstable factors in the process, and corresponding measures can be taken based on the result, such as adjusting process parameters or performing equipment maintenance. If no abnormal mutation points are found, the second detection result confirms that the process is in a normal state, and the system will continue to execute the current process flow.

[0046] P40: Determine the abnormal discharge diagnosis result based on the second detection result, and perform feedback control on the plasma etching process.

[0047] Furthermore, based on the second detection result, the abnormal discharge diagnosis result is determined. In this embodiment, step P40 further includes: P41: Read the second detection result and determine the abnormal sequence, wherein each abnormal sequence includes at least the cause of abnormal discharge, the location of abnormal discharge, and the level of abnormal discharge; P42: Integrate the abnormal sequences as the abnormal discharge diagnosis result, wherein the integration method includes at least fusion based on the same abnormal discharge location.

[0048] Optionally, the abnormal discharge diagnosis result can be determined based on the second detection result, and the plasma etching process can be controlled accordingly.

[0049] First, the second detection result is read and relevant abnormal sequences are extracted. These abnormal sequences are generated by identifying anomalous abrupt change points during deep time-frequency feature extraction and comparison. Their elements include at least the cause of the abnormal discharge, the location of the abnormal discharge, and the level of the abnormal discharge. The cause of the abnormal discharge refers to the specific reason for the abnormal discharge, such as process parameter deviation, component aging, or foreign object intrusion. The location of the abnormal discharge refers to the specific location where the abnormal discharge occurs within the plasma etching equipment. The level of the abnormal discharge is a classification based on the severity of the anomaly, used to indicate the urgency and priority of its handling. For example, first, the specific cause of the abnormal discharge is analyzed, such as process parameter deviation, equipment aging, or external interference, to quickly identify the root cause of the problem. Next, the specific location where the abnormal discharge occurs is calibrated, i.e., a specific area of ​​the plasma etching stage, to help determine whether there are unstable factors in certain parts of the equipment or environment. Finally, the severity of the abnormal discharge is classified, possibly into minor, moderate, or severe, and corresponding abnormal sequences are generated.

[0050] After identifying the abnormal sequences, an integration process is performed to combine multiple abnormal sequences into a comprehensive abnormal discharge diagnostic result. Integration methods can include fusion based on the location of the abnormal discharge. For example, if multiple abnormal sequences all point to an abnormal discharge at the same location, the relevant information from these sequences can be fused to generate a comprehensive abnormal discharge diagnostic result, thereby determining whether there is a systemic fault or potential risk in that area. This method can identify areas where anomalies occur frequently, providing specific basis for equipment maintenance or process adjustments in those areas. This integration method can more effectively organize and present abnormal situations, avoiding low processing efficiency caused by information redundancy or dispersion.

[0051] Finally, based on the integrated abnormal discharge diagnosis results, feedback control is implemented on the plasma etching process, including adjusting etching parameters, pausing the etching process, issuing alarms, and triggering maintenance procedures. This ensures that abnormal situations are corrected in a timely manner, the normal operation of the plasma etching process is maintained, and product quality and process reliability are improved.

[0052] Furthermore, to provide feedback control over the plasma etching process, step P40 in this embodiment of the application also includes: P43: Match the abnormal discharge diagnosis results with the abnormal contingency plan library to determine the abnormal discharge contingency plan; P44: Write the abnormal discharge contingency plan into the control terminal of the plasma etching stage to perform control and management under the plasma etching process.

[0053] In one possible embodiment of this application, in addition to determining the abnormal discharge diagnostic results, feedback control of the plasma etching process can also be performed based on these diagnostic results.

[0054] After confirming the abnormal discharge diagnosis, the results are matched against an anomaly contingency plan database to determine the appropriate contingency plan. The database is a pre-built database containing contingency plans for various possible abnormal discharge scenarios. These plans are developed based on historical data, expert experience, and simulation analysis, covering a range of situations from minor anomalies to severe faults. By matching the abnormal discharge diagnosis results with the plans in the database, the most suitable solution for the current anomaly can be quickly found. The matching process considers multiple factors such as the cause, location, and severity of the abnormal discharge to ensure that the selected plan accurately addresses the current problem.

[0055] After determining the contingency plan for abnormal discharge, it is written into the control terminal of the plasma etching station for implementation and management of the plasma etching process. The control terminal is the core control system of the plasma etching equipment, responsible for adjusting various process parameters and operating the equipment. Once the abnormal discharge contingency plan is written into the control terminal, the system will automatically manage and control the etching process according to the instructions in the plan. This includes adjusting parameters such as etching power, gas flow rate, and pressure; pausing or terminating the etching process; issuing alarms to operators; and even triggering automatic maintenance procedures. Through this automated feedback control mechanism, abnormal situations can be responded to quickly, reducing delays in manual intervention, improving the stability and reliability of the etching process, and reducing the risk of product defects and equipment damage caused by abnormal discharge.

[0056] In summary, the embodiments of this application have at least the following technical effects: This application, by acquiring multispectral data in real time and combining entropy state detection with time-frequency feature analysis, can accurately identify abnormal discharges in the early stages of the plasma etching process, reducing their impact on the process. Employing entropy change analysis and deep time-frequency feature extraction significantly improves the accuracy and sensitivity of abnormal discharge detection, effectively reducing the false alarm rate. Real-time control of the etching process based on the detection results optimizes process parameters, ensuring the stability and efficiency of the plasma etching process. Through early diagnosis and precise control, the impact of abnormal discharges on etching quality can be reduced, improving production efficiency and product quality.

[0057] This technology achieves the technical effect of enabling early diagnosis and high-precision identification of abnormal discharges through multispectral fusion and in-depth analysis, thereby improving the stability and production efficiency of the etching process.

[0058] Example 2, based on the same inventive concept as the multispectral fusion plasma etching stage abnormal discharge diagnosis method in the foregoing examples, such as... Figure 2 As shown, this application provides a multispectral fusion plasma etching stage abnormal discharge diagnostic device. The device and method embodiments in this application are based on the same inventive concept. The device includes: The spectral detection module 11 is used to perform spectral detection at key locations via a miniature spectral probe as the plasma etching process progresses, thereby determining multispectral data.

[0059] The early diagnosis module 12 is used to upload the multispectral data, and according to the entropy state detection component, to perform spectral information entropy field conversion and entropy change analysis on the spectral pixels to determine the first detection result.

[0060] The depth verification module 13 is used to trigger the plasma detection component to perform a verification based on depth time-frequency characteristics if the first detection result is close to the entropy change threshold, and to determine the second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder.

[0061] The feedback control module 14 is used to determine the abnormal discharge diagnosis result based on the second detection result and to perform feedback control on the plasma etching process.

[0062] Furthermore, the early diagnosis module 12 is also used to perform the following steps: The intensity change of a unit spectral pixel at different bands and times is defined as the local information entropy value. A first reconstruction node is built by reconstructing the spectral information entropy field based on the local information entropy value, wherein the reconstruction method is the reconstruction of a one-dimensional spectral sequence to the three-dimensional space of the reaction cavity. A second detection node is built by extracting entropy dynamic behavior features and determining entropy change, wherein the entropy dynamic behavior features include at least the propagation speed of the entropy wave and the growth rate of the abnormally high entropy region. The first reconstruction node and the second detection node are cascaded as the entropy state detection component.

[0063] Furthermore, the early diagnosis module 12 is also used to perform the following steps: The multispectral data is imported into the entropy state detection component, and entropy values ​​are extracted and reconstructed for each spectral pixel to form a real-time spectral information entropy field. Based on the second detection node, entropy dynamic behavior features are extracted to determine whether the entropy change threshold is approached, and the first detection result is generated.

[0064] Furthermore, the early diagnosis module 12 is also used to perform the following steps: If the entropy change threshold is not approached, the first detection result is a normal discharge diagnosis result, and no diagnostic response is made.

[0065] Furthermore, the depth detection module 13 is also used to perform the following steps: A plasma detection component is built by embedding a plasma spectral library and supervising its training to convergence. The plasma detection component is then connected to the entropy state detection component, wherein the plasma detection component is adaptively activated based on the output state of the entropy state detection component. The entropy state detection component and the plasma detection component are central control embedded plug-ins of the plasma etching stage.

[0066] Furthermore, the depth detection module 13 is also used to perform the following steps: Define deep time-frequency features, wherein the deep time-frequency features include at least plasma excited-state particle population density, electron temperature spatial gradient, and reactive group concentration ratio; retrieve historical plasma etching data, and mine a plasma spectrum library based on the deep time-frequency features, wherein the abnormal discharge causes are used to mine normal feature sequences and abnormal mutation sites, and the abnormal discharge causes include at least process parameter deviation, component aging, and foreign object intrusion.

[0067] Furthermore, the depth detection module 13 is also used to perform the following steps: If the first detection result is close to the entropy change threshold, the plasma detection component is triggered, and the multispectral data is imported into the plasma detection component; deep time-frequency feature extraction is performed on the multispectral data, and abnormal mutation points are located by comparing with normal feature sequences; if the abnormal mutation points are not an empty set, the second detection result is generated.

[0068] Furthermore, the feedback control module 14 is also used to perform the following steps: Read the second detection result and determine the abnormal sequence, wherein each abnormal sequence includes at least the cause of abnormal discharge, the location of abnormal discharge, and the level of abnormal discharge; integrate the abnormal sequences as the abnormal discharge diagnosis result, wherein the integration method includes at least fusion based on the same abnormal discharge location.

[0069] Furthermore, the feedback control module 14 is also used to perform the following steps: Based on the abnormal discharge contingency plan database, the abnormal discharge diagnosis results are matched to determine the abnormal discharge contingency plan; the abnormal discharge contingency plan is written into the control terminal of the plasma etching stage to execute the control and management of the plasma etching process.

[0070] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0071] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0072] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A multispectral fusion-based method for diagnosing abnormal discharges in a plasma etching stage, characterized in that, The method includes: As the plasma etching process progresses, a micro-spectral probe is used to perform spectral detection at key locations to determine multispectral data. Upload the multispectral data, and perform spectral information entropy field conversion and entropy change analysis on the spectral pixels according to the entropy state detection component to determine the first detection result; If the first detection result is close to the entropy change threshold, the plasma detection component is triggered to perform a detection based on depth time-frequency characteristics to determine the second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder; Based on the second detection result, the abnormal discharge diagnosis result is determined, and feedback control is performed on the plasma etching process.

2. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 1, characterized in that, The construction of the entropy state detection component includes: The intensity change of a unit spectral pixel in different bands and at different times is defined as the local information entropy value. The first reconstruction node is built by reconstructing the spectral information entropy field based on the local information entropy value, wherein the reconstruction method is the reconstruction of a one-dimensional spectral sequence to the three-dimensional space of the reaction chamber; A second detection node is built by extracting entropy dynamic behavior features and determining entropy change. The entropy dynamic behavior features include at least the propagation speed of entropy waves and the growth rate of abnormally high entropy regions. The first reconstruction node and the second detection node are cascaded together to form the entropy state detection component.

3. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 2, characterized in that, Perform spectral information entropy field transformation and entropy change analysis on spectral pixels to determine the first detection result, including: The multispectral data is imported into the entropy state detection component, and entropy values ​​are extracted and reconstructed for each spectral pixel to serve as the real-time spectral information entropy field. Based on the second detection node, entropy dynamics behavior features are extracted, it is determined whether the entropy change threshold is approached, and the first detection result is generated.

4. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 3, characterized in that, If the entropy change threshold is not approached, the first detection result is a normal discharge diagnosis result, and no diagnostic response is made.

5. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 1, characterized in that, Before triggering the plasma detection component, the construction of the plasma detection component includes: A plasma detection component was built by embedding a plasma spectral library and supervising training to convergence. The plasma detection component is connected to the entropy state detection component in a post-position, wherein the plasma detection component is adaptively activated based on the output state of the entropy state detection component; The entropy state detection component and the plasma detection component are embedded plug-ins in the central control of the plasma etching stage.

6. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 5, characterized in that, Embedded plasma spectral library, including: Define deep time-frequency characteristics, wherein the deep time-frequency characteristics include at least the plasma excited-state particle population density, electron temperature spatial gradient, and reactive group concentration ratio; Historical plasma etching data is retrieved, and a plasma spectrum library based on the time-frequency characteristics of the depth is mined. Among them, the causes of abnormal discharge are used to mine normal feature sequences and abnormal mutation points. The causes of abnormal discharge include at least process parameter deviation, component aging, and foreign object intrusion.

7. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 6, characterized in that, Perform detection based on deep time-frequency features to determine the second detection result, including: If the first detection result is close to the entropy change threshold, the plasma detection component is triggered, and the multispectral data is imported into the plasma detection component; Deep time-frequency feature extraction is performed on the multispectral data, and abnormal mutation points are located by comparing them with normal feature sequences. If the abnormal mutation sites are not an empty set, the second detection result is generated by integrating them.

8. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 1, characterized in that, The abnormal discharge diagnosis result is determined based on the second detection result, including: Read the second detection result and determine the abnormal sequence, wherein each abnormal sequence includes at least the cause of abnormal discharge, the location of abnormal discharge, and the level of abnormal discharge; The abnormal sequence is integrated as the abnormal discharge diagnosis result, wherein the integration method includes at least fusion based on the same abnormal discharge location.

9. The multispectral fusion plasma etching stage abnormal discharge diagnosis method as described in claim 1, characterized in that, Feedback control of the plasma etching process includes: Based on the abnormal discharge contingency plan database, the abnormal discharge diagnosis results are matched to determine the abnormal discharge contingency plan; The abnormal discharge contingency plan is written into the control terminal of the plasma etching stage to perform regulation and management during the plasma etching process.

10. A multispectral fusion plasma etching stage abnormal discharge diagnostic device, characterized in that, The device includes: The spectral detection module is used to perform spectral detection at key locations via a miniature spectral probe as the plasma etching process progresses, in order to determine multispectral data. The early diagnosis module is used to upload the multispectral data, and perform spectral information entropy field conversion and entropy change analysis on the spectral pixels according to the entropy state detection component to determine the first detection result; The depth detection module is used to trigger the plasma detection component to perform a depth-based time-frequency characteristic detection if the first detection result is close to the entropy change threshold, and to determine the second detection result, wherein the entropy change threshold is the critical point from entropy state order to disorder. The feedback control module is used to determine the abnormal discharge diagnosis result based on the second detection result and to control the plasma etching process.

Images