Methods, media, and electronic devices for detecting gas samples within semiconductor process chambers

Abstract

The application relates to a method, medium and electronic device for detecting a gas sample in a semiconductor process chamber, the method comprising: collecting a gas sample in a semiconductor process chamber as a mixed sample to be analyzed; obtaining a background spectrum vector B and a measured spectrum vector M; constructing an interference factor matrix F based on a preset standard spectrum library matrix S, wherein the preset standard spectrum library matrix S comprises standard spectrum vectors S of multiple components k ; constructing a weighted ridge regression optimization model based on the vector M, the vector B, the standard spectrum library matrix S and the interference factor matrix F; solving the model to obtain quantitative detection results of partial pressures of the components; and using the quantitative detection results to analyze real-time changes of gas components in the semiconductor process chamber. The technical scheme can effectively suppress interference and extract weak signals in a complex gas environment, and according to the weighted ridge regression optimization model, real-time changes of gas components in a semiconductor device chamber can be accurately monitored.

Description

Technical Field

[0001] This application relates to the field of semiconductor technology, and in particular to a method, medium, and electronic device for detecting gas samples in a semiconductor process chamber. Background Technology

[0002] In the semiconductor field, chip manufacturing processes require extremely high levels of cleanliness in the process environment. Mass spectrometry plays an irreplaceable core monitoring role. Key process steps such as etching, thin film deposition, and ion implantation are highly susceptible to the influence of gas composition and trace impurities within the process chamber. Even partial pressures as low as parts per billion (ppb) can lead to short circuits, leakage, or structural defects in the chip circuitry, severely restricting chip performance, yield, and reliability.

[0003] A mass analyzer is an analytical instrument that separates and detects substances based on the principle of charged particle deflection in an electromagnetic field, taking into account the mass differences of atoms, molecules, or molecular fragments. The core data carrier of a mass analyzer is a mass spectrum, plotted with mass-to-charge ratio (m / z) on the horizontal axis and relative ion current intensity on the vertical axis, recording characteristic peak information including molecular ion peaks and fragment ion peaks. With its high sensitivity and ability to simultaneously detect and analyze multiple components, the mass analyzer has become a key device in the semiconductor industry for online gas component analysis, vacuum system leak detection, and process control. The accuracy and stability of its quantitative analysis directly affect chip quality and production costs.

[0004] In existing semiconductor process chamber gas component detection schemes, the quantitative analysis methods of mass analyzers mainly include direct quantification and indirect quantification. Direct quantification is only suitable for analyzing simple, interference-free gas mixtures. Indirect quantification is currently the mainstream method in the industry. However, due to the limitations of its mathematical model, when isotopes are present in the gas mixture—that is, when two or more components in the gas mixture have a main peak contribution at the same mass-to-charge ratio—their corresponding standard spectra are mathematically linearly correlated. The model cannot obtain a stable and unique solution, thus failing to yield effective quantitative detection results.

[0005] In some existing improved methods for detecting gas components in semiconductor process chambers, technicians manually introduce fragment peaks to assist in model solving, thereby avoiding the inability to obtain effective quantitative detection results due to the presence of isotopes in the mixed gas. However, this indirect quantitative method, which manually introduces fragment peaks to assist in the solution for quantitative analysis, relies on human experience and intervention, cannot be automated, and may introduce new collinearity problems in complex mixtures, resulting in unstable solution results. Furthermore, this method is dominated by strong signals, making it difficult to accurately extract and quantify trace weak signals over a wide dynamic range, resulting in insufficient reliability of quantitative analysis and thus poor reliability of gas component detection results.

[0006] In summary, existing methods for detecting gas components in semiconductor process chambers suffer from low stability and accuracy when detecting complex mixed gases. Summary of the Invention

[0007] To address the technical problem of low stability and accuracy in gas component detection results when performing gas component detection on mixed gases within semiconductor chambers with complex compositions, this application provides a method, medium, and electronic device for detecting gas samples within semiconductor process chambers.

[0008] The technical solution of this application acquires the background mass spectrum of the vacuum sample chamber and the mass spectrum of the gas sample in the semiconductor process chamber of the gas component analysis system under the same instrument conditions and converts them into vectors, ensuring the consistency of the data benchmark for subsequent quantitative analysis. By constructing a standard spectral library matrix S containing isotopic interference components and an interference factor matrix F, the peak interference relationship between components is quantified. By establishing a weighted ridge regression optimization model with the partial pressure vector C as the solution objective and non-negative constraints, the partial pressure solution is transformed into a mathematically stable problem. The non-negative constraints ensure that the final calculated partial pressure value of each gas component is physically reasonable and meaningful. This effectively suppresses interference and extracts weak signals in complex gas environments with isotopic interference and a wide dynamic range, improving the stability and accuracy of gas component detection results for mixed gases in semiconductor chambers with complex compositions.

[0009] To achieve the above objectives, a first aspect of the present invention provides a method for detecting gas samples in a semiconductor process chamber, applicable to a gas component analysis system, wherein the gas component analysis system includes a vacuum sample chamber, a mass analyzer, and a processing unit, and the method includes:

[0010] A gas sample is collected from the semiconductor process chamber, and the gas sample is used as the mixed sample to be analyzed.

[0011] The background mass spectrum of the vacuum sample chamber and the original mass spectrum of the mixed sample to be analyzed in the vacuum sample chamber are obtained using the mass analyzer, wherein the background mass spectrum and the original mass spectrum are acquired under the same instrument conditions.

[0012] The background mass spectrum and the original mass spectrum are preprocessed to obtain the background spectrum vector B and the measured spectrum vector M.

[0013] Obtain a preset standard spectral library matrix S, and construct an interference factor matrix F based on the preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship between multiple components related to the mixed sample to be analyzed. The preset standard spectral library matrix S includes standard spectral vectors S of the multiple components. k ;

[0014] Based on the measured spectrum vector M, the background spectrum vector B, the preset standard spectrum library matrix S, and the interference factor matrix F, a weighted ridge regression optimization model is constructed with the partial pressure vector C as the solution objective and the non-negativity of the partial pressure vector C as the constraint condition. The partial pressure vector C is used to describe the partial pressure of each component in the mixed sample.

[0015] Solve the weighted ridge regression optimization model to obtain the partial pressure vector C that minimizes the objective function of the weighted ridge regression optimization model. Use the partial pressure vector C as the quantitative detection result of the partial pressure of each component in the mixed sample to be analyzed. The quantitative detection result is used to analyze the real-time changes of gas components in the semiconductor process chamber.

[0016] In this way, by acquiring the background mass spectrum of the vacuum sample chamber and the original mass spectrum of the mixed sample to be analyzed in the vacuum sample chamber under the same instrument conditions, and converting them into vectors, the uniformity of the data benchmark is ensured, providing an accurate foundation for subsequent analysis. By constructing a standard spectral library matrix S and an interference factor matrix F generated based on it, the peak interference relationship between specific components is quantified, providing a basis for subsequent model decoupling. By establishing a weighted ridge regression optimization model with the partial pressure vector C as the solution objective and with non-negative constraints, the partial pressure solution is transformed into a mathematical problem that can be stably calculated, where the non-negative constraints ensure that the solution conforms to physical reality. The synergistic effect of these features enables the technical solution of this application to effectively suppress interference and extract weak signals when detecting gas components in semiconductor process chambers, even in environments with isotopic interference and complex gases with a wide dynamic range, thereby improving the stability and accuracy of gas component detection results in semiconductor process chambers.

[0017] In one possible implementation of the first aspect, the construction of an interference factor matrix F based on the preset standard spectral library matrix S for characterizing the mass spectrometry signal interference relationship among the multiple components includes:

[0018] Obtain the measured spectrum vector M of each component in the preset standard spectrum library matrix S. k and standard spectral vector S k ;

[0019] Based on the measured spectral vector M of each component k and standard spectral vector S kThe scaling factor α and the disturbance vector f are obtained by solving the least squares method. k Make the relational expression Established;

[0020] For the interference vector f k After normalization;

[0021] The interference vector f, after normalization of each component according to the preset component order, is... k The diagonal elements of the interference factor matrix F are set to 1 and respectively, serving as off-diagonal column vectors, to construct the interference factor matrix F.

[0022] By obtaining the measured spectral vector M of each component k With standard spectral vector S k The interference vector f is obtained based on least squares fitting. k It can quantify the inter-peak interference generated by each component in actual measurements; by normalizing the interference vector f k By constructing an interference factor matrix F using off-diagonal column vectors and presetting the diagonal elements to 1, the mutual interference relationships between all components can be fully represented in matrix form, providing an accurate data foundation for subsequent interference subtraction based on this matrix.

[0023] In one possible implementation of the first aspect, the objective function of the weighted ridge regression optimization model is... The following relationship must be satisfied:

[0024]

[0025] Where W is the weight matrix, λ is the regularization parameter, Γ is the constraint matrix, and the partial pressure of each component in the partial pressure vector C is... 0, and the regularization parameter λ is determined by cross-validation.

[0026] Thus, by clearly defining the objective function expression of the weighted ridge regression optimization model, which adjusts the fitting weights of each channel through the weight matrix W, controls the model complexity through the regularization parameter λ, introduces prior constraints through the constraint matrix Γ, and ensures the physical rationality of the pressure divider vector through non-negativity constraints, this mathematical expression provides a clear optimization framework for solving the pressure divider problem.

[0027] In one possible implementation of the first aspect, the weight matrix W is a diagonal matrix, and the i-th diagonal element w of the weight matrix W... i It is calculated based on the signal strength or signal-to-noise ratio of the measured spectrum vector M and the background spectrum vector B on the corresponding i-th quality channel.

[0028] By restricting the weight matrix W to a diagonal matrix, and calculating its diagonal elements based on the signal strength or signal-to-noise ratio of the corresponding quality channel, the fitting weights can be dynamically allocated according to the signal-to-noise level of each channel, which helps to balance the influence of signals of different intensities during the optimization process.

[0029] In one possible implementation of the first aspect, the i-th diagonal element w of the weight matrix W i The calculation formula is:

[0030]

[0031] in, This is the noise variance estimate for the i-th quality channel.

[0032] By further clarifying that the calculation formula for the diagonal elements of the weight matrix is ​​the reciprocal of the noise variance estimate, higher weights can be automatically assigned to low signal-to-noise ratio channels based on the noise level of each quality channel, thereby enhancing the fitting attention to weak signal channels.

[0033] In one possible implementation of the first aspect, solving the weighted ridge regression optimization model to obtain the pressure vector C that minimizes the objective function of the weighted ridge regression optimization model includes:

[0034] The weighted ridge regression optimization model is solved using the iterative reweighted least squares method to obtain the pressure vector C that minimizes the objective function of the weighted ridge regression optimization model:

[0035] Initialize the voltage divider vector C and the weight matrix W;

[0036] Calculate the residual vector R based on the current pressure divider vector C, and update the weight matrix W according to the residual vector R;

[0037] The partial pressures of each component in the partial pressure vector C Under the constraint of 0, the weighted ridge regression optimization model is solved based on the updated weight matrix W to update the pressure vector C;

[0038] Determine whether the updated pressure vector C satisfies the preset convergence condition. If not, repeat the process of updating the weight matrix W and updating the pressure vector C until the pressure vector C satisfies the preset convergence condition.

[0039] The formula for calculating the residual vector R is as follows: .

[0040] By employing the iterative reweighted least squares method, the pressure divider vector C and the weight matrix W are alternately updated during the iteration process, enabling adaptive adjustment of the weights and pressure dividers. By setting convergence conditions to control the termination of the iteration, the optimization process can be guaranteed to converge stably to an effective solution.

[0041] In one possible implementation of the first aspect, the mixed sample to be analyzed includes a gas sample within a chamber during a semiconductor process, the type of which includes thin film, etching, and ion implantation.

[0042] In this way, the mixed sample to be analyzed is clearly defined to include gas samples corresponding to thin film, etching and ion implantation processes, which enables the detection of gas components in key process steps of semiconductor manufacturing and improves the applicability of the technical solution of this application in specific processes.

[0043] In one possible implementation of the first aspect, the real-time changes in the gas composition within the semiconductor process chamber are analyzed in the following manner:

[0044] The percentage of each component in the gas composition within the semiconductor process chamber is obtained based on the quantitative detection results.

[0045] This provides more intuitive data on gas composition distribution, making it easier to directly assess the relative content of each gas component and its potential impact on the process, thereby supporting more efficient qualitative assessment and decision-making regarding process status.

[0046] Secondly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein when the executable computer program in the storage medium is executed by a processor, it is capable of implementing a method for detecting gas samples in a semiconductor process chamber as described in the first aspect and any possible implementation thereof.

[0047] Thirdly, this application provides an electronic device, including a memory and a processor, wherein 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 a method for detecting gas samples in a semiconductor process chamber as described in the first aspect and any possible implementation of the first aspect.

[0048] Compared with the prior art, the beneficial effects of this application are as follows:

[0049] 1. By acquiring and preprocessing background and sample mass spectra under consistent mass analyzer parameter settings, standardized spectral vectors B and M are obtained. This solves the problem of detection errors caused by different instrument conditions and inconsistent quantitative standards caused by instrument state drift, effectively improving the reliability of quantitative analysis basic data.

[0050] 2. By constructing and utilizing the interference factor matrix F to quantify the spectral peak interference relationship between components, a systematic model of complex cross-interference, including isotopes, was established. This enabled the automatic and accurate subtraction of manually introduced fragment peak interference signals, fundamentally solving the problem of empirical operations such as manually introducing fragment peaks in traditional methods and improving the accuracy of process detection.

[0051] 3. By establishing and solving a weighted ridge regression optimization model with the partial pressure vector C as the objective, and combining it with a weight matrix set for the signal-to-noise ratio of different quality channels, the fitting weight and attention of weak signals of trace components are enhanced. This avoids the problem that weak gas components with low partial pressure values ​​in mixed gas samples are overwhelmed by the main gas components with high partial pressure values, resulting in the inability to detect weak gas components. As a result, it can effectively extract and accurately quantify ultra-trace impurities at the ppb (parts per billion) level, with lower detection limits, higher detection sensitivity, and higher detection accuracy.

[0052] 4. The solution process is completed automatically by the model, which can avoid interference from manual pre-judgment and operation, and has higher detection efficiency. It can detect the partial pressure of each component of the gas sample in the semiconductor device chamber, so as to accurately monitor the real-time changes of the gas composition in the semiconductor device chamber.

[0053] 5. By imposing non-negativity constraints on the voltage divider vector C, the physical rationality of the detection results is ensured, completely avoiding the physical meaninglessness of negative voltage divider values ​​that may be generated by traditional algorithms, thus improving the reliability of the detection results. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 According to some embodiments of this application, a scenario for detecting gas samples in a semiconductor process chamber is shown;

[0056] Figure 2 According to some embodiments of this application, a flowchart of a method for detecting gas samples in a semiconductor process chamber is shown;

[0057] Figure 3 According to some embodiments of this application, a structural block diagram of an electronic device is shown. Detailed Implementation

[0058] The illustrative embodiments of this application include, but are not limited to, a method, medium, and electronic device for detecting gas samples in a semiconductor process chamber.

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0060] As mentioned above, the existing technology has the technical problem of low stability and accuracy of the detection results when performing gas component detection on mixed gases in semiconductor chambers with complex compositions.

[0061] In view of this, this application provides a method for detecting gas samples in a semiconductor process chamber. Specifically, the technical solution of this application acquires the background mass spectrum of the vacuum sample chamber in a gas component analysis system and the mass spectrum of the gas sample in the semiconductor process chamber under the same instrument conditions, and converts them into vectors, ensuring the consistency of the data benchmark for subsequent quantitative analysis. By constructing a standard spectral library matrix S and an interference factor matrix F, the peak interference relationship between components is quantified. By establishing a weighted ridge regression optimization model with the partial pressure vector C as the solution objective and non-negative constraints, the partial pressure solution is transformed into a mathematically stable problem, where the non-negative constraints ensure that the final calculated partial pressure value of each gas component is physically reasonable and meaningful. Thus, it can effectively suppress interference and extract weak signals in complex gas environments with isotopic interference and wide dynamic range, improving the stability and accuracy of the detection results obtained for gas component detection in mixed gases with complex compositions in semiconductor chambers.

[0062] This application provides a method for detecting gas samples in a semiconductor process chamber, applicable to, for example... Figure 1 The gas component analysis system 20 shown includes a vacuum sample chamber 26, a mass analyzer 22, and a processing unit 27. (Reference) Figure 1 The gas component analysis system 20 also includes an ion source 21, a detector 23, a sampling interface 25, and a signal amplifier and data acquisition unit 24.

[0063] The gas component analysis system 20 collects a gas sample from the semiconductor process chamber 10 via the sampling interface 25 and introduces the gas sample as the mixed sample to be analyzed into the vacuum sample chamber 26. Subsequently, the gas sample enters the analysis area after pressure regulation and flow control via the sampling interface 25. In the vacuum sample chamber, the background mass spectrum of the vacuum sample chamber 26 and the raw mass spectrum of the mixed sample to be analyzed in the vacuum sample chamber 26 are acquired under the same instrument conditions. The processing unit 27 preprocesses the background mass spectrum and the raw mass spectrum to obtain the background spectral vector B and the measured spectral vector M, respectively. The processing unit 27 acquires a preset standard spectral library matrix S and constructs an interference factor matrix F based on the preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship between multiple components related to the mixed sample to be analyzed, wherein the standard spectral library matrix S includes standard spectral vectors Sk of multiple components. Subsequently, processing unit 27 constructs a weighted ridge regression optimization model based on the measured spectral vector M, the background spectral vector B, the preset standard spectral library matrix S, and the interference factor matrix F. The model uses the partial pressure vector C as the objective and the non-negativity of the partial pressure vector C as a constraint. Processing unit 27 solves the weighted ridge regression optimization model to obtain the partial pressure vector C that minimizes the objective function of the model. This partial pressure vector C is then used as the quantitative detection result of the partial pressures of each component in the mixed sample to be analyzed. Finally, the real-time changes in the gas composition within the semiconductor process chamber 10 are analyzed based on the quantitative detection results.

[0064] It should be noted that the architecture of the gas component analysis system 20 is not limited to... Figure 1 The system architecture shown can also include more devices in the gas component analysis system 20.

[0065] refer to Figure 2 The method for detecting gas components in a semiconductor process chamber 10 provided in this application includes the following steps S11 to S16:

[0066] Step S11: Collect a gas sample from the semiconductor process chamber 10 and use the gas sample as the mixed sample to be analyzed.

[0067] In some embodiments, the mixed sample to be analyzed includes a gas sample from a chamber during a semiconductor process, the types of which include, but are not limited to, thin film deposition, etching, and ion implantation. This allows the technical solution of this application to perform specialized testing of gas components in key process steps of semiconductor manufacturing, improving the applicability of the technical solution in specific processes.

[0068] The applicable types of thin film processes include PVD, CVD, ALD, PECVD, and MOCVD. The applicable types of etching processes include ALE, RIE, and ICP. The applicable types of ion implantation processes include PLAD.

[0069] The components of the mixed sample to be analyzed include, but are not limited to, silane (SiH4), carbon tetrafluoride (CF4), nitrogen trifluoride (NF3), tungsten hexafluoride (WF6), ammonia (NH3), oxygen (O2), nitrogen (N2), carbon dioxide (CO2), hydrogen (H2), carbon monoxide (CO), hydrogen fluoride (HF), and hydrogen chloride (HCl).

[0070] Step S12: Use mass analyzer 22 to obtain the background mass spectrum of vacuum sample chamber 26 and the original mass spectrum of the mixed sample to be analyzed in vacuum sample chamber 26. The background mass spectrum and the original mass spectrum are obtained under the same instrument conditions.

[0071] The same instrument conditions can refer to the same instrument parameters used when scanning with a quality analyzer. These instrument parameters include, but are not limited to, electron emission current, electron energy, scanning range and speed, and detector gain.

[0072] Step S13: Preprocess the background mass spectrum and the original mass spectrum to obtain the background spectrum vector B and the measured spectrum vector M.

[0073] In some embodiments, the background mass spectrum and the raw mass spectrum are preprocessed separately, including: aligning the mass axes of the background mass spectrum and the raw mass spectrum, smoothing the background noise, and extracting the signal intensity values ​​on a preset mass channel sequence. These are then arranged in mass channel order to form a background spectrum vector B and a measured spectrum vector M with consistent dimensions. The preset mass channel sequence covers the characteristic mass-to-charge ratios of all gas components in the mixed sample to be analyzed. This unified preprocessing ensures that vector B and vector M have the same dimensionality and physical meaning.

[0074] Step S14: Obtain a preset standard spectral library matrix S, and construct an interference factor matrix F based on the preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship between multiple components related to the mixed sample to be analyzed. The preset standard spectral library matrix S includes standard spectral vectors S of multiple components. k .

[0075] Each column in the standard spectral library matrix S corresponds to a mass spectrum of a pure component gas, that is, a standard mass spectrum of that gas produced under the same instrument conditions, containing all characteristic peaks and fragment peaks. The standard mass spectra in the standard spectral library matrix S are obtained by calibration of standard gases under the same instrument conditions, or by using data from authoritative commercial databases that have been validated in the industry (such as the NIST Mass Library).

[0076] Among them, the isotopic interference components can be, for example, CO and N2, or CH4 and O2, etc.

[0077] In some embodiments, constructing an interference factor matrix F based on a preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship between multiple components includes: obtaining the measured spectral vector M of each component in the preset standard spectral library matrix S. k and standard spectral vector S k Based on the measured spectral vector M of each component k and standard spectral vector S k The scaling factor α and the disturbance vector f are obtained by solving the least squares method. k Make the relational expression It holds true; for the interference vector f k After normalization, the normalized interference vector f corresponding to each component is processed according to the preset component order. k Each element is used as a non-diagonal column vector, and the diagonal elements of the preset interference factor matrix F are set to 1 to construct the interference factor matrix F.

[0078] In this way, by obtaining the measured spectral vector M of each component... k With standard spectral vector S k The interference vector f is obtained based on least squares fitting. k It can quantify the mutual interference between spectral peaks generated by each component in actual measurements; by normalizing the interference vector f k The interference factor matrix F is constructed by using off-diagonal column vectors and setting the diagonal elements to 1. In other words, setting the diagonal elements of the interference factor matrix F to 1 represents the main signal of the quality channel itself; the off-diagonal elements F ij (i≠j) represents the contribution of a unit partial pressure component at component j to the signal interference in mass channel i. This matrix form comprehensively characterizes the mutual interference relationships among all components, providing a clear data foundation for subsequent interference subtraction based on the interference factor matrix F.

[0079] Step S15: Based on the measured spectrum vector M, the background spectrum vector B, the preset standard spectrum library matrix S, and the interference factor matrix F, construct a weighted ridge regression optimization model with the partial pressure vector C as the solution objective and the non-negativity of the partial pressure vector C as the constraint condition. The partial pressure vector C is used to describe the partial pressure of each component in the mixed sample.

[0080] In some embodiments, the objective function of the weighted ridge regression optimization model The following relationship must be satisfied:

[0081]

[0082] Where W is the weight matrix, λ is the regularization parameter, Γ is the constraint matrix, and the partial pressures of each component in the partial pressure vector C are... 0, and the regularization parameter λ are determined by cross-validation.

[0083] In some embodiments, the cross-validation method specifically includes: dividing historical standard gas sample data into a training set and a validation set according to a preset ratio, such as 7:3 or 8:2; training a model on the training set based on different λ values; evaluating the mean square error between the model's predicted partial pressure and the actual partial pressure on the validation set; and selecting the λ value that minimizes the mean square error as the final regularization parameter. This provides a basis for parameter selection to determine a stable model with strong generalization ability.

[0084] Thus, by clearly defining the objective function expression of the weighted ridge regression optimization model, which adjusts the fitting weights of each channel through the weight matrix W, controls the model complexity through the regularization parameter λ, introduces prior constraints through the constraint matrix Γ, and ensures the physical rationality of the pressure divider vector through non-negativity constraints, this mathematical expression provides a clear optimization framework for solving the pressure divider problem.

[0085] In some embodiments, the weight matrix W is a diagonal matrix, and the i-th diagonal element w of the weight matrix W i It is calculated based on the signal strength or signal-to-noise ratio of the measured spectrum vector M and the background spectrum vector B on the corresponding i-th quality channel.

[0086] By restricting the weight matrix W to a diagonal matrix, and calculating its diagonal elements based on the signal strength or signal-to-noise ratio of the corresponding quality channel, the fitting weights can be dynamically allocated according to the signal-to-noise level of each channel, which helps to balance the influence of signals of different intensities during the optimization process.

[0087] In some embodiments, the i-th diagonal element w of the weight matrix W i The calculation formula is:

[0088]

[0089] in, This is the noise variance estimate for the i-th quality channel.

[0090] By further clarifying that the calculation formula for the diagonal elements of the weight matrix is ​​the reciprocal of the noise variance estimate, higher weights can be automatically assigned to low signal-to-noise ratio channels based on the noise level of each quality channel, thereby enhancing the fitting attention to weak signal channels.

[0091] Step S16: Solve the weighted ridge regression optimization model to obtain the partial pressure vector C that minimizes the objective function of the weighted ridge regression optimization model. Use the partial pressure vector C as the quantitative detection result of the partial pressure of each component in the mixed sample to be analyzed. The quantitative detection result is used to analyze the real-time changes of gas components in the semiconductor process chamber.

[0092] In some embodiments, solving the weighted ridge regression optimization model to obtain the partial pressure vector C that minimizes the objective function of the weighted ridge regression optimization model includes: solving the weighted ridge regression optimization model using an iterative reweighted least squares method to obtain the partial pressure vector C that minimizes the objective function of the weighted ridge regression optimization model: initializing the partial pressure vector C and the weight matrix W; calculating the residual vector R based on the current partial pressure vector C, and updating the weight matrix W according to the residual vector R; and determining the partial pressure of each component in the partial pressure vector C. Under the constraint of 0, the weighted ridge regression optimization model is solved based on the updated weight matrix W to update the partial pressure vector C; it is then determined whether the updated partial pressure vector C satisfies the preset convergence condition. If not, the process of updating the weight matrix W and updating the partial pressure vector C is repeated until the partial pressure vector C satisfies the preset convergence condition; the formula for calculating the residual vector R is: Thus, by employing the iterative reweighted least squares method, the pressure divider vector C and the weight matrix W are alternately updated during the iteration process, enabling adaptive adjustment of the weights and pressure dividers. By setting convergence conditions to control the termination of the iteration, the optimization process can be guaranteed to converge stably to an effective solution.

[0093] In some embodiments, the convergence criteria include: the relative rate of change of the Euclidean norm of the pressure divider vector C is less than a first threshold, and the change in the objective function value of the weighted ridge regression optimization model is less than a second threshold. By setting explicit dual convergence criteria, it can be ensured that the optimization results reach a stable state in both parameters and the objective function.

[0094] In some embodiments, the real-time changes in the gas within the semiconductor process chamber 10 are analyzed by determining the percentage of each component in the gas within the semiconductor process chamber 10 based on quantitative detection results. This provides more intuitive data on the gas composition distribution, facilitating direct assessment of the relative content of each gas component and its potential impact on the process, thereby supporting more efficient qualitative assessment and decision-making regarding the process status.

[0095] In some embodiments, the method for detecting gas within the semiconductor process chamber 10 provided in this application further includes generating a gas composition detection report based on the gas detection results of the semiconductor process chamber 10. This gas composition detection report facilitates subsequent process status assessment and decision-making by engineers.

[0096] In some embodiments, the gas component detection report may be a table that records the detection time, the partial pressure of each gas component in the gas sample, and the percentage of each component. In some embodiments, the table may also include two charts: one chart visually displays the change of the partial pressure of each component in the gas sample over time, and the other chart displays the change of the percentage of each component in the gas sample over time, for the convenience of engineers.

[0097] In this way, by acquiring the background mass spectrum of the vacuum sample chamber 26 and the original mass spectrum of the mixed sample to be analyzed in the vacuum sample chamber 26 under the same instrument conditions, and converting them into vectors, the uniformity of the data benchmark is ensured, providing an accurate basis for subsequent analysis. By constructing a standard spectral library matrix S containing isotopic interference components and an interference factor matrix F generated based on it, the peak interference relationship between specific components is quantified, providing a basis for subsequent model decoupling. By establishing a weighted ridge regression optimization model with the partial pressure vector C as the solution objective and non-negative constraints, the partial pressure solution is transformed into a mathematical problem that can be stably calculated. The non-negative constraints ensure that the solution conforms to physical reality, and the weighting mechanism balances the influence of signals of different intensities. The synergistic effect of these features enables the technical solution of this application to effectively suppress interference and extract weak signals when detecting gas components in the semiconductor process chamber 10, even in the presence of isotopic interference and complex gases with a wide dynamic range in the detection environment, thereby improving the stability and accuracy of the gas component detection results in the semiconductor process chamber 10.

[0098] In a specific embodiment, it is assumed that the mixed sample to be analyzed contains a large amount of argon (Ar) and nitrogen (N2), as well as trace amounts of arsine (AsH3) and phosphine (PH3). The mixed sample may contain carbon monoxide (CO) and nitrogen (N2) forming an isotopic interference pair. When performing step S14 above, the preset standard spectral library matrix S contains standard spectral vectors S of components such as Ar, N2, O2, H2O, CO2, CO, AsH3, and PH3. kWhen constructing the interference factor matrix F based on this matrix S, the interference coefficients of a large number of Ar peak tails on adjacent mass channels were measured and quantified, especially the interference coefficients of the characteristic mass channels of AsH3 and PH3, as well as the interference coefficients between CO and N2. During steps S15 and S16, considering the weak signals of the characteristic mass channels of AsH3 and PH3, the weight matrix W in the weighted ridge regression optimization model automatically assigns higher weights to these channels; simultaneously, the interference factor matrix F is used to accurately subtract interference from the Ar main peak tail and other components. Finally, by solving the weighted ridge regression optimization model, the partial pressure vectors C of all components, including trace amounts of AsH3 and PH3, can be stably and accurately obtained even in the presence of strong background (Ar, N2), with a detection limit on the order of one part per billion. This effectively overcomes the technical challenges of isotope interference and strong signals masking weak signals, achieving high-precision and high-stability detection of complex gas components in semiconductor process chambers.

[0099] It is understood that the execution order of steps S11 to S16 above is only an illustration. In other embodiments, other execution orders may be used, and some steps may be split or combined. This is not limited here.

[0100] Embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon. When the executable computer program in the storage medium is executed by a processor, it can realize the method for detecting gas components in the semiconductor process chamber 10 in any of the above embodiments.

[0101] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements a method for detecting gas components within a semiconductor process chamber 10 as described in the first aspect and any possible implementation thereof.

[0102] Embodiments of this application also provide an electronic device 300, such as... Figure 3 As shown, the electronic device 300 includes a memory 302 and a processor 301. The memory 302 is used to store computer programs executable by the processor 301. The processor 301 is used to execute the computer programs in the memory 302 to implement the method for detecting gas components in the semiconductor process chamber 10 in any of the above embodiments.

[0103] Figure 3 The electronic device 300 shown also includes a communication interface 303. The processor 301, memory 302, and communication interface 303 are connected via a communication bus and communicate with each other.

[0104] Processor 301 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of programs in the above scheme.

[0105] Communication interface 303 is used to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Networks (WLAN), etc.

[0106] Memory 302 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory may exist independently and be connected to the processor via a bus. Memory may also be integrated with the processor.

[0107] In the accompanying drawings, some structural or methodological features may be shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be necessary. Rather, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Furthermore, including structural or methodological features in a particular figure does not imply that such features are required in all embodiments, and in some embodiments, these features may be omitted or may be combined with other features.

[0108] It should be noted that all units / modules mentioned in the device embodiments of this application are logical units / modules. Physically, a logical unit / module can be a physical unit / module, a part of a physical unit / module, or a combination of multiple physical units / modules. The physical implementation of these logical units / modules themselves is not the most important factor; the combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed in this application. Furthermore, to highlight the innovative aspects of this application, the above-described device embodiments of this application have not introduced units / modules that are not closely related to solving the technical problems proposed in this application. This does not mean that the above-described device embodiments do not contain other units / modules.

[0109] It should be noted that in the examples and description of this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0110] Although this application has been illustrated and described with reference to certain preferred embodiments thereof, those skilled in the art should understand that various changes in form and detail may be made thereto without departing from the spirit and scope of this application.

Claims

1. A method for detecting gas samples in a semiconductor process chamber, characterized in that, Applicable to a gas component analysis system, the gas component analysis system including a vacuum sample chamber, a mass analyzer, and a processing unit, the method includes: A gas sample is collected from the semiconductor process chamber, and the gas sample is used as the mixed sample to be analyzed. The background mass spectrum of the vacuum sample chamber and the original mass spectrum of the mixed sample to be analyzed in the vacuum sample chamber are obtained using the mass analyzer, wherein the background mass spectrum and the original mass spectrum are acquired under the same instrument conditions. The background mass spectrum and the original mass spectrum are preprocessed to obtain the background spectrum vector B and the measured spectrum vector M. Obtain a preset standard spectral library matrix S, and construct an interference factor matrix F based on the preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship between multiple components related to the mixed sample to be analyzed. The preset standard spectral library matrix S includes standard spectral vectors S of the multiple components. k ; Based on the measured spectrum vector M, the background spectrum vector B, the preset standard spectrum library matrix S, and the interference factor matrix F, a weighted ridge regression optimization model is constructed with the partial pressure vector C as the solution objective and the non-negativity of the partial pressure vector C as the constraint condition. The partial pressure vector C is used to describe the partial pressure of each component in the mixed sample. Solve the weighted ridge regression optimization model to obtain the partial pressure vector C that minimizes the objective function of the weighted ridge regression optimization model. Use the partial pressure vector C as the quantitative detection result of the partial pressure of each component in the mixed sample to be analyzed. The quantitative detection result is used to analyze the real-time changes of gas components in the semiconductor process chamber.

2. The method for detecting gas samples in a semiconductor process chamber according to claim 1, characterized in that, The interference factor matrix F, constructed based on the preset standard spectral library matrix S to characterize the mass spectrometry signal interference relationship among the multiple components, includes: Obtain the measured spectrum vector M of each component in the preset standard spectrum library matrix S. k and standard spectral vector S k ; Based on the measured spectral vector M of each component k and standard spectral vector S k The scaling factor α and the disturbance vector f are obtained by solving the least squares method. k Make the relational expression Established; For the interference vector f k After normalization; The interference vector f, after normalization of each component according to the preset component order, is... k The diagonal elements of the interference factor matrix F are set to 1 and respectively, serving as off-diagonal column vectors, to construct the interference factor matrix F.

3. The method for detecting gas samples in a semiconductor process chamber according to claim 1, characterized in that, The objective function of the weighted ridge regression optimization model The following relationship must be satisfied: Where W is the weight matrix, λ is the regularization parameter, Γ is the constraint matrix, and the partial pressure of each component in the partial pressure vector C is... 0, and the regularization parameter λ is determined by cross-validation.

4. The method for detecting gas samples in a semiconductor process chamber according to claim 3, characterized in that, The weight matrix W is a diagonal matrix, and the i-th diagonal element w of the weight matrix W i It is calculated based on the signal strength or signal-to-noise ratio of the measured spectrum vector M and the background spectrum vector B on the corresponding i-th quality channel.

5. The method for detecting gas samples in a semiconductor process chamber according to claim 4, characterized in that, The i-th diagonal element w of the weight matrix W i The calculation formula is: in, This is the noise variance estimate for the i-th quality channel.

6. The method for detecting gas samples in a semiconductor process chamber according to claim 3, characterized in that, Solving the weighted ridge regression optimization model to obtain the pressure vector C that minimizes the objective function of the weighted ridge regression optimization model includes: The weighted ridge regression optimization model is solved using the iterative reweighted least squares method to obtain the pressure vector C that minimizes the objective function of the weighted ridge regression optimization model: Initialize the voltage divider vector C and the weight matrix W; Calculate the residual vector R based on the current pressure divider vector C, and update the weight matrix W according to the residual vector R; The partial pressures of each component in the partial pressure vector C Under the constraint of 0, the weighted ridge regression optimization model is solved based on the updated weight matrix W to update the pressure vector C; Determine whether the updated pressure vector C satisfies the preset convergence condition. If not, repeat the process of updating the weight matrix W and updating the pressure vector C until the pressure vector C satisfies the preset convergence condition. The formula for calculating the residual vector R is as follows: .

7. The method for detecting gas samples in a semiconductor process chamber according to claim 1, characterized in that, The mixed sample to be analyzed includes a gas sample in a chamber during a semiconductor process, the type of which includes thin film, etching, and ion implantation.

8. The method for detecting gas samples in a semiconductor process chamber according to claim 1, characterized in that, The real-time changes in gas composition within the semiconductor process chamber are analyzed using the following methods: The percentage of each component in the gas composition within the semiconductor process chamber is obtained based on the quantitative detection results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the executable computer program in the storage medium is executed by a processor, it can implement the method as described in any one of claims 1 to 8.

10. An electronic device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program executable by the processor; and the processor executes the computer program in the memory to implement the method as described in any one of claims 1 to 8.

Images