Production solution for high throughput / precision XPS metrology using unsupervised machine learning

Through unsupervised machine learning and principal component analysis technology, the problem of long acquisition time of XPS and XRF signals is solved, the accuracy and efficiency of layer thickness and component analysis in semiconductor processing are improved, and a faster and more efficient production process is achieved .

CN118901081APending Publication Date: 2024-11-05NOVA MEASURING INSTR LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202280092721.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2021-12-30
Filing Date
2022-12-30
Publication Date
2024-11-05

AI Technical Summary

Technical Problem

During semiconductor processing, X-ray photoelectron spectroscopy (XPS) and X-ray fluorescence spectroscopy (XRF) signal acquisition time is long, resulting in a decrease in production efficiency and profit margin, and the signal is weak, making it difficult to achieve a high signal-to-noise ratio data set with high signal-to-noise ratio .

Method used

Unsupervised machine learning and principal component analysis (PCA) technology are used to filter noise data by reducing half of signal acquisition time and improving data set quality, which is used to monitor layer thickness and component deviations during integrated circuit (IC) processing.

Benefits of technology

It achieves the accuracy and productivity of the analysis results while reducing signal acquisition time, improving productivity of 70% and 30% accuracy, while maintaining measurement accuracy under the same acquisition time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118901081A_ABST
    Figure CN118901081A_ABST
Patent Text Reader

Abstract

Processing deviations in semiconductor processing are determined using unsupervised machine learning of photoelectron emission datasets obtained by XPS or XRF tools. Principal component analysis is applied to the transmission data set, and the variance of each principal component is analyzed, thereby selecting N principal components with the highest variance. All data points of the dataset that do not correspond to any one of the N principal components are removed from the dataset to obtain a filtered dataset. The emission intensity is then calculated from the filtered dataset and drawn on an SPC chart to check for deviations.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Related Applications

[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 295,477, filed on December 30, 2021, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure generally relates to techniques for monitoring semiconductor processing during production by non-destructively measuring layer thickness and composition on structures using photoelectron spectroscopy and x-ray fluorescence. Background Art

[0004] Integrated circuits typically include multiple layers formed on a silicon substrate. As integrated circuits become smaller, and the thickness of the layers comprising the integrated circuit decreases, the performance of the device formed by these layers often depends on the thickness or composition of a particular layer. For example, a transistor formed on a silicon substrate may have different characteristics depending on the thickness or composition of the gate of the transistor. Therefore, during manufacturing, it is very beneficial to monitor the process by determining the thickness and composition of the layers in the manufactured microelectronic devices (such as integrated circuits).

[0005] Techniques that can be used to determine the thickness and / or composition of specific layers in a structure include X-ray photoelectron spectroscopy (XPS) and X-ray fluorescence spectroscopy (XRF). XPS has been used to analyze the surface chemistry of substrates, where a spectrum is obtained by irradiating the substrate with an X-ray beam while measuring the kinetic energy and the number of electrons escaping from the top layer of the substrate. Similarly, XRF has been used for elemental and chemical analysis of samples by sampling the emission of characteristic "secondary" (or fluorescent) X-rays from a material that has been excited by bombardment with high-energy X-rays or gamma rays.

[0006] For both XPS and XRF techniques, the emission is quite weak and in order to achieve sufficient signal-to-noise ratio to produce a meaningful data set, the collection period is relatively long. However, in a production process, any introduction of a monitoring element that requires a long signal acquisition period will slow down production and therefore directly affect the efficiency and profitability of the manufacturing process. Therefore, it would be very beneficial to develop techniques that reduce signal acquisition time without reducing the accuracy of the results. Summary of the invention

[0007] In order to provide a basic understanding of some aspects and features of the present invention, the following summary of the present invention is included. This summary is not a broad overview of the present invention, and therefore it is not intended to specifically identify key or important elements of the present invention or to describe the scope of the present invention. Its only purpose is to present some concepts of the present invention in a simplified form as a preface to the more detailed description presented below.

[0008] Various embodiments of the present invention provide methods and systems for improving signal acquisition in XPS and XRF systems. Embodiments of the present disclosure are particularly suitable for monitoring the consistency of processing during IC production. These embodiments enable the study of the properties of thin films stacked on substrates, such as composition and thickness. Embodiments of the present disclosure have been demonstrated to provide improved analysis results while reducing the standard signal acquisition time by half.

[0009] According to aspects of the invention, an XPS or XRF tool is used to collect emissions from a wafer in a conventional manner, except that the acquisition time is reduced, for example, by 50%. Therefore, the collected data set is quite noisy, with a relatively low signal-to-noise ratio. Therefore, the obtained data set is not suitable for standard analysis of process monitoring. Therefore, the quality of the data set is improved by the operation of unsupervised machine learning. An important aspect of unsupervised machine learning is its ability to separate differences in the data set, which contribute to the noise but do not contribute to the ability to identify deviations in the process. Therefore, unsupervised machine learning is used to reduce the differences in the spectrum caused by noise.

[0010] An embodiment of the present disclosure provides a method for monitoring process deviations in semiconductor processing of an integrated circuit (IC), the method comprising the steps of: irradiating the IC, thereby generating emissions from the IC; collecting the emissions from the IC using one of X-ray photoelectron spectroscopy (XPS) and X-ray fluorescence spectroscopy (XRF), and generating a data set from the emissions, the data set corresponding to counts of kinetic energy per unit time; performing principal component analysis on the data set, thereby obtaining principal component values ​​from the data set; selecting multiple principal components that exhibit high variance contributions compared to the remaining principal components; removing all values ​​corresponding to the remaining principal components from the data set, thereby obtaining a filtered data set; and analyzing the filtered data set to determine the presence of process deviations.

[0011] According to a further aspect, a method for enhanced statistical process control is provided, the method comprising the steps of obtaining a data set corresponding to photoelectron emission from a sample, the data set representing a spectrum of photoelectrons emitted from the sample; performing principal component analysis on the data set to obtain principal component variances for each principal component; examining the variances to select N relevant principal components having the highest variance values; selecting all relevant data points belonging to the N relevant principal components from the data set and using only these data points to update the spectra; calculating the photoelectron emission intensity using the updated spectrum; plotting the calculated photoelectron emission intensity on a statistical process control (SPC) chart; and examining the SPC chart to identify processing deviations.

[0012] According to a further aspect, a metrology module for monitoring a manufacturing process of an integrated circuit (IC) is provided. The metrology module includes: an input module that receives a signal indicating an emission from an IC (e.g., from an XPS or XRF metrology tool); a principal component analysis (PCA) module that receives the signal and calculates therefrom a variance corresponding to each principal component; a PCA number selector module that selects N principal components based on the calculated variances; a filter module that selects only data points corresponding to the selected N principal components from the original XPS signal; an intensity module that calculates an intensity value of the emission using the selected data points; a conversion module that converts the emission intensity into a material parameter; and a display module that displays the calculated parameter on a monitor, typically in the form of an entry in a PCA chart.

[0013] In other aspects, a machine-readable medium is provided, the machine-readable medium having an executable program stored thereon, which, when executed, causes the machine to perform a method for monitoring a semiconductor manufacturing process, the method comprising: receiving a data set of photoelectron counts corresponding to the kinetic energy of emitted photoelectrons per unit time, the data set being obtained by collecting emissions from an IC using X-ray photoelectron spectroscopy (XPS) and generating a data set from the emissions; performing principal component analysis on the data set to obtain principal component values ​​from the data set; selecting a plurality of principal components that exhibit high variance contributions compared to the remaining principal components; removing all values ​​corresponding to the remaining principal components from the data set to obtain a filtered data set; and analyzing the filtered data set to determine the presence of processing deviations.

[0014] Other aspects are disclosed through the detailed description with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present invention and, together with the description, are used to explain and illustrate the principles of the present invention. The accompanying drawings are intended to illustrate the main features of the exemplary embodiments in a diagrammatic manner. The accompanying drawings are not intended to describe every feature of an actual embodiment, nor are they intended to describe the relative sizes of the described elements, and are not drawn to scale.

[0016] One or more embodiments of the invention are illustrated by way of example and not limitation in the accompanying drawings in which like reference numerals indicate similar elements, and in which:

[0017] FIG. 1 is a diagram of a data set obtained using XPS measurement according to an embodiment of the present disclosure;

[0018] FIG. 2 illustrates a diagram of a principal component analysis (PCA) performed on the data set of FIG. 1 , according to one embodiment;

[0019] FIG3 is a plot of the principal component variances of FIG2 versus the logarithm of the number of principal components, according to one embodiment;

[0020] FIG4 shows a graph of a filtered data set obtained by applying PCA to XPS measurements according to an embodiment of the present disclosure;

[0021] FIG. 5A shows a graph of data obtained using full-time acquisition versus data obtained using half-time acquisition according to an embodiment of the present disclosure, and FIG. 5B shows a graph of data obtained using full-time acquisition versus interpolated data, which is data calculated by interpolating half of the full-time data points;

[0022] FIG6A shows the standard deviation of the boron percentage within a wafer calculated from the data sets acquired at full time (solid dots) and half time (circles) according to an embodiment of the present disclosure, while FIG6B shows the standard deviation of the boron percentage calculated from the data set acquired at full time (solid dots) and the filtered half time data set (open circles), i.e., the data set after removing the data corresponding to the unselected principal components in the half time data set;

[0023] FIG7 shows SPC charts for three different XPS tools, wherein the black dots represent the average boron percentage within the wafer calculated from data obtained with full-time acquisition, and the circles represent the average boron percentage within the wafer calculated from enhanced half-time data using embodiments described herein;

[0024] 8 is a flow chart of a method of performing enhanced statistical process control according to one embodiment;

[0025] FIG. 9 is a block diagram illustrating a metrology module for monitoring an integrated circuit manufacturing process according to one embodiment. DETAILED DESCRIPTION

[0026] The description of various aspects of the present invention will be carried out with reference to different embodiments. Each description of a particular embodiment may highlight specific features. However, it should be understood that the features described may also be combined in other embodiments, and different combinations of these features may be assembled to form further embodiments.

[0027] Embodiments of the present disclosure improve the operation and efficiency of XRF and / or XPS process monitors. For the purposes of this disclosure, it is assumed that the reader is fully familiar with the construction and operation of these monitors, which can be learned from, for example, Applicant's U.S. Patents 10,533,961 and 10,801,978, the disclosures of which are incorporated herein in their entirety for completeness.

[0028] As the structural complexity of semiconductor devices increases, the XPS or XRF signal obtained during monitoring becomes "noisier". This results in an increase in signal acquisition time in order to improve the signal-to-noise ratio. For example, an important parameter in the manufacture of SiGe:B raised source drain (RSD) transistors is the percentage of boron in the resulting layer. Another important parameter is the thickness of the TiN layer, such as in the gate electrode. In a manufacturing environment, not only do these parameters need to be measured, but they also need to be done as quickly and efficiently as possible so as not to slow down production. It has been demonstrated that embodiments of the present disclosure are able to cut acquisition time in half without reducing the accuracy of measuring the percentage of boron in SiGe:B structures. In addition, embodiments of the present disclosure have also been demonstrated to increase productivity by 70% when measuring TiN film thickness. Conversely, when utilizing the same acquisition time, these embodiments have been demonstrated to increase accuracy by 30%.

[0029] FIG. 1 is a diagram of a data set of boron concentration in a SiGe:B structure obtained using XPS measurements, showing counts per unit time for the number of bins (here, counts per second), where each bin corresponds to a certain binding energy level (typically in eV). These data can be used for machine training, for example, by performing a standard XPS acquisition at 13 sites on a wafer (e.g., 1.5 to 3 minutes of acquisition time), with 20 measurements at each site. As can be seen, these data are not easily interpreted to determine whether the resulting product meets specifications, or whether deviations occurred in the process that resulted in a change in the boron percentage. According to one embodiment, an unsupervised machine learning process has been implemented to remove data points that do not contribute to the analysis of whether deviations occurred in the manufacturing process.

[0030] Unsupervised machine learning methods are unique in this context because the machine is not provided with reference data that is labeled or scored. For example, in the context of boron concentration, the machine is not provided with reference data that indicates the characteristics of appropriate boron concentration. Instead, the machine needs to self-discover any naturally occurring patterns in the data set and determine which parameters are important for identifying deviations from the desired processing results (e.g., deviations from the desired boron concentration). In the following embodiments, principal component analysis (PCA) is used to implement unsupervised machine learning.

[0031] FIG2 shows a plot of PCA performed on the data set of FIG1 , showing the principal component values ​​versus the number of measurements. In this way, the original spectra are converted into spectra that are independent or orthogonal to each other. The principal component variances are then calculated for each point in the data set. FIG3 is a plot of the principal component variances versus the logarithm of the number of principal components. As indicated by the dashed ellipse, there are N (here four) principal components that contribute significantly to the variance, while the remaining principal components contain data that do not contribute substantially to the variance. The number of principal components, called a hyperparameter, can be set in advance for machine learning, or can be set automatically as part of a selection algorithm during the PCA process.

[0032] Once the principal components that contribute meaningfully to the variance are identified, the reverse PCA process is applied back to the energy space. For example, as shown in FIG4, in this embodiment, the data set plot shown in FIG1 is recreated, but only the data belonging to the N principal components that contribute meaningfully to the variance is used. It can be seen that the plot of FIG4 is "cleaner" than the plot of FIG1, even though it corresponds to the same data set. The data of FIG4 can now be analyzed to determine the quality of the SiGe:B deposition process.

[0033] One result of the embodiments of the present disclosure is that machine learning does not rely on long training cycles and does not require repeated measurements as a reference. Instead, by performing the steps outlined, the data of the selected N principal components becomes the reference itself. Any standard algorithm for determining composition or film thickness can be directly applied to a reconstructed data set that includes only data points from the selected N principal components. In fact, it has been demonstrated that this approach improves accuracy over applying a standard algorithm to the initial data set. In contrast, it has been shown that acquisition time can be reduced, and by adopting the process of the present disclosure, the same accuracy as the standard acquisition time can be achieved. Acquisition time can be reduced by reducing the time for signal acquisition or by taking fewer measurements at each site.

[0034] Figure 5A shows data obtained using full-time acquisition versus data obtained using half-time acquisition (in this case, using half the number of repetitions for each point). Figure 5B shows data obtained using full-time acquisition versus interpolated data obtained using full-time interpolated acquisition using half the number of points from the spectrum.

[0035] Figure 6A shows the standard deviation of the percent boron calculated using the full-time (solid dots) and half-time (open circles) data sets. Figure 6B shows the standard deviation of the percent boron calculated using the full-time (solid dots) and filtered half-time (open circles) data sets (i.e., after removing the data corresponding to the unselected principal components in the half-time data set). As shown, the fit in Figure 6B is tighter after the PCA process is applied to the half-time data set shown in Figure 6A.

[0036] In a production environment, in one embodiment, the process is performed by obtaining XPS spectra from a wafer under inspection. An unsupervised machine learning process then applies PCA to the spectral data. Next, the N principal components that produce the highest variation are selected and the spectral curve is regenerated using the data corresponding to these N principal components. The process then continues by determining the intensity by calculating the area under the curve. The intensity is then converted to an expected value (e.g., boron concentration, TiN thickness, etc.) using a predetermined relationship (i.e., the intensity value is related to the material property being inspected). For example, different boron concentrations in a layer will result in different emission intensities. The intensity or corresponding material property values ​​can be plotted in an SPC (Statistical Process Control) chart to monitor the manufacturing process.

[0037] Figure 7 shows an SPC chart of the boron percentage obtained from three different XPS tools, where the black dots represent the boron concentration calculated using the full-time acquisition data, and the circles represent the boron concentration calculated using the enhanced half-time data in an embodiment of the present invention. It can be seen that the consistency of the data is good, indicating that the process is within the allowable variance range.

[0038] Thus, aspects of the present disclosure achieve enhanced statistical process control by performing the method outlined in FIG. 8 . The process of FIG. 8 begins at step 800 by obtaining a data set corresponding to photoelectron emission from a sample. The data set represents a spectrum of photoelectrons emitted from the sample. At step 805 , the process obtains the principal component variances for each principal component by applying principal component analysis to the data set. At step 810 , the variances are examined to select the N relevant principal components with the highest variance values. This can be done by plotting descending or ascending curves of all variances and detecting inflection points. All variances with values ​​above the inflection point are considered to belong to the relevant principal components and are selected as part of the N principal components. At step 815 , the process continues by restoring to energy space. This is achieved by selecting all data points belonging to the N relevant principal components from the data set and redrawing the spectrum using only these data points. At step 820 , the photoelectron emission intensity is calculated using the updated redrawn spectrum, i.e., only the relevant data points are used for calculation. At step 825 , the calculated photoelectron emission intensity is converted into a layer property value by a known relationship. For example, boron percentage or TiN thickness is related to emission intensity, and this known relationship is used to convert the intensity into a material property value. At step 830, the material property value is plotted on a statistical process control (SPC) chart. Finally, at step 835, the SPC chart is examined to identify process deviations.

[0039] 9 is a block diagram illustrating a metrology module for monitoring an integrated circuit (IC) manufacturing process. As shown, the module includes an input module 900 that receives a signal indicative of photoelectrons emitted from an IC (e.g., from an XPS metrology tool). In this regard, the entire metrology module can be integrated into an XPS or XRF metrology tool, such as implemented as software, hardware, or a combination of both, or can be implemented as part of a controller of a metrology tool, such as a general purpose computer or a specially customized computer.

[0040] The input module 900 passes the signal to the principal component analysis (PCA) module 905. The PCA module 905 receives the signal and calculates the variance of each principal component. The PCA number selector module 910 selects N principal components based on the calculated variance. The filter module 915 selects only the data points corresponding to the selected N principal components from the original XPS signal. The intensity module 920 uses the selected data points to calculate the intensity value of the photoelectron emission, that is, calculates the area under the curve of the selected data points. The conversion module 925 applies a known relationship to convert the calculated intensity into a material property value (e.g., a concentration percentage or a layer thickness). The display module 930 displays the material property values ​​on a monitor, usually in the form of entries in a PCA chart.

[0041] Thus, a method for monitoring process deviations in semiconductor processing of integrated circuits (ICs) is provided, comprising the steps of: irradiating the IC to generate emissions from the IC; collecting the emissions from the IC using one of X-ray photoelectron spectroscopy (XPS) or X-ray fluorescence spectroscopy (XRF), and generating a data set from the emissions corresponding to photon counts per unit time for the kinetic energy of the emitted photons; performing principal component analysis on the data set to obtain principal component values ​​from the data set; selecting N principal components that exhibit high variance contributions compared to the remaining principal components; removing all values ​​corresponding to the remaining principal components from the data set to obtain a filtered data set; analyzing the filtered data set to determine the presence of process deviations. The number N may be pre-set, or selecting the number of N principal components may include calculating the variance of each principal component and plotting the variance against the corresponding principal component to identify an inflection point; and setting N to the number of principal components corresponding to the higher variance from the inflection point. Analyzing the filtered data set may include calculating emission intensities from the filtered data set, converting the emission intensities into material property values, and plotting the material property values ​​on a statistical process control chart.

[0042] In addition, a metrology module for monitoring an integrated circuit (IC) manufacturing process is enabled, the module comprising: an input module that receives a signal indicating photoelectrons emitted from the IC; a principal component analysis (PCA) module that receives the signal and calculates the variance of each principal component; a PCA quantity selector that selects N principal components based on the calculated variance; a filtering module that selects data points corresponding to the N principal components from the signal; an intensity module that calculates emission intensity from the output of the filtering module; a conversion module that converts the emission intensity into a material property value; and a display module that plots the material property value. The input port can be coupled to an X-ray photoelectron spectroscopy (XPS) sensor or an X-ray fluorescence spectroscopy (XRF) sensor. The signal can correspond to the thickness or chemical composition of the sample.

[0043] In addition, a machine-readable medium is provided having executable program code stored thereon, which, when executed, causes a machine to perform a method for determining an anomaly in a property of a fabricated layer of a sample, the method comprising: collecting photoelectron emissions from the sample using one of X-ray photoelectron spectroscopy (XPS) or X-ray fluorescence spectroscopy (XRF), and generating a data set from the emissions corresponding to photoelectron counts per unit time for kinetic energy of emitted photons; performing principal component analysis on the data set to obtain principal component values ​​from the data set; selecting a plurality of principal components that exhibit high variance contributions compared to remaining principal components; removing all values ​​corresponding to the remaining principal components from the data set to obtain a filtered data set; and analyzing the filtered data set to determine the presence of a process deviation. Analyzing the filtered data set may include calculating a total intensity of the photoelectron emissions using the filtered data set, converting the intensity into a material property value, and plotting the material property value on an SPC chart.

[0044] It should be understood that the processes and techniques described herein are not inherently related to any particular apparatus and may be implemented by any suitable combination of components. In addition, various types of general purpose equipment may be used according to the guidance described herein. It may also prove advantageous to construct specialized apparatus to perform the method steps described herein.

[0045] The present invention has been described by specific examples, which are intended to be illustrative and non-restrictive in all respects. It will be appreciated by those skilled in the art that many different combinations of hardware, software, and firmware will be suitable for implementing the present invention. In addition, by considering this specification and practicing the present invention, other implementations of the present invention will be apparent to those skilled in the art. The specification and embodiments are intended to be considered exemplary, and the true scope and spirit of the present invention are indicated by the following claims.

Claims

1. A computerized method for monitoring process deviations in semiconductor processing of an integrated circuit IC, comprising the following steps: illuminating the IC, thereby generating emissions from the IC; collecting the emission from the IC using one of X-ray photoelectron spectroscopy (XPS) and X-ray fluorescence spectroscopy (XRF), and generating a data set from the emission, the data set corresponding to counts of kinetic energy per unit time; performing principal component analysis on the data set to obtain principal component values ​​from the data set; Selecting multiple principal components that exhibit high variance contribution compared to the remaining principal components; removing from the data set all values ​​corresponding to the remaining principal components, thereby obtaining a filtered data set; as well as The filtered data set is analyzed to determine the presence of process deviations.

2. The method according to claim 1, wherein: Performing principal component analysis also includes calculating the principal component variance for each principal component value in the data set.

3. The method according to claim 2, wherein: Selecting a plurality of principal components includes plotting the calculated principal component variances in descending or ascending order and detecting an inflection point, and selecting all principal component variances having values ​​higher than the inflection point.

4. The method according to claim 2, wherein: Obtaining the filtered data set further comprises selecting from the emission values ​​of the data corresponding to counts of kinetic energy per unit time only those belonging to the N principal components that contribute meaningfully to the variance. 5 . The method of claim 4 , further comprising plotting a spectral curve of the filtered data set and determining emission intensity by calculating an area under the spectral curve. The method of claim 5 , further comprising converting the emission intensity into a material property value of the IC.

7. The method of claim 6, further comprising plotting the material property values ​​in a statistical process control chart.

8. The method according to claim 7, wherein: The material property value includes one of material concentration and thickness.

9. A metrology module for monitoring a manufacturing process of an integrated circuit IC, the metrology module comprising: an input module that receives a signal indicative of an emission from the IC; A principal component analysis (PCA) module receives the signal and calculates the variance corresponding to each principal component; PCA number selector, selects N principal components based on the calculated variance; A filtering module, selecting from signal data points corresponding to the N principal components; an intensity module that calculates the emission intensity of the output from the filtering module; A conversion module, converting the emission intensity into a material property value; as well as A display module is used to draw the material property value.

10. The metering module according to claim 9, wherein: The signal includes one of an X-ray photoelectron spectroscopy (XPS) signal and an X-ray fluorescence spectroscopy (XRF) signal.

11. The metering module according to claim 9, wherein: The input module is coupled to an X-ray photoelectron spectroscopy XPS sensor or an X-ray fluorescence spectroscopy XRF sensor.

12. The metering module according to claim 9, wherein: The PCA Quantity Selector plots the principal component variances in descending or ascending order to determine the inflection point, and selects all principal component variances having values ​​above the inflection point.

13. The metering module according to claim 9, wherein: The intensity module plots a spectral curve of the data points selected by the filter module and determines the emission intensity by calculating the area under the spectral curve.

14. The metering module according to claim 9, wherein: The conversion module converts the emission intensity into one of a material concentration and a layer thickness.

15. A machine-readable medium having executable program code stored thereon, which, when executed, causes a machine to perform a process for determining an anomaly in a property of a fabricated layer of a sample, the process comprising: collecting photoelectron emission from the sample and generating from the emission a data set of photoelectron counts corresponding to kinetic energies of emitted photoelectrons per unit time; performing principal component analysis on the data set to obtain principal component values ​​from the data set; Obtain N principal components; Removing all values ​​that do not belong to the N principal components from the data set, thereby obtaining a filtered data set; as well as The filtered data set is analyzed to determine the presence of process deviations.

16. The machine-readable medium of claim 15, wherein: Collecting the photoelectron emission includes using one of X-ray photoelectron spectroscopy XPS and X-ray fluorescence spectroscopy XRF.

17. The machine-readable medium of claim 15, wherein: The process also includes calculating the principal component variance for each principal component value in the data set.

18. The machine-readable medium of claim 15, wherein: Obtaining N principal components includes selecting principal components that exhibit high variance contribution compared to the remaining principal components.

19. The machine-readable medium of claim 15, wherein: Obtaining N principal components includes selecting a preset number N.

20. The machine-readable medium of claim 15, wherein: Analyzing the filtered data set includes calculating a total intensity of photoelectron emissions using the filtered data set, converting the intensity into material property values, and plotting the material property values ​​on a statistical process control chart.