Spectroscopic analysis-based online monitoring method for quality of developer for semiconductor 4D equipment

Abstract

The application discloses a semiconductor 4D equipment developing solution quality online monitoring method based on spectrum analysis, relates to the technical field of developing solution quality monitoring, and comprises the following steps: establishing an etching light path baseline calibration model, tracking a developing solution spectrum baseline, obtaining a baseline drift trend, and not interrupting collection when the drift trend is less than or equal to a set drift threshold; otherwise, collecting a wafer factory environment data, creating a calibration necessity prediction strategy, and outputting calibration necessity; according to the developing solution spectrum characteristics, introducing local rapid calibration instead of full-width calibration, performing high-precision scanning on selected wave bands, and outputting developing solution quality coefficients; setting a small temperature control calibration liquid bin, automatically and periodically switching quantitative calibration reference liquid to flow through a sensor probe according to the developing solution quality coefficients, and quickly switching back to production liquid after calibration is completed. The application solves the problem that the traditional full-width spectrum calibration method is time-consuming, frequent calibration affects the production line rhythm and reduces the production efficiency.

Description

Technical Field

[0001] This invention relates to the field of developer quality monitoring technology, specifically to an online monitoring method for developer quality in semiconductor 4D equipment based on spectral analysis. Background Technology

[0002] Even slight fluctuations in the concentration of the developer can cause image deviations, requiring real-time, high-precision online monitoring. Traditional offline sampling or online electrochemical or titration monitoring is slow to respond and susceptible to corrosion and baseline drift. While 4D monitoring technologies such as snapshot-type spectral imaging can provide high spatiotemporal resolution, they also bring new challenges such as data redundancy, cumbersome calibration, and sensor corrosion.

[0003] While offline sampling titration or electrochemical methods are accurate, their long response cycles make them difficult to meet the demands of high-speed production lines. Furthermore, the strong alkalinity of the developer can corrode quartz and metal surfaces, generating corrosion products that adhere and cause baseline nonlinear drift.

[0004] Therefore, this invention provides an online monitoring method for the quality of developer in semiconductor 4D equipment based on spectral analysis. Summary of the Invention

[0005] The purpose of this invention is to provide an online monitoring method for the quality of the developer in semiconductor 4D equipment based on spectral analysis, so as to solve the existing problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring method for the quality of developer in semiconductor 4D equipment based on spectral analysis, comprising the following steps: S1. Establish a baseline verification model for the corrosion optical path, track the spectral baseline of the developer, and obtain the baseline drift trend. When the drift trend is greater than the set drift threshold, run S2. When the drift trend is less than or equal to the set drift threshold, do not interrupt the acquisition and continue normal operation. S2. Collect wafer fab environmental data, create a calibration necessity prediction strategy, and output calibration necessity; S3. Based on the spectral characteristics of the developer, local rapid calibration is introduced to replace full-amplitude calibration. High-precision scanning is performed on the selected band, and the developer quality coefficient is output. S4. Set up a small temperature-controlled calibration liquid tank, and automatically and periodically switch the quantitative calibration reference liquid flow through the sensor probe according to the developer quality coefficient. After calibration, quickly switch back to the production liquid.

[0007] A further improvement of this invention is that the corrosion optical path baseline verification model is based on the optical characteristics of corrosion products from accelerated aging tests, and establishes a corrosion-baseline drift comparison library. This data was embedded as prior data into the corrosion optical path baseline verification model; the wavelength of the "zero zone" where the developer absorbs the least in the current band was selected. As a baseline measurement reference point, the real-time baseline B(n) is calculated using RLS based on the real-time detection signal d(n). The cumulative weighted error of the baseline drift is calculated by first accumulating the values ​​from step 0 to step n for each time step. At each step, the square of the change in the baseline measurement value between two consecutive moments is calculated and multiplied by a weighting factor. All weighted squared changes are summed to obtain the baseline drift trend E(n) at the current moment n. In the update of the corrosion optical path baseline verification model, the corrosion-baseline drift reference library is used. As process noise, slow drift is compensated in real time, and narrowband linear regression correction is performed on the M bands with the largest drift to complete the updated B'(n).

[0008] A further improvement of this invention lies in that the calibration necessity prediction strategy, through interfacing with the wafer field environment monitoring system, acquires environmental data in real time, including temperature, humidity, vibration, and cleanliness; extracts the baseline drift trend E(n) and drift threshold as drift labels; uses environmental data as feature input and drift labels as supervision signals, and establishes a necessity prediction model based on an LSTM model to capture the dynamic impact of time-series environmental changes on baseline stability. The final hidden state of the LSTM model extracts the degree of impact of the current environmental changes on baseline stability. After passing through a fully connected layer, the internal state is compressed into a real number output and normalized to the (0,1) interval by a Sigmoid activation function. In this process, the cleanliness weight is 0.5, and the sum of the weights of the other environmental data is 0.5, thus obtaining the final score as the calibration necessity score Cns.

[0009] A further improvement of the present invention is that the calibration necessity prediction strategy does not include setting a calibration necessity threshold. When the calibration necessity score is greater than the calibration necessity threshold, the calibration necessity prediction strategy intelligently determines whether there is a production line idle period window. If there is, then S3 is executed. If there is no, or when the calibration necessity score is less than or equal to the calibration necessity threshold, then the pending calibration status is recorded, and the calibration is executed immediately when the next production line idle period window exists.

[0010] A further improvement of this invention is that the local rapid calibration includes a key band intelligent selection strategy, a segmented scanning protocol design strategy, a dynamic calibration parameter update strategy, and a developer quality coefficient output strategy. The intelligent band selection strategy includes constructing a reference value Con for the developer TMAH concentration and its corresponding historical spectral feature matrix SPM; and calculating the first weighting coefficient matrix of each wavelength variable using least squares. The first 5% of bands in the first weighting coefficient matrix of the current spectral image are selected as the key calibration area. Within the band range of the key calibration area, the signal-to-noise ratio (SNR) is calculated, and bands with an SNR greater than 50 are retained as high-precision scanning bands and fed into the high-precision scanning band sequence (HPS). A spectral reflectance score is calculated using a spectral combination strategy, which calculates the combined feature values ​​of the high-precision scanning bands. The spectral reflectance score sequence Res is calculated by using the correlation coefficient (cor) between all combinations and the developer TMAH concentration reference value Con. The formula for calculating the characteristic value of high-precision scanning band combinations is as follows: , and This represents any combination of band reflectance in the high-precision scan band sequence (HPS); the maximum value of the feature value of the high-precision scan band combination is extracted as the spectral reflectance score Srs.

[0011] A further improvement of this invention is that the segmented scanning protocol design strategy includes high-precision scanning of key bands and low-precision sampling of non-key areas.

[0012] A further improvement of this invention is that the dynamic calibration parameter update strategy includes continuously optimizing the calibration model by combining real-time production data, and updating the first weight coefficient matrix after each calibration is completed: , The weight increments were calculated using the latest calibration data; subsequently, they were calculated using Mahalanobis distance. Detect abnormal bands and automatically remove them. This band will be temporarily blocked until the next calibration verification.

[0013] A further improvement of this invention is that the developer quality coefficient output strategy includes extracting spectral reflectance scores and retrieving the current developer TMAH concentration using a high-precision scanning band sequence HPS. optical path l =1mm, where Indicates the mean of the calibration values. This represents the average band reflectance in a high-precision scanned band sequence. Extract the ensemble peaks (CPI) of the metal contaminant interference bands and establish the theoretical absorbance of the pure developer in the metal contaminant interference bands. ,in, Indicating the first peak in the set of interference bands from metallic pollutants Reflectance of each wavelength band; calculate the relative deviation between the actual absorbance and the theoretical value for each wavelength. ,in This represents the actual absorbance value of the interference peak band of metallic pollutants; the pollutant index is obtained based on the relative deviation between the actual absorbance and the theoretical value at each wavelength. ; Developer quality factor .

[0014] A further improvement of the present invention is that step S4 specifically includes using a microchannel integrated piezoelectric drive valve made of PDMS material, and using a flow mode switching equation to control the switching time of the calibration liquid and the production liquid flow path.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention firstly shortens the calibration time by performing instantaneous scanning of key narrow bands through local rapid calibration, which can eliminate production line stoppages caused by calibration and ensure that the wafer processing cycle is not affected; and combined with environmental prediction and intelligent scheduling, calibration is inserted only when necessary, further reducing the calibration frequency and realizing continuous online monitoring. 2. Secondly, the corrosion-baseline drift control library obtained from the accelerated aging test is embedded into the Kalman filter process through the corrosion optical path baseline verification model. This compensates for the slow drift caused by the micro-corrosion of the quartz flow cell by the strong alkali or organic solvent of the developer in real time, making up for the drift effect of equipment corrosion on the optical path baseline. Furthermore, through dual compensation of adaptive filtering and differential reference path, the measurement error of the developer concentration can be reduced. Attached Figure Description

[0016] Figure 1 This is a flowchart of the online monitoring method for the quality of the developer in a semiconductor 4D device based on spectral analysis, as described in this invention. Figure 2 This is a flowchart of the local rapid calibration process for the online monitoring method of developer quality in semiconductor 4D equipment based on spectral analysis, as described in this invention. Detailed Implementation

[0017] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0018] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0019] Example 1 Figure 1 The flowchart of the online monitoring method for developer quality in semiconductor 4D equipment based on spectral analysis disclosed in this embodiment is shown, and the steps are as follows: S1. Establish a baseline verification model for the corrosion optical path, track the spectral baseline of the developer, and obtain the baseline drift trend. When the drift trend is greater than the set drift threshold, run S2. When the drift trend is less than or equal to the set drift threshold, do not interrupt the acquisition and continue normal operation. The corrosion optical path baseline verification model is based on the optical characteristics of corrosion products from accelerated aging tests, establishing a corrosion-baseline drift comparison library. This data is embedded as prior data into the corrosion optical path baseline verification model to address the problem that semiconductor developers often contain strong alkaline or organic solvents (such as tetramethylammonium hydroxide), which may cause corrosion of the detector flow cell or sensor material (such as quartz) after long-term contact, thereby changing the light transmittance or electrochemical response characteristics of the optical path and causing baseline drift. The corrosion optical path baseline verification model is based on a Kalman filter simulation of corrosion experiments on a quartz sensor at high temperature using a developing solution, establishing a relationship between the corrosion layer thickness dcl and the transmittance I. , This represents the quartz corrosion coefficient; the wavelength of the "zero zone" where the developer absorbs the least in the current band is selected. As a baseline measurement reference point, the real-time baseline B(n) is calculated using RLS based on the real-time detection signal d(n). The cumulative weighted error of the baseline drift is calculated by first accumulating the values ​​from step 0 to step n for each time step. At each step, the square of the change in the baseline measurement value between two consecutive moments is calculated and multiplied by a weighting factor. All weighted squared changes are summed to obtain the baseline drift trend E(n) at the current moment n. In the update of the corrosion optical path baseline verification model, the corrosion-baseline drift reference library is used. As process noise, slow drift is compensated in real time, and narrowband linear regression correction is performed on the M bands with the largest drift to complete the updated B'(n).

[0020] S2. Collect wafer fab environmental data, create a calibration necessity prediction strategy, and output the calibration necessity. The calibration necessity prediction strategy connects with the wafer fab environmental monitoring system to acquire environmental data in real time, including temperature, humidity, vibration, and cleanliness. Extract the baseline drift trend E(n) and drift threshold as drift labels. Use environmental data as feature input and drift labels as supervision signals. Establish a necessity prediction model based on an LSTM model to capture the dynamic impact of time-series environmental changes on baseline stability. The final hidden state of the LSTM model extracts the degree of impact of the current environmental changes on baseline stability. After passing through a fully connected layer, the internal state is compressed into a real number output and normalized to the (0,1) interval by the Sigmoid activation function. In this process, since the wafer has high cleanliness requirements, the cleanliness weight is 0.5, and the sum of the weights of the other environmental data is 0.5, thus obtaining the final score as the calibration necessity score Cns.

[0021] The calibration necessity prediction strategy does not yet include setting a calibration necessity threshold. When the calibration necessity score is greater than the calibration necessity threshold, the calibration necessity prediction strategy intelligently determines whether there is a production line idle period window. If there is, then run S3. If there is no, or when the calibration necessity score is less than or equal to the calibration necessity threshold, then record the pending calibration status and execute immediately when the next production line idle period window exists.

[0022] This invention utilizes the acquisition and dynamic correlation modeling of wafer fab environmental data; a lightweight time series prediction model; and intelligent inference of calibration necessity and reasonable insertion of production line calibration. This significantly reduces unnecessary calibration frequency and ensures the stability of the semiconductor 4D equipment developer spectral monitoring system and the production line cycle efficiency.

[0023] S3. Based on the spectral characteristics of the developer, local rapid calibration is introduced to replace full-amplitude calibration. High-precision scanning is performed on the selected band, and the developer quality coefficient is output. S4. Set up a small temperature-controlled calibration liquid tank, and automatically and periodically switch the quantitative calibration reference liquid flow through the sensor probe according to the developer quality coefficient. After calibration, quickly switch back to the production liquid.

[0024] Example 2 Based on the inventive concept of Embodiment 1, this embodiment proposes a specific implementation process for the local rapid calibration in step S3. Figure 2 The present invention illustrates a flowchart of a method for rapid local calibration of the online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis, which specifically includes: The local rapid calibration includes a key band intelligent selection strategy, a segmented scanning protocol design strategy, a dynamic calibration parameter update strategy, and a developer quality coefficient output strategy. In spectral imaging, snapshot spectral imaging chips can acquire global image and spectral information within the field of view in a single exposure, resulting in higher image acquisition efficiency. However, this method also generates a large amount of redundant data during calibration, increasing the burden of data transmission, storage, and processing. The intelligent band selection strategy includes constructing a reference value Con for the developer TMAH concentration and its corresponding historical spectral feature matrix SPM; and calculating the first weighting coefficient matrix of each wavelength variable using least squares. Select the top 5% of bands in the first weighting coefficient matrix of the current spectral image as the key calibration region, and calculate the signal-to-noise ratio within the band range of the key calibration region. ,in, This represents the average signal strength in the critical calibration area. and The mean and standard deviation of the region near the baseline are represented; bands with a signal-to-noise ratio greater than 50 are retained as high-precision scanning bands and sent to the high-precision scanning band sequence HPS. The spectral reflectance score is calculated using a spectral combination strategy, which calculates high-precision combined characteristic values ​​of scanning bands. The spectral reflectance score sequence Res is calculated by using the correlation coefficient (cor) between all combinations and the developer TMAH concentration reference value Con. The formula for calculating the characteristic value of high-precision scanning band combinations is as follows: , and This represents any combination of band reflectance in the high-precision scan band sequence (HPS); the maximum value of the feature value of the high-precision scan band combination is extracted as the spectral reflectance score Srs.

[0025] The segmented scanning protocol design strategy includes high-precision scanning of key bands and low-precision sampling of non-key areas.

[0026] The dynamic calibration parameter update strategy includes continuously optimizing the calibration model by combining real-time production data, and updating the first weight coefficient matrix after each calibration is completed: , The weight increments were calculated using the latest calibration data; subsequently, they were calculated using Mahalanobis distance. Detecting abnormal bands, specifically manifested as follows: And automatically remove abnormal bands, if This band will be temporarily blocked until the next calibration verification. The developer quality coefficient output strategy includes extracting spectral reflectance scores and retrieving the current developer TMAH concentration from the high-precision scan band sequence HPS using the Beer-Lambert law. optical path l =1mm, where Indicates the mean of the calibration values. This represents the average band reflectance in a high-precision scanned band sequence. Extract the set of peak bands (CPI) of metal contaminants, and establish the theoretical absorbance value of the pure developer in the metal contaminant interference peak band by formulating the value by those skilled in the art, through experimental calibration, or based on existing data. ,in, Indicating the first peak in the set of interference bands from metallic pollutants Reflectance of each wavelength band; calculate the relative deviation between the actual absorbance and the theoretical value for each wavelength. ,in The actual absorption value of the interference peak band of metallic pollutants is expressed by the following formula: Updated every 100ms This indicates the intensity of transmitted light after passing through the developer solution being tested; This represents the original light intensity value of the light emitted by the light source when it passes through a blank reference sample (such as a pure solvent or air) without being absorbed or scattered; The pollutant index is obtained based on the relative deviation between the actual absorbance and the theoretical value at each wavelength. ; Obtain the developer quality coefficient .

[0027] This invention solves the problem that 4D systems need to continuously calibrate the spectrometer online, which affects the production line cycle time if the calibration process takes up too much production time.

[0028] This embodiment also provides the implementation process of step S4 in embodiment 1, specifically including using a microchannel integrated piezoelectrically driven valve made of PDMS material, and controlling the switching time of the calibration liquid and production liquid flow paths through a flow mode switching equation; the flow mode switching equation is expressed as: Where r represents the radius of the flow channel cross-section, This represents the pressure difference between the fluid inlet and outlet, which is the driving force of the flow. It can be set to 50 kPa. The value represents the fluid dynamic viscosity, which can be set to 1 cP, and L represents the length of the fluid flow path from the inlet to the outlet.

[0029] The threshold and weight settings can be set by default according to the present invention, or they can be set by those skilled in the art.

[0030] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0031] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0032] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0033] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0034] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for online monitoring of the quality of a developer solution for semiconductor 4D equipment based on spectral analysis, characterized in that: Includes the following steps: S1. Establish a baseline verification model for the corrosion optical path, track the spectral baseline of the developer, and obtain the baseline drift trend. When the drift trend is greater than the set drift threshold, run S2. When the drift trend is less than or equal to the set drift threshold, do not interrupt the acquisition and continue normal operation. S2. Collect wafer fab environmental data, create a calibration necessity prediction strategy, and output calibration necessity; S3. Based on the spectral characteristics of the developer, local rapid calibration is introduced to replace full-amplitude calibration. High-precision scanning is performed on the selected band, and the developer quality coefficient is output. S4. Set up a small temperature-controlled calibration liquid tank, and automatically and periodically switch the quantitative calibration reference liquid flow through the sensor probe according to the developer quality coefficient. After calibration, quickly switch back to the production liquid. The local rapid calibration includes a key band intelligent selection strategy, a segmented scanning protocol design strategy, a dynamic calibration parameter update strategy, and a developer quality coefficient output strategy. The intelligent band selection strategy includes constructing a reference value Con for the developer TMAH concentration and its corresponding historical spectral feature matrix SPM; and calculating the first weighting coefficient matrix of each wavelength variable using least squares. The first 5% of bands in the first weighting coefficient matrix of the current spectral image are selected as the key calibration area. Within the band range of the key calibration area, the signal-to-noise ratio (SNR) is calculated, and bands with an SNR greater than 50 are retained as high-precision scanning bands and fed into the high-precision scanning band sequence (HPS). A spectral reflectance score is calculated using a spectral combination strategy, which calculates the combined feature values ​​of the high-precision scanning bands. The spectral reflectance score sequence Res is calculated by using the correlation coefficient (cor) between all combinations and the developer TMAH concentration reference value Con. The formula for calculating the characteristic value of high-precision scanning band combinations is as follows: , and The reflectance combination of any band in the high-precision scanned band sequence (HPS) is represented; the maximum value of the feature value of the high-precision scanned band combination is extracted as the spectral reflectance score Srs. The developer quality coefficient output strategy includes extracting spectral reflectance scores and using high-precision scanning band sequence HPS to retrieve the current developer TMAH concentration. optical path l =1mm, where Indicates the mean of the calibration values. This represents the average band reflectance in a high-precision scanned band sequence; where, For the updated baseline; Extract the ensemble peaks (CPI) of the metal contaminant interference bands and establish the theoretical absorbance of the pure developer in the metal contaminant interference bands. ,in, Indicating the first peak in the set of interference bands from metallic pollutants Reflectance of each wavelength band; calculate the relative deviation between the actual absorbance and the theoretical value for each wavelength. ,in This represents the actual absorbance value of the interference peak band of metallic pollutants; the pollutant index is obtained based on the relative deviation between the actual absorbance and the theoretical value at each wavelength. ; Developer quality factor .

2. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 1, characterized in that: The corrosion optical path baseline verification model is based on the optical characteristics of corrosion products from accelerated aging tests, establishing a corrosion-baseline drift comparison library. This data was embedded as prior data into the corrosion optical path baseline verification model; the zero-region wavelength with the minimum absorption of the developer in the current band was selected. As a baseline measurement reference point, the real-time baseline B(n) is calculated using RLS based on the real-time detection signal d(n). The cumulative weighted error of the baseline drift is calculated by first accumulating the values ​​from step 0 to step n for each time step. At each step, the square of the change in the baseline measurement value between two consecutive moments is calculated and multiplied by a weighting factor. All weighted squared changes are summed to obtain the baseline drift trend E(n) at the current moment n. In the update of the corrosion optical path baseline verification model, the corrosion-baseline drift reference library is used. As process noise, slow drift is compensated in real time, and narrowband linear regression correction is performed on the M bands with the largest drift to complete the updated B'(n).

3. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 2, characterized in that: The calibration necessity prediction strategy interfaces with the wafer field environmental monitoring system to acquire environmental data in real time, including temperature, humidity, vibration, and cleanliness. It extracts the baseline drift trend E(n) and drift threshold as drift labels. Using environmental data as feature input and drift labels as supervisory signals, a necessity prediction model is established based on an LSTM model to capture the dynamic impact of time-series environmental changes on baseline stability. The final hidden state of the LSTM model extracts the degree of impact of current environmental changes on baseline stability. After passing through a fully connected layer, the internal state is compressed into a real number output and normalized to the (0,1) interval using a Sigmoid activation function. During this process, the cleanliness weight is 0.5, and the sum of the weights of the other environmental data is 0.5, thus obtaining the final score as the calibration necessity score Cns.

4. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 3, characterized in that: The calibration necessity prediction strategy does not yet include setting a calibration necessity threshold. When the calibration necessity score is greater than the calibration necessity threshold, the calibration necessity prediction strategy intelligently determines whether there is a production line idle period window. If there is, then run S3. If there is no, or when the calibration necessity score is less than or equal to the calibration necessity threshold, then record the pending calibration status and execute immediately when the next production line idle period window exists.

5. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 1, characterized in that: The segmented scanning protocol design strategy includes high-precision scanning of key bands and low-precision sampling of non-key areas.

6. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 1, characterized in that: The dynamic calibration parameter update strategy includes continuously optimizing the calibration model by combining real-time production data, and updating the first weight coefficient matrix after each calibration is completed: , The weight increment calculated from the latest calibration data; Subsequently, the Mahalanobis distance was used. Detect abnormal bands and automatically remove them. This band will be temporarily blocked until the next calibration verification.

7. The method for online monitoring of developer quality in semiconductor 4D equipment based on spectral analysis according to claim 1, characterized in that: Step S4 specifically includes using a microchannel integrated piezoelectrically driven valve made of PDMS material, and controlling the switching time of the calibration liquid and production liquid flow paths through a flow mode switching equation.

Images