Metal pollution online monitoring method and system, electronic equipment, storage medium and computer program product

Abstract

The invention relates to a metal pollution online monitoring method and system, electronic equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring spectral data in a process cavity in a semiconductor process; calculating the SPE statistic of the spectral data according to a principal component analysis model, taking the spectral data as a pollution spectrum when the SPE statistic is judged to exceed a preset control limit, and entering a metal pollution analysis step which comprises the following steps: acquiring a closest reference spectrum based on a process time stage of the pollution spectrum, and calculating a difference spectrum between the pollution spectrum and the reference spectrum; calculating the pollution index of the to-be-detected metal based on the original spectral data and the differential spectrum; and matching the pollution index of the to-be-detected metal with a preset metal characteristic spectrum database, when a matching result meets a preset condition, determining that metal pollution exists in the process cavity, and outputting an alarm signal. According to the application, early warning and quick response to metal pollution can be realized, so that the process control level and the product yield of semiconductor manufacturing are improved.

Description

Technical Field

[0001] This invention relates to the field of semiconductor process monitoring technology, specifically to an online monitoring method, system, electronic device, storage medium, and computer program product for metal contamination based on optical emission spectroscopy. Background Technology

[0002] In semiconductor manufacturing, plasma technology is widely used in critical process steps such as etching, chemical vapor deposition (CVD), and ion implantation. However, during plasma processes, components such as the inner walls of the reaction chamber, electrodes, and fixtures may release trace amounts of metallic contaminants (such as Fe, Cu, Al, Ni, and Cr) under the continuous bombardment and corrosion of high-temperature and highly reactive plasma. Once these metallic contaminants deposit on the wafer surface, they can lead to degradation of device electrical performance, failure of gate oxide integrity, increased leakage current, and even a significant decrease in the yield of the entire batch of products. Therefore, strict monitoring and control of metallic contamination in the plasma process environment is a crucial step in ensuring semiconductor manufacturing yield and product reliability.

[0003] Currently, the monitoring of metal pollution mainly relies on various offline detection methods, including: 1) Wafer Acceptance Test (WAT): This test is performed after a specific stage or all processes of the wafer are completed. It indirectly infers the contamination that may exist in the process by measuring the electrical parameters (such as threshold voltage, leakage current, breakdown voltage, etc.) of the test structure (such as transistors, capacitors). However, it has a serious lag and can only detect problems after the wafer is processed. It cannot intervene at the moment when contamination occurs, which may lead to the scrapping of a large number of wafers. 2) Offline Particle Monitoring (OPM): Periodically extracting gas samples from the process chamber or using a portable particle detector to sample the chamber environment to assess the level of particulate contamination; however, it lacks real-time capability. 3) Preventive Maintenance (PM): The plasma equipment is shut down and disassembled according to a preset time cycle or cumulative operating time. Then, the wear and corrosion of the inner wall of the chamber and components are manually inspected and cleaned or replaced. This method is costly and will lead to a decrease in equipment utilization. Moreover, it is a preventive maintenance and cannot predict or deal with sudden metal contamination caused by component abnormalities between two maintenance cycles. 4) Time-of-Flight Secondary Ion Mass Spectrometry (TOF-SIMS): This method uses a primary ion beam (such as Bi) to...+ Ga + The process involves bombarding the surface of the wafer under test, causing sputtering of surface atoms, molecules, and atomic groups (i.e., generating secondary ions). Subsequently, time-of-flight mass spectrometry analysis is performed on these secondary ions, and the elemental and even molecular composition is determined based on their mass-to-charge ratio, thereby determining the type and distribution of metal contamination on the wafer surface. Although TOF-SIMS has high surface sensitivity and rich chemical information, its analysis process requires offline processing of the sample in a vacuum environment, which is time-consuming and can cause some damage to the sample surface, making it unsuitable for real-time monitoring in the process. 5) Inductively Coupled Plasma Mass Spectrometry (ICP-MS): This method typically requires sampling contaminants on the wafer surface. The sample solution is atomized into an aerosol, which is then introduced into a high-temperature argon plasma generated by a high-frequency radio frequency induction coil by a carrier gas. The sample undergoes evaporation, decomposition, atomization, and ionization in the plasma center region, forming monovalent positive ions. These ions are then introduced into the mass spectrometer for separation and detection according to their mass-to-charge ratio, thereby accurately determining the concentration of metal elements. Although ICP-MS has extremely high sensitivity, its sample pretreatment is complex and the analysis cycle is long, making it impossible to achieve real-time online analysis during the process.

[0004] In summary, none of the aforementioned detection methods can achieve real-time, online, and highly sensitive monitoring of metal contamination during plasma processing. They either suffer from significant time lag, requiring post-contamination confirmation, or necessitate interrupting production for sampling or equipment disassembly, leading to reduced efficiency and increased costs. This lack of monitoring capability prevents process engineers from obtaining early warning information the instant contamination occurs, hindering timely adjustments to process parameters or shutdowns for investigation, thus failing to effectively prevent batch quality incidents caused by sudden contamination.

[0005] Therefore, there is a need for a method that can be integrated with plasma process equipment, without interrupting production, and can monitor the level of metal contamination in the process environment in real time and online while the process is in progress. Summary of the Invention

[0006] The technical problem to be solved by this application is to provide a method, system, electronic device, storage medium and computer program product for online monitoring of metal contamination based on optical emission spectrum, which can monitor the level of metal contamination in process chambers in real time and online, so as to realize early warning and rapid response to metal contamination, and fundamentally improve the process control level and product yield of semiconductor manufacturing.

[0007] According to a first aspect of the embodiments of this application, an online monitoring method for metal pollution is provided, comprising: Acquire spectral data within the process cavity during semiconductor manufacturing; The SPE statistic of the spectral data is calculated based on the principal component analysis model. When the SPE statistic exceeds the preset control limit, the spectral data is used as a contamination spectrum and the process proceeds to the metal contamination analysis step. The metal contamination analysis step includes: obtaining the closest reference spectrum based on the process time stage of the contamination spectrum, and calculating the difference spectrum between the contamination spectrum and the reference spectrum. The reference spectrum is obtained from the raw spectral data collected by running the semiconductor process multiple times in a clean process chamber. Based on the original spectral data and the difference spectrum, the pollution index of the metal to be tested is calculated; The contamination index of the metal to be tested is matched with a preset metal characteristic spectrum database. When the matching result meets the preset conditions, it is determined that there is metal contamination in the process chamber, and an alarm signal is output.

[0008] In one implementation, obtaining the closest reference spectrum based on the process time stage of the pollution spectrum includes: obtaining the original spectral data; Based on preset time stage criteria, a subset of spectra that meet the target time stage is selected from the original spectral data; The mean spectrum and variance spectrum of the spectral subset are calculated and standardized to generate a reference spectrum that represents the spectral characteristics of the target time period, which is the closest reference spectrum.

[0009] In one embodiment, the steps for establishing the preset metal characteristic spectrum database are as follows: Based on the semiconductor process and the original spectral data, characteristic spectral lines of at least two metals to be tested are determined; reference spectra corresponding to different process time stages in the semiconductor process are obtained; and the reference index CL corresponding to the characteristic spectral line of each metal to be tested in each reference spectrum is calculated. m The process time stage and the benchmark index CL for each metal being tested will be used. m Establish corresponding relationships to form a metal characteristic spectrum database; Specifically, the contamination index of the metal to be tested is matched with a preset metal characteristic spectrum database. When the matching result meets preset conditions, it is determined that metal contamination exists in the process chamber, including: Based on the process time stage of the pollution spectrum, the reference index CL of the metal to be tested is obtained from the preset metal characteristic spectrum database. m ; The pollution index CL of the metal to be tested is compared with the benchmark index CL corresponding to the metal to be tested. mIf the comparison result meets the preset condition, it is determined that metal contamination exists in the process chamber. The preset condition is that CL is greater than CL within a preset time period. m .

[0010] In one embodiment, calculating the pollution index of the metal to be tested based on the original spectral data and the difference spectrum includes: Obtain the reference intensity ratio of the first characteristic peak and the second characteristic peak of the inert gas, wherein the first characteristic peak and the second characteristic peak of the inert gas are selected according to the semiconductor process, and the reference intensity ratio is the spectral intensity ratio of the first characteristic peak and the second characteristic peak of the inert gas in the original spectral data collected in the clean process chamber; Based on the difference spectrum, the ratio of the spectral intensity of the first characteristic peak of the inert gas to the second characteristic peak of the inert gas is calculated and used as the current ratio. When the current ratio is compared with the reference intensity ratio and the difference between the two is within a preset range, the ratio of the characteristic wavelength spectral intensity of the pollutant to be tested in the difference spectrum to the spectral intensity of the first characteristic peak of the inert gas in the original spectral data is calculated as the pollution index of the metal to be tested.

[0011] In one embodiment, the online monitoring method for metal pollution further includes: When the current ratio is compared with the reference intensity ratio and the difference between the two exceeds a preset range, the reference spectrum is calibrated based on the difference spectrum to generate a background spectrum; Based on the difference spectrum, the background spectrum is subtracted from the characteristic wavelength spectrum of the metal to be tested and the spectral lines corresponding to the first characteristic peak of the inert gas, respectively, and the respective background-removed peak areas are calculated and denoted as I'1 and I'2. The ratio of the area of ​​the background-removed peak corresponding to the metal to be tested to the area of ​​the background-removed peak corresponding to the first characteristic peak of the inert gas is calculated as the pollution index of the metal to be tested.

[0012] In one embodiment, the steps for establishing the principal component analysis model include: performing N semiconductor processes within a cleanroom, acquiring raw spectral data using an OES sensor, wherein the raw spectral data is acquired at M time points during each process, and each time point covers a full-band spectrum of P wavelengths; constructing a two-dimensional data matrix from the raw spectral data, wherein the dimension of the two-dimensional data matrix is ​​(N × M) rows × P columns; standardizing the two-dimensional data matrix column-wise to obtain a reference spectral matrix; calculating its covariance matrix based on the reference spectral matrix, and solving for the eigenvalues ​​and eigenvectors of the covariance matrix, defining the eigenvectors as principal component directions; The step of calculating the SPE statistic of the spectral data based on the principal component analysis model includes: The spectral data is projected onto the principal component direction to calculate the SPE statistic.

[0013] In one embodiment, when it is determined that the SPE statistic exceeds a preset control limit, the spectral data is used as the pollution spectrum and the process proceeds to the metal pollution analysis step. The process further includes: when it is determined that all the SPE statistics within a certain time period exceed the preset control limit, the average value of the corresponding spectral data within the certain time period is used as the pollution spectrum and the process proceeds to the metal pollution analysis step.

[0014] According to a second aspect of the embodiments of this application, an online monitoring system for metal pollution is provided, which includes a spectral acquisition module, a first processing module, and a second processing module connected together, wherein... The spectral acquisition module is used to acquire spectral data within the process chamber in real time. The first processing module is used to calculate the SPE statistic of the spectral data according to the principal component analysis model, and when it is determined that the SPE statistic exceeds the preset control limit, output a trigger command to enter the metal pollution analysis step. The second processing module is used to execute the metal contamination analysis step when the trigger command is received. The metal contamination analysis step includes: obtaining the closest reference spectrum based on the process time stage of the contamination spectrum, and calculating the difference spectrum between the contamination spectrum and the reference spectrum, wherein the reference spectrum is obtained from the raw spectral data collected by running the semiconductor process multiple times in the clean process chamber. Based on the original spectral data and the difference spectrum, the pollution index CL of the metal to be tested is calculated; The contamination index of the metal to be tested is matched with a preset metal characteristic spectrum database. When the matching result meets the preset conditions, it is determined that there is metal contamination in the process chamber, and an alarm signal is output.

[0015] In one embodiment, the online monitoring system further includes a data storage module connected to the first processing module and the second processing module. The data storage module is used to store the principal component analysis model, the preset metal characteristic spectrum database, and the reference spectrum.

[0016] According to a third aspect of the embodiments of this application, an electronic device is provided, comprising a memory and a processor, the memory being configured to store a computer program executable by the processor; the processor being configured to execute the computer program in the memory to implement the method as described in any of the preceding claims.

[0017] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, wherein when the executable computer program in the storage medium is executed by a processor, it is capable of implementing the method described in any of the preceding claims.

[0018] According to a fifth aspect of the embodiments of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the method as described in any of the preceding claims.

[0019] Compared with the prior art, the beneficial effects of this application are as follows: This application first performs principal component analysis on the real-time acquired spectral data. Only when the SPE statistic exceeds the preset control limit is the metal contamination analysis step triggered. This eliminates the need to calculate the metal contamination analysis index for all real-time spectral data, enabling real-time, online monitoring of metal contamination levels within the process chamber. This avoids performing metal contamination level analysis on every real-time spectrum, improving analytical efficiency and accuracy. Furthermore, the metal contamination analysis step employs differential spectral analysis, and the reference spectrum used for differential spectral analysis is selected to be closest to the process time stage of the contamination spectrum. Through a time alignment mechanism, only the original spectral data (i.e., historical data) corresponding to the process time stage in which the contamination spectrum occurred is analyzed. This eliminates the need for global processing of the original spectral data, improving analytical efficiency and accuracy. It enables early warning and rapid response to metal contamination, fundamentally improving the process control level and product yield in semiconductor manufacturing. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating an online monitoring method for metal pollution according to an exemplary embodiment; Figure 2 This is a flowchart illustrating a metal contamination analysis according to an exemplary embodiment; Figure 3 This is a block diagram illustrating an online monitoring system for metal pollution according to an exemplary embodiment; Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0021] Unless otherwise defined, the technical or scientific terms used in this specification and claims should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. Specific embodiments of the invention will be described below with reference to the accompanying drawings. It should be noted that, in order to provide a concise description, this specification cannot provide a detailed description of all features of the actual embodiments. Without departing from the spirit and scope of the invention, those skilled in the art can make modifications and substitutions to the embodiments of the invention, and the resulting embodiments are also within the protection scope of the invention.

[0022] In related technologies, the monitoring of metal contamination during plasma processes suffers from severe time lag, and can only be confirmed retrospectively after contamination occurs; or the production process needs to be interrupted for sampling or equipment dismantling, resulting in reduced production efficiency and increased costs.

[0023] To address the aforementioned technical problems, this application proposes a method, device, electronic equipment, and storage medium for online monitoring of metal contamination. This method can monitor the level of metal contamination in process chambers in real time and online, enabling early warning and rapid response to metal contamination, thereby fundamentally improving the process control level and product yield in semiconductor manufacturing.

[0024] One embodiment of this application provides a method for online monitoring of metal contamination. This method can be applied to electronic devices such as control computers in OES equipment and central controllers in chip manufacturing processes. Please refer to... Figure 1 The online monitoring method for metal pollution may include the following steps 201 to 203: Step 201: Obtain spectral data within the process cavity of the semiconductor process; specifically, optical emission spectroscopy, i.e., an OES sensor, can be used to collect spectral signals within the process cavity in real time. The OES sensor can be embedded in the monitoring window of the process cavity. The semiconductor process can be a plasma process (such as etching, chemical vapor deposition, or resist removal), and is not limited to this step. Step 202: Calculate the SPE statistic of the spectral data according to the principal component analysis model (i.e., PCA model). When it is determined that the SPE statistic exceeds the preset control limit, the spectral data is used as the pollution spectrum and the process proceeds to the metal pollution analysis step, i.e., step 203. When it is determined that the SPE statistic does not exceed the preset control limit, continue to collect spectral data in real time and calculate the SPE statistic. Step 203, the metal contamination analysis includes the following steps 301 to 303: Step 301: Based on the process time stage of the contamination spectrum, obtain the closest reference spectrum, and calculate the difference spectrum between the contamination spectrum and the reference spectrum, wherein the reference spectrum is obtained from the original spectral data collected by running the semiconductor process multiple times in a clean process chamber; Step 302: Calculate the pollution index of the metal to be tested based on the original spectral data and the difference spectrum; Step 303: Match the contamination index of the metal to be tested with a preset metal characteristic spectrum database. When the matching result meets the preset conditions, it is determined that there is metal contamination in the process chamber, and an alarm signal is output.

[0025] This embodiment uses real-time principal component analysis (PCA) to analyze spectral data acquired during semiconductor manufacturing processes. Based on the real-time calculated SPE statistic, a preliminary assessment of metal contamination is made. The metal contamination analysis step is triggered only when the SPE statistic exceeds a preset control limit. This eliminates the need to calculate the metal contamination index for all real-time spectral data, enabling real-time, online monitoring of metal contamination levels within the process chamber. This avoids performing metal contamination level analysis on every real-time spectrum, improving analysis efficiency and accuracy. Furthermore, the metal contamination analysis step employs differential spectral analysis, using a reference spectrum closest to the process time stage of the contamination spectrum. Through a time alignment mechanism, only the original spectral data (i.e., historical data) corresponding to the process time stage where the contamination spectrum occurred is analyzed, eliminating the need for global processing of the original spectral data. This improves analysis efficiency and accuracy, enabling early warning and rapid response to metal contamination, fundamentally enhancing process control and product yield in semiconductor manufacturing.

[0026] In one embodiment, step 201, the sampling process for acquiring spectral data, is as follows: For real-time acquired spectral signals, a sampling window of 5 is set. For example, if 10 frames are continuously acquired, then 1-5 constitute one sampling window, and 2-6 constitute another. The spectral signal within each sampling window is calculated in real time, that is, the average spectrum of the 5 frames within the sampling window is calculated as the current real-time spectrum. This embodiment sets a sampling window to reduce the processing load of spectral data. The sampling window is not limited to 5 frames; it can also be 6, 8, or 10 frames, depending on the semiconductor process requirements.

[0027] To facilitate real-time monitoring, in one embodiment, the aforementioned principal component analysis model and reference spectrum are both formed based on pre-collected historical data, which can be pre-stored in the control system of the OES equipment or the local control system of the semiconductor process. Specifically, N semiconductor processes are performed in the cleanroom, and raw spectral data is collected using OES sensors. This raw spectral data is collected at M time points during each process, with each time point covering the full-band spectrum of P wavelengths. The raw spectral data is constructed into a two-dimensional data matrix, where the dimension of the two-dimensional data matrix is ​​(N × M) rows × P columns. The two-dimensional data matrix is ​​standardized column-wise. This standardization process may include calculating the mean μ and standard deviation σ of all elements in each column of the two-dimensional data matrix, and subtracting the mean μ from each raw value in that column and dividing by the standard deviation σ to obtain a standard matrix. The wavelength variables in the standard matrix are then centered and scaled to obtain the reference spectral matrix. Based on the reference spectral matrix, its covariance matrix is ​​calculated, and the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvectors are defined as the principal component directions. The original spectral data (i.e., historical data) is projected along the principal component directions to obtain the dimensionality-reduced score matrix, and its T² and SPE statistics are calculated. The formula for T² is as follows: Where x is the spectral vector in the original spectral data, P is the principal component loading matrix, and Λ is the eigenvalue diagonal matrix of the principal components. The SPE calculation formula is as follows: Where x is the spectral vector in the spectral data, This is the spectral vector reconstructed through the principal component directions. The aforementioned baseline spectral matrix, along with its corresponding T² and SPE statistics, are used as key parameters of the principal component analysis model, available for use by online monitoring methods. The T² and SPE statistics calculated from the original spectral data, and then determined according to the semiconductor process formulation, can be used to determine the critical values ​​of the T² and SPE statistics, which serve as the preset control limits for the process, enabling early warning analysis of metal contamination.

[0028] Specifically, step 202 above may include projecting the real-time acquired spectral data (i.e., the spectral data corresponding to the current sampling window) onto the principal component direction determined by the aforementioned benchmark spectral matrix to obtain a dimension-reduced score matrix. The T² and SPE statistics are then calculated according to the aforementioned formulas, and the calculated T² and SPE statistics are compared with preset control limits. If the real-time calculated SPE statistics do not exceed the preset control limits, the current spectral data can be considered normal data, and real-time spectral acquisition can continue for PCA analysis. If the real-time calculated SPE statistics exceed the preset control limits, the current spectral data is considered a contamination spectrum, and the process proceeds to the metal contamination analysis step for further analysis to determine whether it is indeed metal contamination.

[0029] Furthermore, in step 202 above, when it is determined that all SPE statistics exceed the preset control limits within a certain time period, the average value of the corresponding spectral data within that time period is taken as the contamination spectrum, and the process proceeds to the metal contamination analysis step. This involves calculating the similarity between the real-time acquired spectral data and multiple original spectra for the corresponding process time period, comparing them with the aforementioned benchmark spectral library to find the closest benchmark spectrum for that time period, and performing numerical averaging and window smoothing on the spectral data acquired within a certain time period (e.g., spectral data corresponding to 5 sampling windows), using the average value as the contamination spectrum to reduce noise interference.

[0030] To better monitor and analyze in real time, in one embodiment, the above-mentioned raw spectral data is aligned and divided based on the time dimension to establish a benchmark spectral library, specifically including the following steps: S402, Obtain the above-mentioned raw spectral data; S404, based on a preset time stage criterion, selects a subset of spectra that meet the target time stage from the original spectral data; specifically, the preset time stage criterion can be set according to the semiconductor process, such as plasma etching process, which can be divided into the initial, middle and late stages of main etching. For example, the spectral data corresponding to the initial stage of main etching in multiple processes can be selected to form an initial spectral subset. S406, calculate the average spectrum and variance spectrum of the spectral subset and perform standardization processing to generate a reference spectrum representing the spectral characteristics of the target time period, that is, to form a reference spectral library divided by the time dimension. For example, the average spectrum and variance spectrum of the above initial spectral data are calculated and standardized. The standardization processing includes, but is not limited to, vector normalization, standard normal variable transformation, or maximum value normalization to form the initial reference spectrum.

[0031] Based on this benchmark spectral library, in real-time online monitoring of metal contamination, the closest spectrum can be found at the process time stage where the contamination spectrum is generated by the semiconductor process. This combines spectral analysis with precise process timing, improving the traditional monitoring system from offline analysis of metal contamination to online detection and predictive monitoring of "when metal contamination occurred". This significantly improves the yield control capability, production efficiency and automation level of semiconductor manufacturing.

[0032] Specifically, in step 301 above, based on the process time stage of the contamination spectrum, the closest reference spectrum is obtained. Specifically, according to the current process time stage of the contamination spectrum, the closest reference spectrum is searched from the aforementioned reference spectrum library. If step 202 detects that the SPE statistic of the real-time spectrum during the main etching period exceeds a preset control limit, then step 301 obtains the reference spectrum of the main etching period as the background spectrum, and performs difference spectral analysis with the real-time spectrum. This embodiment compares the real-time spectrum with a reference spectrum library aligned by time stage, which can immediately determine which process time stage's reference spectrum is most similar to the currently acquired contamination spectrum. When metal contamination is detected, the process time stage can be output synchronously, i.e., the alarm signal can be displayed as "metal contamination occurs in the middle of the main etching." Thus, when contamination has just occurred and has not yet caused irreversible damage to the entire batch of wafers, an alarm signal can be immediately identified and triggered, achieving early monitoring, diagnosis, and intervention.

[0033] To better monitor the process chamber environment online, in one embodiment, the process equipment and metal materials used in the semiconductor process are considered as contamination sources, and a preset metal characteristic spectrum database is used for detection and tracking. Specifically, the establishment of the preset metal characteristic spectrum database may include the following steps S501-S503: Step S501: Based on the semiconductor process and the raw spectral data collected above, determine the characteristic spectral lines of at least two metals to be tested; for example, determine the potential metal contamination elements that need to be monitored based on the materials used in the semiconductor process (such as Al, Ti, W, Cu, Fe, Ni, etc.); if the main components are aluminum alloy electrodes and stainless steel cavity walls, then the monitored metals are Al and Cu; specifically, select one or more atomic or ion characteristic emission spectral lines with high sensitivity and low interference for each metal to be tested (such as Al and Cu) as the monitoring targets. Step S502: Obtain the reference spectrum corresponding to different process time stages in the semiconductor process. The reference spectrum is obtained from the reference spectrum library obtained in step S403 above. Step S503: Calculate the reference index CL corresponding to the characteristic spectral line of each metal to be tested in each reference spectrum. m The process time stage and the benchmark index CL for each metal being tested will be used. m Establish corresponding relationships to form a metal characteristic spectrum database.

[0034] Specifically, the reference index CL corresponds to the characteristic spectral lines of each metal being tested. m The calculation process is as follows: In the aforementioned reference spectral library, the reference spectrum corresponding to each process time point t is searched. For example, if the original spectral data contains 100 process data points (i.e., N is 100), and there are 100 frames corresponding to time point t, a background index Cl can be calculated for each spectrum in the 100 frames. a Value, for 100 Cl a The average value is calculated as Mean, and the standard deviation is calculated as σ. Mean and σ are used as the benchmark index CL. m Storage. Wherein, Cl a The value is the ratio of the intensity of the characteristic emission line of the metal element to be tested to the intensity of the characteristic emission line of the inert gas selected according to the semiconductor process.

[0035] To eliminate the impact of overall spectral intensity variations caused by fluctuations in process power and pressure, a stable background gas spectral line is first identified. The specific selection depends on the semiconductor process, and inert gases are typically used as background gases. These gases do not participate in reactions, have stable chemical properties, and their spectra are only affected by the plasma physical state (such as power and pressure) and are not interfered with by metal contamination. Nitrogen is generally the inert gas selected based on the semiconductor process. To improve monitoring accuracy, in this embodiment, during the acquisition of raw spectral data, two different spectral lines of an inert gas are selected as dual references. That is, when establishing the reference spectral library, the following steps are also included: Step S403: Determine two different spectral lines of the inert gas according to the semiconductor process, such as two wavelengths of nitrogen as reference characteristic emission spectral lines, such as the first characteristic peak of the inert gas being 750.4 nm and the second characteristic peak of the inert gas being 575.5 nm, and use these as background gases for monitoring. Step S405: Based on the raw spectral data obtained in step S402, calculate the spectral intensity ratio corresponding to the two reference characteristic emission lines within each process time stage, denoted as the reference intensity ratio of the first characteristic peak and the second characteristic peak of the inert gas. Each process time stage can be consistent with the preset time stage in step S406. Specifically, for the reference spectrum corresponding to each process time stage, calculate the reference intensity ratio of the first characteristic peak and the second characteristic peak of the inert gas to obtain the alignment data between the reference intensity ratio and the time node. This data can be stored in the reference spectrum library for easy retrieval and searching during real-time online monitoring, simplifying the processing time of on-site process data. In the dual-reference system, the reference index CL corresponding to the characteristic spectral line of each metal to be tested... m The benchmark index CL m The spectral intensity of the first characteristic peak corresponding to the inert gas is obtained by calculation.

[0036] Another embodiment of this application provides a method for online monitoring of metal pollution, see [link to application]. Figure 2As shown, the metal contamination analysis in step S203 above may include: Step 301: Based on the process time stage of the contamination spectrum, find the corresponding reference spectrum from the aforementioned reference spectral library, calculate the difference spectrum between the contamination spectrum and the reference spectrum, and find the reference intensity ratio between the first characteristic peak and the second characteristic peak of the inert gas; and find the reference index CL of the corresponding metal to be tested from the aforementioned metal characteristic spectral database. m The reference spectral library can be established by steps S402-S406, and the metal characteristic spectral database can be obtained by steps S501-S503. Step 302: Based on the difference spectrum from step 301, calculate the pollution index CL of the metal to be tested, specifically including: Step 3021: Based on the difference spectrum, calculate the ratio of the spectral intensity of the first characteristic peak of the inert gas to the second characteristic peak of the inert gas, and use it as the current ratio. Step 3022: Calculate the difference between the current ratio and the reference intensity ratio found in step 301, and compare it with a preset range. If the difference is within the preset range, calculate the ratio of the characteristic wavelength spectral intensity of the pollutant to be tested in the difference spectrum to the spectral intensity of the first characteristic peak of the inert gas in the original spectral data, and use this ratio as the pollution index CL of the metal to be tested. Specifically, the spectral intensity of the first characteristic peak of the inert gas is selected for calculation based on the reference index CL calculated from the reference spectral library. m The selected characteristic peaks are chosen to ensure consistency between the two comparison objects. Understandably, if the reference index CL is calculated in the reference spectral library... m Since the selected characteristic peak is the second characteristic peak of inert gas, when calculating the pollution index CL in this step, it is also necessary to select the spectral intensity of the second characteristic peak of inert gas in the original spectral data for calculation. If the difference between the two exceeds a preset range, the reference spectrum is calibrated based on the difference spectrum to generate a background spectrum; based on the difference spectrum, the background spectrum is subtracted from the characteristic wavelength spectrum of the metal to be tested and the spectrum of the first characteristic peak of the inert gas, respectively, and the respective de-background peak areas are calculated and denoted as I'1 and I'2; the ratio of the de-background peak area corresponding to the metal to be tested to the de-background peak area corresponding to the first characteristic peak of the inert gas is calculated as the pollution index CL of the metal to be tested; Step 303: Compare the pollution index CL of the metal to be tested with the corresponding benchmark index CL of the metal to be tested. m If the comparison result meets a preset condition, it is determined that metal contamination exists in the process chamber, and an alarm signal is output. The preset condition is that CL is greater than CL within a preset time period. m The preset time period can be 5-10 sampling windows, and each sampling window can include 5 frames of data.

[0037] In summary, the technical solution provided in this application has the following advantages: By first performing principal component analysis on the real-time acquired spectral data, and then triggering and entering the metal contamination analysis step only when the calculated SPE statistic exceeds the preset control limit, it is not necessary to calculate the metal contamination analysis index for all real-time spectral data. This allows for real-time, online monitoring of the metal contamination level within the process chamber, avoiding the need to perform metal contamination level analysis on all real-time spectra, thus improving analysis efficiency and accuracy. In addition, the metal contamination analysis step uses differential spectral analysis, and the reference spectrum used for differential spectral analysis is selected to be closest to the process time stage of the contamination spectrum. Through the time alignment mechanism, only the original spectral data (i.e., historical data) corresponding to the process time stage in which the contamination spectrum occurred is analyzed, eliminating the need for global processing of the original spectral data, thus improving analysis efficiency and accuracy. This enables early warning and rapid response to metal contamination, fundamentally improving the process control level and product yield in semiconductor manufacturing.

[0038] Another exemplary embodiment of this application provides an online monitoring system for metal pollution. For example... Figure 3 As shown, in this embodiment, the online monitoring system for metal pollution includes a connected spectral acquisition module 100, a first processing module 200, and a second processing module 300, wherein... The spectral acquisition module 100 is used to acquire spectral data within the process chamber in real time. The first processing module 200 is used to calculate the SPE statistic of the spectral data according to the principal component analysis model, and when it is determined that the SPE statistic exceeds the preset control limit, output a trigger command to enter the metal pollution analysis step. The second processing module 300 is used to execute the metal contamination analysis step when the trigger command is received. The metal contamination analysis step includes: obtaining the closest reference spectrum based on the process time stage of the contamination spectrum, and calculating the difference spectrum between the contamination spectrum and the reference spectrum, wherein the reference spectrum is obtained from the raw spectral data collected by running the semiconductor process multiple times in the clean process chamber. Based on the original spectral data and the difference spectrum, the pollution index CL of the metal to be tested is calculated; The contamination index of the metal to be tested is matched with a preset metal characteristic spectrum database. When the matching result meets the preset conditions, it is determined that there is metal contamination in the process chamber, and an alarm signal is output.

[0039] In this embodiment of the online metal contamination monitoring system, the first processing module 200 acts as a trigger for the second processing module 300. The second processing module is triggered only when the SPE statistic exceeds the preset control limit. This eliminates the need for global processing of the original spectral data, improving analysis efficiency and accuracy. It enables early warning and rapid response to metal contamination, fundamentally improving the process control level and product yield in semiconductor manufacturing.

[0040] In one embodiment, the online monitoring system for metal pollution further includes: A data storage module 400 is connected to the first processing module and the second processing module. The data storage module is used to store the principal component analysis model, the preset metal characteristic spectrum database, and the reference spectrum.

[0041] Embodiments of this application also propose an electronic device, including a processor and a memory; the memory is used to store a computer program executable by the processor; the processor is used to execute the computer program in the memory to implement the online monitoring method for metal pollution described in any of the above embodiments.

[0042] Embodiments of this application also propose a computer-readable storage medium that, when an executable computer program in the storage medium is executed by a processor, enables the implementation of the online metal pollution monitoring method described in any of the above embodiments.

[0043] Embodiments of this application also propose a computer program product, including a computer program that, when executed by a processor, implements a driving method for the universal serial bus interface of any of the above embodiments.

[0044] Regarding the apparatus in the above embodiments, the specific manner in which the processor performs the operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0045] Figure 4 This is a block diagram illustrating an electronic device according to an exemplary embodiment. For example, the electronic device (i.e., device 900) may be provided as a server. (Refer to...) Figure 4 The device 900 includes a processing component 922, which further includes one or more processors, and memory resources represented by a memory 932 for storing instructions, such as application programs, that can be executed by the processing component 922. The application programs stored in the memory 932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 922 is configured to execute instructions to perform the aforementioned method for online monitoring of metal contamination.

[0046] Device 900 may also include a power supply component 926 configured to perform power management of device 900, a wired or wireless network interface 950 configured to connect device 900 to a network, and an input / output (I / O) interface 958. Device 900 can operate on an operating system stored in memory 932, such as Windows Server™, MacOS X™, Unix™, Linux™, FreeBSD™, or similar.

[0047] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 932 including instructions, which can be executed by a processing component 922 of the device 900 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0048] In this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The term "multiple" refers to two or more unless otherwise expressly defined.

[0049] The above description of the embodiments is intended to enable those skilled in the art to understand and apply this application. It will be apparent to those skilled in the art that various modifications can be easily made to these embodiments, and the general principles described herein can be applied to other embodiments without creative effort. Therefore, this application is not limited to the embodiments described herein, and any improvements and modifications made by those skilled in the art based on the disclosure of this application without departing from the scope and spirit of this application are within the scope of this application.

Claims

1. A method for on-line monitoring of metal contamination, characterized by, The method comprises the following steps: acquiring spectral data in a process cavity in a semiconductor process; calculating an SPE statistic of the spectral data according to a principal component analysis model, and when the SPE statistic exceeds a preset control limit, regarding the spectral data as a pollution spectrum and entering a metal pollution analysis step, the metal pollution analysis step comprising: acquiring a closest reference spectrum based on a process time stage of the pollution spectrum, calculating a difference spectrum between the pollution spectrum and the reference spectrum, wherein the reference spectrum is obtained from original spectral data collected by running the semiconductor process multiple times in a clean process cavity; calculating a pollution index of a to-be-tested metal based on the original spectral data and the difference spectrum; matching the pollution index of the to-be-tested metal with a preset metal characteristic spectrum database, and when a matching result meets a preset condition, determining that there is metal pollution in the process cavity and outputting an alarm signal.

2. The metal contamination on-line monitoring method according to claim 1, wherein The step of acquiring the closest reference spectrum based on the process time stage of the pollution spectrum comprises: acquiring the original spectral data; screening a spectrum subset meeting a target time stage from the original spectral data based on a preset time stage criterion; calculating an average spectrum and a variance spectrum of the spectrum subset, and performing standardization processing to generate a reference spectrum representing spectral characteristics of the target time stage as the closest reference spectrum.

3. The metal contamination on-line monitoring method according to claim 2, wherein The step of establishing the preset metal characteristic spectrum database comprises: determining characteristic spectral lines of at least two to-be-tested metals based on the semiconductor process and the original spectral data; acquiring reference spectra corresponding to different process time stages in the semiconductor process; calculating a reference index CL corresponding to the characteristic spectral line of each of the metals to be measured in each of the reference spectra m ; The process time phase and the reference index CL of each metal to be tested m The corresponding relationship is established to form a metal characteristic spectrum database; wherein the step of matching the pollution index of the to-be-tested metal with the preset metal characteristic spectrum database when the matching result meets the preset condition to determine that there is metal pollution in the process cavity comprises: obtaining a reference index CL of the metal to be measured from the preset metal characteristic spectrum database based on the process time stage of the pollution spectrum m ; The contamination index CL of the metal to be tested is compared with the reference index CL corresponding to the metal to be tested m When the comparison result meets the preset condition, it is determined that there is metal contamination in the process cavity, and the preset condition is that CL is greater than CL within a preset time period m .

4. The metal contamination on-line monitoring method according to claim 1, wherein The step of calculating the pollution index of the to-be-tested metal based on the original spectral data and the difference spectrum comprises: acquiring a reference intensity ratio of an inert gas first characteristic peak and an inert gas second characteristic peak, wherein the inert gas first characteristic peak and the inert gas second characteristic peak are selected according to the semiconductor process, and the reference intensity ratio is a spectral intensity ratio of the inert gas first characteristic peak and the inert gas second characteristic peak in the original spectral data collected in a clean process cavity; calculating a current ratio of spectral intensity ratios of the inert gas first characteristic peak and the inert gas second characteristic peak in the difference spectrum as a current ratio; when the current ratio and the reference intensity ratio are compared and a difference therebetween is within a preset range, calculating a spectral intensity ratio of a characteristic wavelength of a to-be-tested pollutant in the difference spectrum to spectral intensity of the inert gas first characteristic peak in the original spectral data as a pollution index of the to-be-tested metal.

5. The metal contamination on-line monitoring method according to claim 4, wherein The method further comprises: when the current ratio and the reference intensity ratio are compared and a difference therebetween exceeds the preset range, calibrating the reference spectrum based on the difference spectrum to generate a background spectrum; Subtracting the background spectrum from the characteristic wavelength spectrum of the metal to be detected and the spectral line of the first characteristic peak of the inert gas respectively, and calculating the background-subtracted peak area of each, denoted as I'1 and I'2; Calculating the ratio of the background-subtracted peak area of the metal to be detected to the background-subtracted peak area of the first characteristic peak of the inert gas as the contamination index of the metal to be detected.

6. The metal contamination on-line monitoring method according to claim 1, wherein The step of establishing the principal component analysis model comprises: performing N semiconductor processes in a clean process chamber, collecting original spectrum data using an OES sensor, the original spectrum data being collected at M time points during each process, and each time point collecting a full-band spectrum covering P wavelength points; constructing the original spectrum data into a two-dimensional data matrix, wherein the dimensions of the two-dimensional data matrix are (N × M) rows × P columns; performing standardization processing on the two-dimensional data matrix by column to obtain a reference spectrum matrix; based on the reference spectrum matrix, calculating a covariance matrix thereof, and solving the eigenvalues and eigenvectors of the covariance matrix, and defining the eigenvectors as principal component directions; The step of calculating the SPE statistic of the spectrum data according to the principal component analysis model comprises: Projecting the spectrum data to the principal component directions to calculate the SPE statistic.

7. The metal contamination on-line monitoring method of claim 1, wherein, When it is judged that the SPE statistic exceeds the preset control limit, the spectrum data is regarded as a pollution spectrum and enters the metal contamination analysis step, and the method further comprises: when it is judged that each of the SPE statistics in a certain time period exceeds the preset control limit, the average value of the corresponding spectrum data in the certain time period is regarded as a pollution spectrum and enters the metal contamination analysis step.

8. A metal contamination on-line monitoring system characterized by, The method comprises the steps of: The spectrum acquisition module is configured to acquire spectrum data in the process chamber in real time; The first processing module is configured to calculate the SPE statistic of the spectrum data according to the principal component analysis model, and output a trigger instruction for entering the metal contamination analysis step when it is judged that the SPE statistic exceeds the preset control limit; The second processing module is configured to execute the metal contamination analysis step when the trigger instruction is received, and the metal contamination analysis step comprises: based on the process time stage of the pollution spectrum, acquiring the closest reference spectrum, and calculating the difference spectrum between the pollution spectrum and the reference spectrum, wherein the reference spectrum is obtained from the original spectrum data collected by running multiple semiconductor processes in the clean process chamber; Based on the original spectrum data and the difference spectrum, the contamination index CL of the metal to be detected is calculated; The contamination index of the metal to be detected is matched with the preset metal characteristic spectrum database, and when the matching result meets the preset condition, it is determined that there is metal contamination in the process chamber, and an alarm signal is output.

9. The metal contamination on-line monitoring system of claim 8, wherein the metal contamination on-line monitoring system further comprises a second light source and a second light detector. The data storage module is connected to the first processing module and the second processing module, and is configured to store the principal component analysis model, the preset metal characteristic spectrum database, and the reference spectrum.

10. An electronic device, comprising: A computer program product comprising a memory for storing a computer program executable by a processor; and the processor for executing the computer program in the memory to implement the method of any one of claims 1-7.

11. A computer readable storage medium having stored thereon a computer program, characterized in that A computer program product comprising a memory for storing a computer program executable by a processor; and the processor for executing the computer program in the memory to implement the method of any one of claims 1-7.

12. A computer program product comprising a computer program, characterized in that, A computer program product comprising a memory for storing a computer program executable by a processor; and the processor for executing the computer program in the memory to implement the method of any one of claims 1-7.

Images