A data processing method and system for quality change detection of semiconductor process parts

Abstract

The present application discloses a data processing method and system for detecting degradation of semiconductor process components. The method comprises the following steps: obtaining multiple Raman spectral data of a semiconductor process component, and performing multiple operations on the Raman spectral data to obtain a first degradation state simulation parameter for indicating the degree of aging of the semiconductor process component and a second degradation state simulation parameter for indicating the degree of variation in the internal molecular structure of the semiconductor process component.

Description

Technical Field

[0001] The present application relates to the field of data processing, and more specifically, to a data processing method and system for detecting quality changes in semiconductor process components. Background Art

[0002] The yield rate of semiconductor manufacturing is significantly impacted by the availability of semiconductor process components used in semiconductor production equipment. Any tiny defects on a semiconductor process component, or any residual molecules on its surface, can become a source of contamination in the semiconductor production process. In advanced processes, as process line diameters shrink, tiny molecules that previously had no impact on yield can now severely impact semiconductor manufacturing quality. Furthermore, extremely harsh semiconductor process environments can easily cause degradation of semiconductor process components. Therefore, monitoring the degradation status of semiconductor process components at all times has become a critical requirement in advanced processes.

[0003] Currently, understanding the degradation status of semiconductor process components is typically achieved through empirical testing or the use of testing tools. However, relying solely on empirical testing often results in components being retired even when degradation is minimal, resulting in low efficiency. Furthermore, currently available testing tools can only measure dimensional changes in these components and cannot accurately determine the changes in properties before micro-defects occur. For advanced processes, these tools are insufficient to accurately determine the true quality of semiconductor process components. In particular, even new semiconductor process components can experience degradation, and simply comparing the degradation status of components based on new and old components is no longer sufficient to meet the needs of advanced processes.

[0004] In order to solve the above problems and meet the needs of high-end processes, the applicant in this case has previously proposed a detection system that can accurately detect the quality change status of semiconductor process parts and components. However, the detection results are multi-dimensional spectral data, which is not very convenient to interpret. In order to improve the convenience of interpretation, a data processing tool that can be used with the detection system is also needed to convert these spectral data into data that is easy to interpret. In addition, it is hoped that these data can be used to predict the life span of various semiconductor process parts and components, so as to grasp the usage status of each semiconductor process part and component in advance, so that each semiconductor process part and component can be used in the best condition, avoiding the use of semiconductor process parts and components with severe quality changes in the semiconductor process and affecting the production yield. Summary of the Invention

[0005] In view of the above problems, the present application proposes a data processing method and system for quality change detection of semiconductor process components, which are used to generate quality change state simulation parameters that can predict the life span of various semiconductor process components.

[0006] In one aspect, the present invention provides a data processing method for detecting degradation of semiconductor processing components. In one embodiment, the data processing method includes the following steps: obtaining multiple Raman spectral data of a semiconductor processing component; and performing multiple computations on the Raman spectral data to obtain a first degradation state simulation parameter representing the degree of degradation of the semiconductor processing component and a second degradation state simulation parameter representing the degree of internal molecular structure variation of the semiconductor processing component.

[0007] In one embodiment, the computational processing implemented in the proposed data processing method includes the following steps: performing an alignment operation in at least two dimensions on the obtained Raman spectral data; performing a derivative operation on the Raman spectral data after the alignment operation to obtain a first parameter; performing scaling and normalization processing on the first parameter obtained after the derivative operation to obtain a second parameter; performing a dimensionality reduction operation on the second parameter obtained after the scaling and normalization processing to obtain a third parameter; and performing a standard deviation operation on the third parameter obtained after the dimensionality reduction operation to obtain a first qualitative change state simulation parameter and a second qualitative change state simulation parameter.

[0008] In one embodiment, the derivative operation implemented in the proposed data processing method has the following steps: performing a ratio operation on the differences between all spectral peaks of each Raman spectral data in different dimensions to obtain a first parameter that can represent the change in the intensity values ​​of all spectral peaks of each Raman spectral data as the Raman spectral shift changes.

[0009] In one embodiment, one of the alignment operation, scaling and normalization operation, and dimensionality reduction operation implemented in the proposed data processing method includes a matrix operation.

[0010] In one embodiment, the first qualitative change state simulation parameter obtained by the proposed data processing method is related to the average value obtained after the standard deviation calculation, and the second qualitative change state simulation parameter is related to the discrete value obtained after the standard deviation calculation.

[0011] In another aspect, the present invention provides a data processing system for detecting quality changes in semiconductor process components. In one embodiment, the proposed data processing system for detecting quality changes in semiconductor process components comprises an alignment operation unit, a derivative operation unit, a scaling and normalization unit, a dimensionality reduction operation unit, and a standard deviation operation unit. The alignment operation unit receives multiple Raman spectral data from a semiconductor process component and performs a unified error calculation on the Raman spectral data across all dimensions to eliminate inconsistencies in detection results from different quality change detection systems for the semiconductor process component. The derivative operation unit is electrically connected to the alignment operation unit and receives the Raman spectral data processed by the alignment operation unit. The derivative operation unit performs a ratio calculation on the differences between all spectral peaks in different dimensions on each received Raman spectral data to obtain a first parameter representing the change in the intensity of all spectral peaks in each Raman spectral data as the Raman spectral shift changes. The scaling and normalization unit is electrically connected to the derivative operation unit and receives the first parameter and performs scaling and normalization on the received first parameter to obtain a second parameter. The dimension reduction operation unit is electrically connected to the scale normalization unit and is configured to receive a second parameter and perform a dimension reduction operation on the received second parameter to obtain a third parameter. The standard deviation operation unit is electrically connected to the dimension reduction operation unit and is configured to receive a third parameter and perform a standard deviation operation on the received third parameter to obtain a first qualitative change state simulation parameter representing the degree of aging of the semiconductor process component and a second qualitative change state simulation parameter representing the degree of variation in the internal molecular structure of the semiconductor process component.

[0012] In one embodiment, the first quality change state simulation parameter is a life index of a semiconductor process component, and a smaller life index indicates a lower degree of aging of the semiconductor process component.

[0013] In one embodiment, the vital indicator is related to an average value obtained after being processed by a standard deviation calculation unit.

[0014] In one embodiment, the second qualitative change state simulation parameter is the dispersion of semiconductor process components, and a smaller dispersion indicates a smaller degree of variation in the internal molecular structure of the semiconductor process components.

[0015] In one embodiment, the dispersion is related to the discrete value obtained after being processed by the standard deviation calculation unit.

[0016] Beneficial effects

[0017] According to various embodiments of the present invention, a data processing method and system for detecting degradation of semiconductor process components, described herein, utilizes an algorithm, including alignment, derivative calculations, scale normalization, dimensionality reduction, and standard deviation calculations, to apply to Raman spectral data to obtain a lifespan indicator and dispersion that can be used to predict the lifespan of various semiconductor process components. Because the lifespan indicator and dispersion are one-dimensional data converted from Raman spectral data, the complexity of interpreting Raman spectral data is simplified. Furthermore, inspectors can assess the operational status of various semiconductor process components without interpreting Raman spectral data. The lifespan of various semiconductor process components can be predicted solely based on the values ​​of the lifespan indicator and dispersion, thereby improving the efficiency of component utilization.

[0018] In order to make the above features and advantages of the present application more obvious and easy to understand, embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a system architecture diagram showing a data processing system for a quality change detection system for semiconductor process components according to an embodiment of the present invention.

[0020] Figure 2 It is a plan view showing a semiconductor process component to be tested. Figure 1 Multiple target areas under inspection by the quality change detection system for semiconductor process parts.

[0021] Figure 3 The diagram is a plan view showing a plurality of Raman spectral data before and after being processed by an alignment operation unit of a data processing system of a quality change detection system for semiconductor process components according to an embodiment of the present invention.

[0022] Figure 4 The diagram is a plan view showing Raman spectral data of three different target areas of a semiconductor process component received by a derivative operation unit of a data processing system of a quality change detection system for semiconductor process components according to an embodiment of the present invention.

[0023] Figure 5 1 is a schematic plan view showing Raman spectral data of a semiconductor process component of the same specification in three different external environments simulated by a data processing system of a quality change detection system for a semiconductor process component according to an embodiment of the present invention.

[0024] Figure 6 The diagram is a plan view showing Raman spectral data of three different target areas of the same semiconductor process component simulated by a data processing system of a quality change detection system for semiconductor process components according to an embodiment of the present invention.

[0025] Figure 7 The present invention is a flow chart showing a data processing method for a quality change detection system for semiconductor process components according to an embodiment of the present invention.

[0026] Figure 8 The diagram is a plan view showing the relationship between the vital indicators, the dispersion and the degree of quality change of semiconductor process components obtained by the data processing system of the quality change detection system for semiconductor process components according to one embodiment of the present invention. DETAILED DESCRIPTION

[0027] This application discloses a data processing method and system for detecting quality changes in semiconductor manufacturing components. The following descriptions, such as those related to Raman spectroscopy, are readily apparent to those skilled in the art and will not be fully described. Furthermore, if the meaning of technical terms described below differs from common usage in the art, the meaning in this text shall prevail. The accompanying drawings are intended to convey information related to the features of the present invention and are not drawn to their actual dimensions, which is hereby acknowledged.

[0028] In the present invention, the so-called semiconductor process parts refer to process parts used in semiconductor processes, especially process parts made of inorganic materials, including but not limited to: silicon (Si) rings used in etching processes; quartz (Quartz) parts used in etching processes or deposition processes, such as quartz furnace tubes, quartz crystal boats, quartz rings, quartz troughs, quartz heaters, etc.; alumina ceramic parts (Al2O3) used as semiconductor chamber process parts or wafer transfer process parts; and protective parts with resistance to plasma etching such as yttrium oxide (Y2O3) and yttrium aluminum garnet (YAG) used as internal coating parts of semiconductor devices or in semiconductor cavities.

[0029] Figure 1 This is a system architecture diagram showing a data processing system for a quality change detection system for semiconductor process parts according to an embodiment of the present invention. Figure 1In one embodiment, a data processing system 1 for a semiconductor process component quality change detection system receives detection data from a semiconductor process component quality change detection system 2, and includes an alignment operation unit 11, a derivative operation unit 12, a scale normalization unit 13, a dimension reduction operation unit 14, and a standard deviation operation unit 15. The alignment operation unit 11 is connected to the derivative operation unit 12, the derivative operation unit 12 is connected to the scale normalization unit 13, the scale normalization unit 13 is connected to the dimension reduction operation unit 14, and the standard deviation operation unit 15 is connected to the dimension reduction operation unit 14. The semiconductor processing component quality change detection system 2 comprises at least a Raman spectrometer 21, an optical detection unit 22 coupled to the Raman spectrometer 21, a control and calculation unit 23 coupled to the Raman spectrometer 21, and a Raman spectrum database unit 24 coupled to the control and calculation unit 23. The optical detection unit 22 is coupled to the Raman spectrometer 21, and both the Raman spectrometer 21 and the Raman spectrum database unit 24 are coupled to the control and calculation unit 23. In this embodiment, the semiconductor processing component quality change detection system 2 primarily utilizes the Raman spectrometer 21 and the optical detection unit 22 coupled to the Raman spectrometer 21 to perform quality change detection on the semiconductor processing component 3 to be tested. The Raman spectroscopy database unit 24 has the functions of receiving, outputting, and storing Raman spectroscopy data. For example, it is a memory, hard drive, or cloud system. It stores a plurality of Raman spectroscopy data for various semiconductor process components with known usage hours, known materials, known material compounds, or known material degradation states, as well as Raman spectroscopy data measured from the semiconductor process component 3 to be tested. The data processing system 1 for the semiconductor process component degradation detection system can not only receive Raman spectroscopy data from the Raman spectroscopy database unit 24 but also receive additional input Raman spectroscopy data. Ultimately, it generates degradation state simulation parameters 16 that can be used to predict the life time of the semiconductor process component 3 to be tested or a specific semiconductor process component, thereby pre-determining the degradation state of the semiconductor process component to be tested or the specific semiconductor process component. The qualitative change state simulation parameters 16 include at least two parameters: a life index and a discrete degree. The life index is used to indicate the degree of aging of the semiconductor process component 3 to be tested or a specific semiconductor process component in different external environments. The discrete degree is used to indicate the degree of molecular structure variation between different internal positions of the semiconductor process component 3 to be tested or a specific semiconductor process component.

[0030] Please continue to refer to Figure 1In one embodiment, the Raman spectrometer 21 emits a light beam, such as laser light, which passes through the optical detection unit 22 and is then projected onto a semiconductor process component 3 to be tested, causing the semiconductor process component 3 to emit scattered light, and then the Raman spectrum of the scattered light is measured. The optical detection unit 22 may have two optical fibers, one of which is used to guide the output light beam from the Raman spectrometer 21 through the light projection end 222 and project it onto the target area A of the semiconductor process component 3 to be tested, and the other optical fiber is used to guide the Raman scattered light (Raman Scattering Light) with a Raman spectrum excited from the target area A of the semiconductor process component 3 to be tested back to the Raman spectrometer 21 through the light projection end 222. The target area A here refers to the effective projection area where the output light beam from the Raman spectrometer 21 is projected onto the semiconductor process component 3 to be tested through the optical detection unit 22. In one embodiment, the projected light within this area is a circular light spot with a limited area.

[0031] Please continue to refer to Figure 1 In one embodiment, the Raman spectrum database unit 24 is coupled to the control operation unit 23. The Raman spectrum database unit 24 has stored therein a plurality of Raman spectrum data corresponding to a plurality of known usage hours of the semiconductor process component 3 to be tested, such as 50, 100, 150...500 hours, or a plurality of known materials, a plurality of known material compounds, or a plurality of known material transformation states or transformation parameters (such as a transformation percentage used to indicate the degree of transformation) thereof.

[0032] Figure 2 It is a plan view showing a semiconductor process component to be tested. Figure 1 The quality change detection system for semiconductor process parts is used to detect multiple target areas. Figure 1 and Figure 2 The size of the semiconductor process component 3 to be tested can be large or small without limitation. In one embodiment, multiple Figure 2 The validity of the test results is ensured by testing the dispersed target areas A1, A2, and A3 in the test area. The number of target areas is determined by the size of the test area 31. As the test area 31 increases, the number of dispersed target areas may need to be increased. The so-called test area 31 refers to the area on the semiconductor process component 3 to be tested that is most likely to undergo quality changes, such as the area that is usually subjected to plasma etching in the semiconductor process. Figure 2 The target areas A1 , A2 , and A3 shown in FIG. 1 can have the same or different sizes, and are preferably the same.

[0033] Please continue to refer to Figure 1In one embodiment, the alignment operation unit 11 of the data processing system 1 of the quality change detection system for semiconductor process parts is used to receive various Raman spectral data measured when the same semiconductor process part 3 to be tested is detected by different semiconductor process part quality change detection systems 2, including Raman spectral data of multiple target areas A1, A2, and A3 of the semiconductor process part 3 to be tested stored in the Raman spectral database unit 24, and perform all-dimensional error uniform operations on these Raman spectral data, such as a matrix operation, to eliminate the inconsistency of the detection results of the same semiconductor process part 3 to be tested by different semiconductor process part quality change detection systems 2. Since Raman spectral data is multidimensional data of at least two dimensions, that is, data composed of coordinate values ​​of at least two coordinate axes, such as the X-axis and the Y-axis, the alignment operation at least includes the alignment operation of the coordinate values ​​of the Raman spectral data on the X-axis and the Y-axis. Figure 1 and Figure 3 As shown, the original two-dimensional Raman spectrum data 100 with various errors is converted into two-dimensional Raman spectrum data 200 with the same error after alignment operation.

[0034] Please continue to refer to Figure 1 In one embodiment, the derivative operation unit 12 of the data processing system 1 of the quality change detection system for semiconductor process parts is used to receive the Raman spectrum data processed by the alignment operation unit 11, and perform a ratio operation on the differences between all spectral peaks in different dimensions on each received Raman spectrum data to obtain a first parameter 300 that can represent the change of all spectral peak intensity values ​​of each Raman spectrum data with the change of Raman spectrum offset. Since the Raman spectrum data is multidimensional data of at least two dimensions, that is, data composed of coordinate values ​​of at least two coordinate axes such as the X-axis and the Y-axis, the derivative operation at least includes a ratio operation on the differences between all spectral peak intensity values ​​of the Raman spectrum data on the X-axis and the Y-axis. Figure 4 As shown, in the Raman spectrum data C1, C2 and C3 of the three different target areas of the semiconductor process component 3 to be tested, the slope of the difference between the peak intensity value M and the peak intensity value N of the Raman spectrum data C1 is Δy1 / Δx1, which represents the ratio of the components of the difference between the peak intensity value M and the peak intensity value N on the Y-axis and the X-axis respectively; the slope of the difference between the peak intensity value N and the peak intensity value O of the Raman spectrum data C1 is Δy2 / Δx2, which represents the ratio of the components of the difference between the peak intensity value N and the peak intensity value O on the Y-axis and the X-axis respectively. Figure 4 In the X-axis, the physical quantity is Raman shift, and the unit is cm -1 , the physical quantity of the Y-axis is intensity, and its unit is au.

[0035] Please continue to refer to Figure 1 In one embodiment, the scaling and normalization unit 13 of the data processing system 1 for a quality change detection system for semiconductor process components receives the first parameter 300 obtained after processing by the derivative operation unit 12 and performs scaling and normalization on the received first parameter 300 to obtain a second parameter 400. Scaling refers to adjusting the scale of the first parameter 300 in each dimension to ensure that the data reading conforms to common practices, for example, scaling the scale of a dimension from 1 to 75 to 1 to 100. Normalization refers to removing duplicate and abnormal portions from the first parameter 300 to improve the usability of the second parameter 400. In one embodiment, the scaling and normalization performed by the scaling and normalization unit 13 on the first parameter 300 includes a matrix operation. In other embodiments, the scaling and normalization unit 13 may be placed between the dimensionality reduction operation unit 14 and the standard deviation operation unit 15, or after the standard deviation operation unit 15.

[0036] Please continue to refer to Figure 1 In one embodiment, the dimensionality reduction unit 14 of the data processing system 1 for a quality change detection system for semiconductor process components receives the second parameter 400 obtained after processing by the scale normalization unit 13 and performs a dimensionality reduction operation on the received second parameter 400 to reduce the complexity of the second parameter 400 and simplify subsequent operations, thereby obtaining a third parameter 500. In one embodiment, the dimensionality reduction operation performed is, for example, a principal component analysis (PCA) operation that reduces the dimensionality of the data signal into a combination of a specific basis and its corresponding independent variable (index). Because Raman spectral data is multidimensional data of at least two dimensions, and the second parameter 400 is also multidimensional data, the dimensionality reduction operation uses a matrix operation to convert the matrix data of the second parameter 400 into matrix data of lower dimensionality, such as one-dimensional matrix data.

[0037] Please continue to refer to Figure 1In one embodiment, the standard deviation calculation unit 15 of the data processing system 1 of the quality change detection system for semiconductor process parts is used to receive the third parameter 500 obtained after processing by the dimension reduction calculation unit 14, and perform standard deviation calculation on the received third parameter 500 to obtain two quality change state simulation parameters 16: life index and dispersion. The life index is used to represent the aging degree of the semiconductor process parts to be tested or specific in different external environments, and the dispersion is used to represent the degree of molecular structure variation between different internal positions of the semiconductor process parts to be tested or specific. Please refer to Figure 5 , which shows Raman spectra E1, E2, and E3 of semiconductor process components of the same specification in three different external environments simulated by a data processing system 1 of a quality change detection system for semiconductor process components according to an embodiment. The physical quantity on the X-axis is the Raman shift, with the unit being cm -1 , the physical quantity of the Y axis is intensity, and its unit is au. Figure 5 As shown, the peak intensity value of the simulated Raman spectrum data reflects the degree of aging of the semiconductor process parts. The lower the peak intensity value, the lower the aging degree. Therefore, the aging degree of the semiconductor process parts can be judged from the peak intensity value of the simulated Raman spectrum data, and the life index obtained by the standard deviation calculation unit 15 is related to the average value obtained after the standard deviation calculation, which represents the average value of the peak intensity of the simulated Raman spectrum data. In the present invention, the smaller the life index, the better, preferably below 90. On the other hand, please refer to Figure 6 , which shows Raman spectra data of the same semiconductor process component at three different target areas A1, A2, and A3 simulated by a data processing system 1 of a quality change detection system for semiconductor process components according to another embodiment. The physical quantity on the X-axis is the Raman shift, with the unit being cm -1 , the physical quantity of the Y axis is intensity, and its unit is au. Figure 6 As shown, the differences between the three peak intensity values ​​corresponding to each peak position X1, X2, and X3 of the simulated Raman spectrum data reflect the degree of molecular structure variation between different internal positions of semiconductor process components. Lower differences usually indicate lower degree of molecular structure variation. For example, Figure 6If the difference between the peak intensity values ​​of A1, A2, and A3 corresponding to position X1 in the graph and the difference between the peak intensity values ​​of A1, A2, and A3 corresponding to position X2 are consistent or not significantly different, it indicates that the molecular structure variation between different positions within the semiconductor process component is low. However, if the difference is large, it indicates that the molecular structure variation between different positions within the semiconductor process component is high, indicating a significant degree of qualitative change. Therefore, the degree of qualitative change of the semiconductor process component can be determined by the difference between the peak intensity values ​​corresponding to each peak position of the Raman spectrum data simulated after the standard deviation calculation. In other words, the dispersion obtained by the standard deviation calculation unit 15 is related to the discrete value obtained after the standard deviation calculation, indicating the difference in the peak intensity values ​​of the simulated Raman spectrum data. In the present invention, the smaller the dispersion, the better, and is preferably less than 1.5.

[0038] Therefore, if Figure 1 As shown, in addition to directly using the semiconductor process component quality change detection system 2 to detect quality changes in a semiconductor process component 3, the Raman spectral data obtained from the detection can also be input into the data processing system 1 of the semiconductor process component quality change detection system, via or without the Raman spectral database unit 24, to obtain quality change state simulation parameters 16 including life indicators and dispersion. The life indicators and dispersion can then be used to predict the life span of each semiconductor process component. For example, when the semiconductor process component 3 to be tested is new, due to the influence of the external environment, the degree of aging reflected is still low and insufficient to reflect the actual quality change state of the semiconductor process component 3 to be tested. The obtained dispersion can also be used to determine whether the component has undergone serious quality changes and is actually defective, thereby making a low estimate of its life span to avoid premature elimination and replacement of defective components. Furthermore, when the semiconductor process component 3 to be tested is an old product, and the degree of aging reflected by it is very high due to the influence of the external environment, which over-reflects the actual quality change state of the semiconductor process component 3 to be tested, the obtained discreteness can be used at the same time to judge whether only a slight quality change has occurred inside it and it is actually a valid product, and then a high evaluation prediction of the life time is made to avoid premature replacement of old products.

[0039] Figure 7 This is a flow chart showing a data processing method for a quality change detection system for semiconductor process parts according to an embodiment of the present invention. Figure 7As shown, in one embodiment, a data processing method for a quality change detection system for semiconductor process parts has the following steps: obtaining multiple Raman spectral data of a semiconductor process part; and performing multiple operations on the obtained Raman spectral data to obtain a first quality change state simulation parameter for representing the aging degree of the semiconductor process part in different external environments and a second quality change state simulation parameter for representing the degree of molecular structure variation between different internal positions of the semiconductor process part. The obtained Raman spectral data can be from Figure 1 The semiconductor process component quality change detection system 2 shown in FIG. 1 may detect data of a semiconductor process component 3 to be detected, or may have stored data in a storage medium such as FIG. Figure 1 The Raman spectroscopy database unit 24 includes multiple Raman spectral data sets of various semiconductor process components with known usage hours, known materials, known material compounds, or known material transformation states, or other input Raman spectral data of the semiconductor process components. The computational processing performed on the acquired Raman spectral data includes the following steps.

[0040] Step 701: Perform an alignment operation on the acquired Raman spectral data in at least two dimensions. In one embodiment, the alignment operation performed on the input Raman spectral data includes performing an error unification operation on all dimensions of the input Raman spectral data, including a matrix operation, to eliminate inconsistencies in the output results of different semiconductor process component quality change detection systems for the same semiconductor process component. Because Raman spectral data is multidimensional data with two or more dimensions, the alignment operation includes at least two dimensions of the input Raman spectral data: the X-axis and the Y-axis.

[0041] Step 702: Perform a derivative operation on the aligned Raman spectral data to obtain a first parameter. In one embodiment, the derivative operation is performed on each aligned Raman spectral data by performing a ratio operation on the differences between all spectral peaks in different dimensions to obtain a first parameter that represents how all spectral peak intensity values ​​of each Raman spectral data change with changes in Raman spectral shift. Because Raman spectral data is multidimensional data with two or more dimensions, the derivative operation at least includes performing a ratio operation on the differences between all spectral peak intensity values ​​of the aligned Raman spectral data on the X-axis and the Y-axis.

[0042] Step 703: Scaling and normalizing the first parameter obtained after the derivative operation to obtain the second parameter. In one embodiment, the scaling process involves adjusting the scale of each dimension of the first parameter to ensure that the data reading conforms to common practices, for example, scaling the scale of a dimension from 1 to 75 to 1 to 100. The normalization process involves removing duplicate and abnormal components from the first parameter to improve the usability of the second parameter. In one embodiment, the scaling and normalization process includes a matrix operation.

[0043] Step 704: Perform a dimensionality reduction operation on the second parameter obtained after scaling and normalization to obtain a third parameter. In one embodiment, the dimensionality reduction operation performed is a principal component analysis operation that reduces the second parameter to a combination of a specific basis and its corresponding independent variable. Because Raman spectral data is multidimensional data of at least two dimensions, the dimensionality reduction operation uses a matrix operation to convert the matrix data of the second parameter into a matrix data of a lower dimension, such as a one-dimensional matrix data.

[0044] Step 705: Perform a standard deviation calculation on the third parameter obtained after the dimensionality reduction operation to obtain a qualitative change state simulation parameter. In one embodiment, the qualitative change state simulation parameter includes a lifespan indicator and a dispersion. The lifespan indicator is used to indicate the degree of aging of the semiconductor process component under test or in different external environments, and the dispersion is used to indicate the degree of molecular structure variation between different internal locations of the semiconductor process component under test or in test. The lifespan indicator is related to the average value obtained after the standard deviation calculation and represents the average value of the peak intensity of the simulated Raman spectral data. The dispersion is related to the discrete value obtained after the standard deviation calculation and represents the degree of variability of the peak intensity values ​​of the simulated Raman spectral data.

[0045] Figure 8 This is a schematic plan view showing the relationship between the vital indicators, dispersion and the degree of quality change of semiconductor process parts obtained by the data processing system of the quality change detection system for semiconductor process parts according to one embodiment of the present invention. Figure 8 As shown, with dispersion D as the horizontal axis and life indicator L as the vertical axis, among the four regions on the DL plane, Region III has the lowest life indicator and dispersion, indicating minimal quality degradation and optimal use of semiconductor process components. Regions I, II, and IV have higher life indicators or dispersion, indicating significant quality degradation and potentially poor use. Based on this, the life span of each semiconductor process component can be predicted solely by interpreting the life indicator and dispersion, providing a pre-determined understanding of each component's quality degradation.

[0046] In summary, the data processing method and system for a semiconductor process component quality change detection system described in various embodiments of the present invention utilizes an algorithm, including alignment, derivative, scale normalization, dimensionality reduction, and standard deviation, on Raman spectral data to obtain life indicators and dispersions that can predict the lifespan of various semiconductor process components. Because the life indicators and dispersions are converted from one-dimensional Raman spectral data, the complexity of interpreting Raman spectral data is simplified. Furthermore, inspectors can assess the operational status of various semiconductor process components without interpreting Raman spectral data. Instead, they can predict the lifespan of various semiconductor process components solely based on the values ​​of the life indicators and dispersions, thereby improving the efficiency of component utilization.

[0047] The above detailed description has provided specific descriptions of some feasible embodiments of the present application. These embodiments are not intended to limit the patent scope of the present application. Any equivalent implementation or modification that does not depart from the technical spirit of the present application should be included in the patent scope of the present application.

Claims

1. A data processing method for detecting quality changes in semiconductor process parts, characterized in that: include: Acquiring multiple Raman spectral data of a semiconductor process component; and Performing multiple operations on the Raman spectrum data to obtain a first quality change state simulation parameter representing the degree of aging of the semiconductor process component and a second quality change state simulation parameter representing the degree of variation of the internal molecular structure of the semiconductor process component; The Raman spectral data is obtained when the semiconductor process component is detected by different semiconductor process component quality change detection systems, and the calculation processing includes: Performing an alignment operation in at least two dimensions on the obtained Raman spectral data to eliminate errors and inconsistencies in detection results of the semiconductor process component and the quality change detection system of different semiconductor process components; Performing a derivative operation on the Raman spectrum data after the alignment operation to perform a ratio operation on the differences between all spectral peaks of each Raman spectrum data in different dimensions to obtain a first parameter capable of representing how the intensity values ​​of all spectral peaks of each Raman spectrum data change with the change of Raman spectrum shift; performing scaling and normalization on the first parameter to obtain a second parameter; performing a dimensionality reduction operation on the second parameter to obtain a third parameter; and Performing a standard deviation operation on the third parameter to obtain an average value and a dispersion value; The first qualitative change state simulation parameter is related to the average value, and the second qualitative change state simulation parameter is related to the discrete value.

2. The data processing method according to claim 1, wherein: One of the alignment operation, the scaling and normalization process, and the dimensionality reduction operation includes a matrix operation.

3. A data processing system for detecting quality changes in semiconductor process parts, characterized in that: include: An alignment operation unit receives data from different semiconductor process component quality change detection systems for a semiconductor process component. Measured a plurality of Raman spectral data, and performing a unified error calculation of all dimensions on the Raman spectral data to eliminate the inconsistency of the errors in the detection results of the semiconductor process component and the quality change detection system of different semiconductor process components; a derivative operation unit electrically connected to the alignment operation unit, configured to receive the Raman spectrum data processed by the alignment operation unit, and perform a ratio operation on the differences between all spectral peaks in different dimensions on each of the received Raman spectrum data to obtain a first parameter representing how the intensity values ​​of all spectral peaks of each of the Raman spectrum data change with the change of the Raman spectrum shift; a scaling and normalization unit electrically connected to the derivative operation unit, configured to receive the first parameter and perform scaling and normalization processing on the received first parameter to obtain a second parameter; a dimensionality reduction operation unit electrically connected to the scale normalization unit, configured to receive the second parameter and perform a dimensionality reduction operation on the received second parameter to obtain a third parameter; and a standard deviation operation unit electrically connected to the dimensionality reduction operation unit, configured to receive the third parameter and perform a standard deviation operation on the received third parameter to obtain a first quality change state simulation parameter representing the degree of aging of the semiconductor process component and a second quality change state simulation parameter representing the degree of variation of the internal molecular structure of the semiconductor process component; The first qualitative change state simulation parameter is related to an average value obtained after the standard deviation calculation, and the second qualitative change state simulation parameter is related to a discrete value obtained after the standard deviation calculation.

4. The data processing system according to claim 3, wherein: The first qualitative change state simulation parameter is a life index of the semiconductor process component, and a smaller life index indicates a smaller aging degree of the semiconductor process component.

5. The data processing system according to claim 3, wherein: The second qualitative change state simulation parameter is the dispersion of the semiconductor process component, and the smaller the dispersion is, the smaller the degree of variation of the internal molecular structure of the semiconductor process component is.

Images