Method for detecting abnormality in semiconductor manufacturing process and program for performing same

Abstract

The present invention provides an abnormality detection method for detecting an abnormality in a semiconductor manufacturing process by analyzing light in a process chamber for processing a substrate by using plasma. The method may comprise: a cluster classification step of clustering reference light intensity data, which varies according to the type of an object to be removed from a substrate, into a plurality of clusters; a reference data generation step of generating reference data including the distribution of abnormality determination values for each of the plurality of clusters; and an abnormality detection step of comparing target data for detecting the abnormality with the reference data to determine whether the target data is abnormal.

Description

Method for detecting anomalies in a semiconductor manufacturing process and a program for performing the same

[0001] The present invention relates to a method for detecting anomalies in a semiconductor manufacturing process and a program for performing the same.

[0002] Semiconductor manufacturing processes include plasma-based etching. This process removes specific layers from the wafer surface. An Optical Emission Spectroscopy (OES) analyzer is used to accurately detect the end point of the etching process. The OES analyzer is based on specific OES wavelengths generated from the decomposition or reaction products of the etching gas according to the wafer process recipe. This allows for the determination of the end point detection (EPD) time of the process.

[0003] However, background knowledge of the element is required to select the appropriate OES wavelength for each element in OES analysis. Each element emits a different wavelength, and accurate analysis is difficult if these are not properly matched. Furthermore, since the precision of wavelengths varies among OES measurement instruments, the measured wavelengths can differ by decimal points. For this reason, it is extremely difficult to match and verify every single wavelength individually.

[0004] Furthermore, even with the same process recipe, the OES response pattern varies depending on the composition of the wafer layers. For example, the OES signal generated when a specific layer on the wafer is etched varies depending on the material of the layer. If anomalies are detected without considering these differences or classifying them according to wafer layers, there is a high probability of misclassification. In particular, such errors can occur when detecting ashing rate defects.

[0005] Furthermore, to detect process anomalies more effectively, it is common practice to use additional monitoring means, such as sensors, in addition to the OES. For example, sensors monitoring temperature or pressure within the chamber can be added to monitor environmental changes. However, such additional equipment incurs high installation costs and increases process complexity. This reduces system efficiency and presents a problem of being unreasonable in terms of cost.

[0006] One objective of the present invention is to provide an anomaly detection method for a semiconductor manufacturing process and a program for performing the same, which can calculate the end point detection time and the end point detection area even without knowing the wavelength information of the light generated when a layer to be removed is removed by plasma.

[0007] In addition, the present invention has the objective of providing a method for detecting abnormalities in a semiconductor manufacturing process and a program for performing the same, which can detect whether there are abnormalities in a semiconductor manufacturing process performed within a process chamber using an optical analyzer.

[0008] The problems that the present invention aims to solve are not limited to those described above, and problems not mentioned will be clearly understood by those skilled in the art from this specification and the attached drawings.

[0009] The present invention provides an anomaly detection method for detecting anomalies in a semiconductor manufacturing process by analyzing light within a process chamber that processes a substrate using plasma. The method may include: a clustering step of clustering reference light intensity data, which vary depending on the type of object to be removed from the substrate, into a plurality of clusters; a reference data generation step of generating reference data including a distribution of anomaly judgment values ​​for each of the plurality of clusters; and an anomaly detection step of determining whether the target data is anomaly by comparing the target data for detecting the anomaly with the reference data.

[0010] According to one embodiment, the anomaly detection step may include: a cluster determination step for determining whether the target data belongs to a cluster among the clusters distinguished in the cluster distinction step; an anomaly determination value calculation step for calculating the anomaly determination values ​​of the target data; and an anomaly determination step for determining whether there is an anomaly in the semiconductor manufacturing process when collecting the target data by comparing the anomaly determination values ​​of the target data with the anomaly determination values ​​of the reference data.

[0011] According to one embodiment, the above abnormality judgment values ​​may include: End Point time; and End Point area.

[0012] According to one embodiment, the method may further include a wavelength selection step of selecting at least one wavelength satisfying a reference condition among the collectible wavelengths collectible within the process chamber as selected wavelengths.

[0013] According to one embodiment, the method may further include a selected wavelength data collection step for collecting reference light intensity data collected when processing a substrate with respect to the selected wavelength.

[0014] According to one embodiment, the reference condition measures the Wasserstein distance between a first section and a second section different from the first section in the light intensity data for each of the collectible wavelengths, lists them in order of increasing Wasserstein distance, and then selects the up to the nth collectible wavelength as the selected wavelength by prioritizing the wavelength with the largest Wasserstein distance.

[0015] According to one embodiment, the clustering method used in the clustering step may be a K-Means method or a Hierarchical method.

[0016] In addition, the present invention provides a method for detecting anomalies in a semiconductor manufacturing process. The method determines whether there is an anomaly in the process by comparing pre-stored reference data with target data for detecting anomalies, wherein the reference data may include distribution data regarding endpoint time and endpoint area.

[0017] According to one embodiment, the reference data can be derived by clustering reference light intensity data, which vary depending on the type of object to be removed from the substrate, into a plurality of clusters, and calculating the endpoint time and endpoint area of ​​the reference light intensity data included in each cluster.

[0018] In addition, the present invention provides a program stored on a recording medium for detecting abnormalities in a semiconductor manufacturing process by analyzing light within a process chamber that processes a substrate using plasma. The program performs an abnormality detection operation by comparing pre-stored reference data with target data for detecting abnormalities to determine whether there is an abnormality in the process, wherein the reference data may include distribution data regarding endpoint time and endpoint area.

[0019] According to one embodiment, the program may further perform a wavelength selection operation of selecting at least one wavelength satisfying a reference condition among the collectible wavelengths collectible within the process chamber as selected wavelengths.

[0020] According to one embodiment, the program may further perform: a clustering operation for clustering reference light intensity data for the selected wavelength into a plurality of clusters; and a reference data generation operation for each of the plurality of clusters for generating reference data including a distribution of abnormality judgment values.

[0021] According to one embodiment, the anomaly detection operation may include: a cluster determination operation for determining whether the target data belongs to which cluster among the clusters distinguished in the cluster distinction operation; and an anomaly determination numerical calculation operation for calculating the endpoint time and the endpoint area of ​​the target data.

[0022] According to one embodiment, the anomaly detection operation may further include an anomaly determination operation that determines whether there is an anomaly in the semiconductor manufacturing process when collecting the target data by comparing the endpoint time and the endpoint area of ​​the target data with the endpoint time and the endpoint area of ​​the reference data.

[0023] According to one embodiment of the present invention, when a layer to be removed is removed by plasma, the end point detection time and the end point detection area can be calculated even without knowing the wavelength information of the light generated.

[0024] In addition, according to one embodiment of the present invention, an abnormality in a semiconductor manufacturing process performed in a process chamber can be detected using an optical analyzer.

[0025] The effects of the present invention are not limited to the effects described above, and unmentioned effects will be clearly understood by those skilled in the art from this specification and the attached drawings.

[0026] FIG. 1 is a diagram schematically illustrating a substrate processing apparatus according to one embodiment of the present invention.

[0027] FIG. 2 is a flowchart for explaining an anomaly detection method according to one embodiment of the present invention.

[0028] Figure 3 is a diagram illustrating wavelength candidates that can be selected as the selected wavelength in the wavelength selection step of Figure 2.

[0029] FIG. 4 is a diagram illustrating a method for selecting a selected wavelength among wavelength candidates in the wavelength selection step of FIG. 2.

[0030] Figure 5 is an example of light intensity data of a wavelength that can be selected as a selected wavelength in the wavelength selection step of Figure 2.

[0031] Figure 6 is an example of light intensity data of a wavelength that is difficult to select as a selected wavelength in the wavelength selection step of Figure 2.

[0032] Figure 7 is a graph illustrating light intensity data that can be collected when processing multiple substrates with the same process recipe.

[0033] Figure 8 is a graph showing clustered light intensity data that can be collected when processing multiple substrates with the same process recipe.

[0034] Figure 9 is a distribution graph showing the endpoint time and endpoint area distributions generated for each cluster.

[0035] Figure 10 is a graph illustrating the endpoint time and endpoint area distribution of Figure 9.

[0036] Figure 11 is a flowchart for explaining the anomaly detection step of Figure 2.

[0037] The various features and benefits of the non-limiting embodiments of this specification may become more apparent from a review of the detailed description in conjunction with the accompanying drawings. The accompanying drawings are provided for illustrative purposes only and should not be construed as limiting the claims. Unless expressly stated otherwise, the accompanying drawings are not to be drawn to scale. For clarity, various dimensions in the drawings may be exaggerated.

[0038] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. Exemplary embodiments are provided to ensure that the present disclosure is thorough and will fully convey its scope to those skilled in the art. To provide a complete understanding of the embodiments of the present disclosure, many specific details, such as examples of specific components, devices, and methods, are presented. It will be apparent to those skilled in the art that specific details are not necessary, that exemplary embodiments may be implemented in many different forms, and that neither should be interpreted as limiting the scope of the present disclosure. In some exemplary embodiments, known processes, known device structures, and known technologies are not described in detail.

[0039] The terms used herein are merely for describing specific exemplary embodiments and are not intended to limit exemplary embodiments. Singular expressions or expressions where singularity is not specified, as used herein, are intended to include plural expressions unless the context clearly indicates otherwise. The terms “comprising,” “comprising,” “having,” and “having” are open-ended and thus specify the presence of the mentioned features, components, steps, operations, elements, and / or components, and do not exclude the presence or addition of one or more other features, components, steps, operations, elements, components, and / or groups thereof. Method steps, processes, and operations in this specification are not to be interpreted as necessarily being performed in the specific order discussed or described unless the order of performance is specified. Additionally, additional or alternative steps may be selected.

[0040] When an element or layer is referred to as being "on," "connected," "combined," "attached," "adjacent," or "covering" another element or layer, it may be directly on, connected to, combined with, attached to, adjacent to, or covering said other element or layer, or intermediate elements or layers may exist. Conversely, when an element is referred to as being "directly on," "directly connected to," or "directly combined" with another element or layer, it should be understood that intermediate elements or layers do not exist. Throughout the specification, the same reference numerals refer to the same elements. The term "and / or" as used in the present invention includes all combinations and non-combinations of one or more of the listed items.

[0041] Although terms such as first, second, third, etc., may be used to describe various elements, regions, layers, and / or sections in the present invention, it should be understood that these elements, regions, layers, and / or sections are not limited by these terms. These terms are used merely to distinguish one element, region, layer, or section from another element, region, layer, or section. Accordingly, the first element, first region, first layer, or first section discussed below may be referred to as the second element, second region, second layer, or second section without departing from the teachings of the exemplary embodiments.

[0042] Spatially relative terms (e.g., "below," "under," "lower," "above," "top," etc.) may be used for convenience of explanation to describe the relationship between one element or feature and another element(s) or feature(s) as illustrated in the drawings. It should be understood that spatially relative terms are intended to include not only the orientations illustrated in the drawings but also other orientations of the device in use or operation. For example, if the device in the drawings is inverted, elements described as "below" or "under" other elements or features will be oriented "above" other elements or features. Thus, the term "below" may include both upper and lower orientations. The device may be oriented differently (rotated 90 degrees or in a different orientation), and the spatially relative descriptive terms used in the present invention may be interpreted accordingly.

[0043] It should be understood that there may be some inaccuracy when the terms "identical" or "same" are used in the description of the embodiments. Therefore, if one element or value is referred to as identical to another element or value, it should be understood that said element or value is identical to another element or value within a manufacturing or operating tolerance (e.g., ±10%).

[0044] Where the words “approximately” or “substantially” are used in this specification with respect to figures, it should be understood that such figures include a manufacturing or operational tolerance (e.g., ±10%) of the figures mentioned. Additionally, where the words “generally” and “substantially” are used with respect to geometric forms, it should be understood that while geometric accuracy is not required, freedom of form (latitude) is within the scope of disclosure.

[0045] Unless otherwise defined, all terms used in the present invention (including technical and scientific terms) have the same meaning as generally understood by those skilled in the art to which the exemplary embodiments belong. Furthermore, terms, including those defined in commonly used dictionaries, should be interpreted as having a meaning consistent with that meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the present invention.

[0046] FIG. 1 is a diagram schematically illustrating a substrate processing apparatus according to one embodiment of the present invention.

[0047] The substrate processing device (1) according to the present invention includes a process chamber (10), a light sensor (20), a light analyzer (30), and a controller (40). Through these components, the substrate (W) is processed using plasma, and abnormalities in the process can be detected based on data generated during the process.

[0048] The process chamber (10) is a chamber capable of processing a substrate, such as a wafer (W), using plasma. Plasma is an ionized gas in a high-energy state and can be used for operations such as etching, deposition, and cleaning in semiconductor manufacturing processes.

[0049] Plasma processes can be achieved by applying electrical energy to a gas to ionize it. When the gas is ionized, electrons and ions are generated, and these high-energy particles collide with the surface of a substrate to remove specific materials or deposit new materials. In the etching process, unnecessary layers can be removed, and in the deposition process, new layers can be formed on the surface of the substrate. Additionally, in the cleaning process, residues or impurities remaining on the surface of the substrate can be effectively removed.

[0050] The process chamber (10) may have a power supply that generates plasma by supplying power. The power supply may use high-frequency (RF) power or microwave power to transfer energy to a gas and ionize it. The density and stability of the plasma can be controlled through this power supply. RF power generated by the power supply may be applied into the process chamber (10) through components such as an antenna or a counter electrode, thereby effectively ionizing the gas to form plasma. The power supply, the antenna, or the power supply and the counter electrode may be defined as a plasma source.

[0051] Additionally, the process chamber (10) may include a gas supply system that supplies process gas necessary to form plasma. Gases used in the etching process include CF₄, SF₆, O₂, etc., and gases such as SiH₄ or NH₃ may be used in the deposition process.

[0052] The process chamber (10) may also include a pressure control system for controlling the pressure inside the chamber. Since the plasma can be stably maintained under specific pressure conditions, the plasma can be properly formed during the process through a pressure control system including a pump, and the desired process can be consistently performed.

[0053] Additionally, the process chamber (10) may also include a substrate holder that accurately controls the position of the substrate. The substrate holder can fix the substrate during the process and ensure uniform plasma treatment and stable process results.

[0054] In addition, the process chamber (10) may include various components necessary to perform a process of treating a substrate using plasma. For example, a magnetic field generating device to ensure uniformity of plasma generation may be added.

[0055] The optical sensor (20) can detect light generated inside the process chamber through a view port provided on the side wall of the process chamber (10). Light emitted during the plasma process is important information indicating the process state, and the optical sensor can detect this light in real time to collect data necessary for process monitoring.

[0056] The detected light can be transmitted to a light analyzer (30) via an optical cable. The optical sensor (20) can convert the detected light signal into an analog or digital signal and transmit it to the light analyzer (30). These signals can be processed by the light analyzer (30) to provide important data for analyzing the process status or determining whether there is an abnormality.

[0057] Various technologies can be used for the optical sensor (20). Photodiode sensors can convert light into electrical signals based on semiconductor materials and are characterized by high-speed response. Charge-Coupled Device (CCD) sensors can capture high-resolution optical images, enabling precise optical analysis. Complementary Metal-Oxide-Semiconductor (CMOS) sensors can provide low power consumption and high integration. Optical fiber sensors can transmit optical signals stably even in high temperatures and harsh environments.

[0058] The optical analyzer (30) is a device that analyzes the optical emission spectrum emitted during a plasma process within a process chamber (10) and is based on Optical Emission Spectroscopy (OES) technology. Plasma emits light of various wavelengths in a high-energy state, and this light can provide important information regarding the state of the process and chemical reactions. The optical analyzer (30) analyzes these optical signals to monitor the state of the process in real time or detect abnormalities.

[0059] The light analyzer (30) receives light emitted from the plasma collected through the light sensor (20) and can analyze the wavelength, intensity, distribution, etc. of the light. Each wavelength corresponds to a specific element or molecule, and through this, it is possible to determine what reaction is taking place in the current process and what chemical components are being generated or removed. For example, if light of a specific wavelength is detected in the plasma, this may be a signal that a specific gas or substance is reacting during the process.

[0060] OES plays an important role in End Point Detection (EPD) in the semiconductor manufacturing process. The optical analyzer (30) can determine whether the process has reached a desired level by monitoring the light emitted while a specific layer of the substrate (W) is being removed. As the layer is removed during the etching process, the spectrum of the light changes, and by analyzing this change, the accurate endpoint can be detected.

[0061] The optical analyzer (30) provides various spectral resolutions and analysis ranges, allowing it to detect and analyze a wide range of wavelengths. This enables the detection of even minute changes occurring in the plasma and allows for the real-time detection of small fluctuations or abnormalities in the process. The results of this analysis are transmitted to the controller (40), which can automatically adjust the process state or enable immediate response when an abnormality occurs during the process.

[0062] The optical analyzer (30) can also be configured to suit various substrates and process conditions. For example, since the response of each wavelength may differ depending on the layer characteristics of a specific substrate, reference data for each wavelength used during the process can be pre-set, and based on this, it can be evaluated whether the process is proceeding normally. Through such settings, customized analysis according to the substrate material or process gas is possible.

[0063] In addition, the optical analyzer (30) is equipped with a high-speed analysis function, so it can respond immediately to rapid changes in the plasma process. In particular, it can contribute to increasing the accuracy of the process by performing precise wavelength analysis even in complex process environments.

[0064] The controller (40) may include memory, a processor, a display, an interface unit, and a bus.

[0065] Various components such as memory, processor, display, and interface unit can be connected and communicate with each other (i.e., transmit control messages and data) via the bus.

[0066] Memory may include volatile memory (e.g., DRAM, SRAM, or SDRAM) and / or non-volatile memory (e.g., OTPROM (one-time programmable ROM), PROM, EPROM, EEPROM, mask ROM, flash ROM, flash memory, PRAM, RRAM, MRAM, hard drive, or solid-state drive (SSD)). Memory may include internal memory and / or external memory. Memory may store instructions or data related to at least one other component of the electronic device, for example. Additionally, memory may store software and / or programs. Programs may include, for example, a kernel, middleware, an application programming interface (API), and / or an application program (or "application"). At least part of the kernel, middleware, or API may be referred to as an operating system.

[0067] Additionally, the controller (40) may be provided with a non-transitory computer-readable medium. A non-transitory computer-readable medium refers to a medium that stores data semi-permanently and is readable by a computer, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, the various applications or programs described above may be provided by being stored on a non-transitory computer-readable medium such as a CD, DVD, hard disk, Blu-ray disk, USB, memory card, or ROM. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0068] The processor may include one or more of a central processing unit, an application processor, or a communication processor (CP). The processor may perform operations or data processing regarding the control and / or communication of at least one other component of a computing device or a non-transient computer-readable medium, for example.

[0069] The display may include, for example, a liquid crystal display (LCD), a light-emitting diode (LED) display, an organic light-emitting diode (OLED) display, a micro-electromechanical system (MEMS) display, or an electronic paper display. The display may display various content (e.g., text, images, videos, icons, and / or symbols, etc.) to a user. The display may include a touch screen and may receive touch, gesture, proximity, or hovering input using, for example, an electronic pen or a part of the user's body.

[0070] The interface unit enables the computing device to communicate with the outside world via a network. Here, the network includes both wired and wireless methods. In particular, wireless communication may include cellular communication using at least one of, for example, LTE, LTE-A (LTE Advance), CDMA (code division multiple access), WCDMA (wideband CDMA), UMTS (universal mobile telecommunications system), WiBro (Wireless Broadband), or GSM (Global System for Mobile Communications). Alternatively, wireless communication may include at least one of WiFi (wireless fidelity), LiFi (light fidelity), Bluetooth, Bluetooth Low Energy (BLE), Zigbee, NFC (near field communication), Magnetic Secure Transmission, Radio Frequency (RF), or Body Area Network (BAN). Alternatively, wireless communication may include GNSS. GNSS may be, for example, GPS (Global Positioning System), Glonass (Global Navigation Satellite System), Beidou Navigation Satellite System (hereinafter "Beidou") or Galileo, the European global satellite-based navigation system.Wired communication may include at least one of, for example, USB (universal serial bus), HDMI (high definition multimedia interface), RS-232 (recommended standard 232), power line communication, or POTS (plain old telephone service), computer networks (e.g., LAN or WAN).

[0071] Hereinafter, an anomaly detection method according to one embodiment of the present invention will be described in detail. In addition, the anomaly detection method described below can be implemented by a processor executing a program stored in a recording medium equipped with the controller (40) described above.

[0072] FIG. 2 is a flowchart for explaining an anomaly detection method according to one embodiment of the present invention.

[0073] Referring to FIG. 2, an anomaly detection method according to one embodiment of the present invention may include a wavelength selection step (S10), a selected wavelength data collection step (S20), a cluster classification step (S30), a reference data generation step (S40), and an anomaly detection step (S50).

[0074] In the wavelength selection step (S10), at least one wavelength satisfying a reference condition among the collectible wavelengths that can be collected by the optical analyzer (30) within the process chamber (10) can be selected as the selected wavelength.

[0075] FIG. 3 is a diagram illustrating wavelength candidates that can be selected as selected wavelengths in the wavelength selection step of FIG. 2. Specifically, as shown in FIG. 3, while performing a plasma treatment process on a substrate (W) in a process chamber (10), light having various wavelengths may be generated. The range of wavelengths that can be collected by the light analyzer (30) within the process chamber (10) may be 200 nm to 1000 nm, and the number of variables (i.e., the number of wavelength candidates) may be 1500 or more. That is, there may be multiple wavelength candidates (W1, W2, W3, …, WN, N is a natural number) that can be collected by the light analyzer (30) within the process chamber (10).

[0076] Among these wavelength candidates (WN), a wavelength to be used for process abnormality judgment can be selected. Below, the wavelength selected to be used for process abnormality judgment can be defined as the selected wavelength.

[0077] Figure 4 is a diagram illustrating a method for selecting a selected wavelength among wavelength candidates in the wavelength selection step of Figure 2.

[0078] Referring to FIG. 4, among the wavelength candidates (WN), the selected wavelength used for determining process abnormalities may be a wavelength that satisfies a preset criterion. The preset criterion, the criterion condition, may be defined as a case where the difference between the light intensity at the beginning of the process and the light intensity at the end of the process is relatively large. In order to select a wavelength that satisfies the criterion condition among the wavelength candidates (WN), multiple substrates (W) are processed, and light intensity data for the light generated at this time can be collected through a light analyzer (30) as normal process samples. This can then be stored in the controller (40).

[0079] The graph illustrated in FIG. 4 is a graph showing the change in light intensity of a specific wavelength in light intensity data collected when the substrate (W) is processed normally. When the process is performed, the process can be divided into multiple sections. The user can set the size of each section. For example, if the size of each section is set to 10% of the total process time, the process can be divided into 10 sections when one process is performed. At this time, the controller (40) can measure the Wasserstein distance between the starting section (an example of the first section, initial 10%) and the last section, the representative section (an example of the second section, final process later 10%), using a pre-stored algorithm. Wasserstein distance may be a measure used to measure the distance between two probability distributions.

[0080] The larger the Wasserstein distance, the greater the change in light intensity between the beginning and end sections of the process may be for the wavelength. The controller (40) measures the Wasserstein distance from the normal process light intensity data for each of the wavelength candidates (WN) and lists them in order from the largest to the smallest measured Wasserstein distance. Then, the user can pre-set up to the nth wavelength candidate (WN) and select it as the selected wavelength. A larger Wasserstein distance means that the difference in light intensity between the beginning and end of the process is greater. A large difference in intensity between the beginning and end of the process for light of a specific wavelength range allows it to be estimated that the light of that wavelength is light generated as the removal target being removed from the substrate (W) reacts with the plasma.

[0081] In addition, the user may measure not only the Wasserstein distance between the starting section and the representative section as needed, but also the distance between the starting section and the section immediately preceding the representative section (an example of the third section, the section immediately preceding the last process, 10%). Depending on the process, some process noise may occur in the representative section where the process ends, so the Wasserstein distance between the section immediately preceding the representative section and the starting section may be measured to resolve the above problem.

[0082] In some cases, the sum of Wasserstein distances between the starting interval and the representative interval, and between the starting interval and the immediate preceding interval, may be calculated, and the selected wavelength may be chosen based on this. For instance, the wavelength candidate with the largest sum can be selected as the final wavelength. This is intended to ensure that if noise occurs in the optical intensity data in either the representative interval or the immediate preceding interval, the other can compensate for it.

[0083] In the manner described above, a wavelength candidate (WN) satisfying the standard conditions is selected as the selected wavelength. For example, as shown in FIG. 5, a wavelength with a large difference in light intensity between the beginning and end of the process is selected as the selected wavelength, and as shown in FIG. 6, a wavelength with a small difference in light intensity between the beginning and end of the process may be difficult to select as the selected wavelength.

[0084] Referring again to FIG. 2, in the selected wavelength data collection step (S20), light intensity data for the selected wavelength selected in the wavelength selection step (S10) can be collected. For example, if the selected wavelengths are selected as 200 nm and 300 nm, light intensity data for 200 nm and light intensity data for 300 nm collected while the substrate (W) is processed in the process chamber (10) can be collected for each selected wavelength.

[0085] The light intensity data described above can be defined as reference light intensity data. The reference light intensity data may be collected by performing a process on an additional substrate (W) after the wavelength selection step (S10) is completed, or alternatively, by collecting light intensity data for the selected wavelength from the normal process sample described above.

[0086] Reference light intensity data may exhibit somewhat different characteristics, as illustrated in FIG. 7. FIG. 7 may show a change in light intensity over time at a specific wavelength (e.g., 200 nm wavelength) when the process chamber is operated with the same process recipe (e.g., when the process is performed by controlling the power intensity, pressure inside the chamber, flow rate of the supplied process gas, etc., in the same way). As illustrated in FIG. 7, L1 and L2 exhibit similar characteristics, and L3 to L8 exhibit similar characteristics. L1 to L8 may be light intensities collected when different substrates (W) are each processed with the same process recipe, and may be the previously defined reference light intensity data.

[0087] In this way, when the process chamber (10) is operated with the same process recipe and changes in light intensity of the same wavelength are observed, reference light intensity data showing different aspects can be collected, which may vary depending on the type of material of the object to be removed from the substrate (W). For example, the light intensity change data observed when the object to be removed from the substrate (W) is a first layer provided with material A, and when the object to be removed is a second layer provided with material B, may be different from each other.

[0088] Referring again to FIG. 2, in the clustering step (S30), reference light intensity data that vary depending on the type of object to be removed from the substrate (W) can be clustered into multiple clusters.

[0089] Specifically, the controller (40) can cluster the reference light intensity data as described above through a pre-stored cluster algorithm. A cluster algorithm is a technique that defines clusters of similar data. As for the clustering technique, if the user wants to define the number of clusters, the K-means methodology, which is a supervised learning methodology, can be used, and if the number of clusters is unknown, the Hierarchical methodology can be used. Through such a cluster algorithm, the reference light intensity data can be divided into multiple clusters.

[0090] In an embodiment of the present invention, as shown in FIG. 8, if the object to be removed is provided as material A in the first layer, it is classified as a first cluster; if the object to be removed is provided as material B in the second layer, it is classified as a second cluster; otherwise, it is determined to be an outlier. Then, when each cluster is formed, a representative waveform for each cluster to be used in the cluster determination step (S51) described below can be separately stored in the controller (40). For example, the representative waveform for each cluster may be the average of the reference data belonging to each cluster.

[0091] Referring again to FIG. 2, in the reference data generation step (S40), reference data can be generated to be used when determining whether there is an anomaly in the target data in the anomaly detection step (S50) described later. The reference data can be derived by clustering reference light intensity data, which vary depending on the type of object to be removed from the substrate (W), into multiple clusters, and calculating the end point time and end point area of ​​the reference light intensity data included in each cluster.

[0092] Figure 9 is a distribution graph showing the endpoint time and endpoint area distributions generated for each cluster.

[0093] FIG. 9 illustrates the distribution of abnormality judgment values ​​of reference light intensity data belonging to the first cluster and abnormality judgment values ​​of reference light intensity data belonging to the second cluster, and the distribution shown in FIG. 9 may be one example of reference data defined in the present invention. The abnormality judgment values ​​may include end point time and end point area. Based on the reference data as shown in FIG. 9 and the abnormality judgment values ​​obtained from the target data described below, it is possible to determine whether there is a process abnormality when collecting target data.

[0094] Figure 10 is a graph illustrating the endpoint time and endpoint area distribution of Figure 9.

[0095] Referring to FIG. 10, first, in the total process time, the time set when the process recipe starts and the time set before the process recipe ends are considered dead time, and data collected during these times are excluded when calculating the end point time and end point area. The dead time can be designated as the time set after the plasma is turned on (i.e., the power supply of the power supply unit is turned on) and the time set before the plasma is turned off (i.e., the power supply of the power supply unit is turned off).

[0096] Then, the remaining time interval is divided into multiple segments while excluding dead time. For example, among the multiple segments (SE1 to SE10), the segment (SE10) belonging to the last 10% of the segment can be set as the representative segment (an example of the first segment), and the remaining segments (SE1 to SE9) can be set as comparison segments (an example of the second segment). In the case of the etching (or ashing) process, in order to completely etch (or ashing) the object to be removed on the substrate (W), over-etching (ashing) is performed beyond the existing process time, which usually corresponds to the last 10% of the process. Therefore, the segment (SE10) belonging to the last 10% of the segment is set as the representative segment.

[0097] Then, the average value of the light intensity of the reference light data belonging to each cluster is calculated. Then, a reference time point (ET) is selected to detect the end point, which is the end of the etching process. The point in time when the slope of the light intensity of the wavelength in the representative section (SE10) enters within the set slope can be selected as the reference time point (ET).

[0098] When the user derives the aforementioned setting slope from the average light intensity value, the reference time point (ET) of each reference light data is derived based on the said setting slope.

[0099] In addition, the time taken from the start of the process recipe excluding dead time to the above reference time point (ET) is defined as the end point time, and the area under the corresponding graph is defined as the end point area.

[0100] In the distribution diagram of Figure 9 described above, the endpoint time and endpoint area of ​​each reference light data belonging to each cluster are displayed as previously defined.

[0101] Figure 11 is a flowchart for explaining the anomaly detection step of Figure 2.

[0102] Referring to FIGS. 2 and FIGS. 11, in the anomaly detection step (S50), the target data to be judged is compared with reference data to determine whether there is an anomaly in the target data. If an anomaly occurs in the target data, it can be presumed that an anomaly occurred in the semiconductor manufacturing process when the target data was collected.

[0103] When performing the anomaly detection step (S50), the previously performed wavelength selection step (S10), selected wavelength data collection step (S20), cluster classification step (S30), and reference data generation step (S40) may have been performed in advance, and reference data may have been generated in advance through this.

[0104] In order to perform the anomaly detection step (S50), the process chamber (10) performs a processing process on the substrate (W), thereby allowing target data, which is light intensity data for a selected wavelength, to be collected. When the target data is collected through the light analyzer (30), the controller (40) can extract light intensity data for a pre-selected wavelength from the target data. Then, a cluster determination step (S51) is performed on the extracted light intensity data to determine whether the extracted light intensity data belongs to a first cluster and a second cluster. The same cluster algorithm used in the cluster classification step (S30) described above can be applied to the cluster determination step (S51). Through this, when collecting the target data, the material of the object to be removed from the substrate (W) can be estimated.

[0105] Afterward, the controller (40) may perform an abnormality judgment value calculation step (S52) to calculate an abnormality judgment value from the target data. The abnormality judgment value may be the endpoint time and endpoint area described above.

[0106] Subsequently, the controller (40) may perform an abnormality determination step (S53) to detect whether there is a process abnormality when collecting target data by comparing the abnormality determination values ​​of the target data with the abnormality determination values ​​of the reference data. For example, the controller (40) may connect the numerical distributions located at the outermost edge of each cluster in the abnormality determination value distribution shown in FIG. 9 into a closed loop, and determine whether there is a process abnormality based on whether the abnormality determination value of the target data is located within the area defined by the aforementioned closed loop. If it is located within the area, it may be determined as a normal process, and if it is located outside the area, it may be determined as an abnormal process.

[0107] In contrast, the average value of each cluster is calculated from the reference data, and if the distance between the calculated average value and the abnormal judgment numerical distribution of the target data is greater than or equal to a set distance, it is determined to be an abnormal process, and if it is less than the set distance, it is determined to be a normal process. In particular, the present invention determines process abnormalities by considering not only the end point time but also the end point area together, because if process abnormalities are determined solely by the end point time, process abnormalities occurring in the middle of the process cannot be properly identified, and if process abnormalities are determined solely by the end point area, cases where the process time is excessively long cannot be properly detected.

[0108] In addition, the present invention classifies each data into multiple clusters using previously collected normal process sample data and first determines which of the multiple clusters the target data belongs to. This enables accurate determination of process abnormalities even if the type of object to be removed changes.

[0109] It should be understood that exemplary embodiments are disclosed herein and that other variations may be possible. Individual elements or features of a particular embodiment are generally not limited to that particular embodiment, but may be interchangeable and used in selected embodiments where applicable, even if not specifically illustrated or described. Such variations should not be construed as departing from the spirit and scope of the disclosure, and all such variations that are obvious to a person skilled in the art are intended to be included within the scope of the following claims.

[0110] [Explanation of the symbol]

[0111] S10: Wavelength selection step

[0112] S20: Data collection step by selected wavelength

[0113] S30: Cluster separation step

[0114] S40: Reference data generation step

[0115] S50: Anomaly detection stage

[0116] 1: Substrate processing device

[0117] 10: Process chamber

[0118] 20: Optical sensor

[0119] 30: Optical analyzer

[0120] 40: Controller

Claims

1. An anomaly detection method for detecting anomalies in a semiconductor manufacturing process by analyzing light within a process chamber that processes a substrate using plasma: A clustering step for clustering reference light intensity data, which vary depending on the type of object to be removed from the substrate, into multiple clusters; A reference data generation step for generating reference data including a distribution of abnormal judgment values ​​for each of the plurality of clusters above; and A step for detecting the above anomaly, comprising an anomaly detection step for determining whether the target data is anomaly by comparing the target data and the reference data for detecting the above anomaly. Anomaly detection method.

2. In Paragraph 1, The above anomaly detection step is: A cluster determination step for determining which cluster the above target data belongs to among the clusters distinguished in the cluster distinction step; An abnormality judgment value calculation step for calculating the abnormality judgment values ​​of the above target data; and A step for determining abnormality that compares the abnormality determination values ​​of the target data with the abnormality determination values ​​of the reference data to determine whether there is an abnormality in the semiconductor manufacturing process when collecting the target data, comprising an abnormality determination step. Anomaly detection method.

3. In Paragraph 2, The above abnormal judgment values ​​are: End Point Time; and Including the end point area, Anomaly detection method.

4. In any one of paragraphs 1 through 3, The method further comprises a wavelength selection step of selecting at least one wavelength satisfying a reference condition among the collectible wavelengths collectible within the process chamber as selected wavelengths. Anomaly detection method.

5. In Paragraph 4, A selection wavelength data collection step further comprising collecting the reference light intensity data collected when processing the substrate for the above selection wavelength, Anomaly detection method.

6. In Paragraph 5, The above standard condition measures the Wasserstein distance between a first section and a second section different from the first section in the optical intensity data for each of the above collectible wavelengths, lists the Wasserstein distances in order of largest to smallest, and then selects the up to the nth collectible wavelength as the selected wavelength by prioritizing the wavelength with the largest Wasserstein distance. Anomaly detection method.

7. In any one of paragraphs 1 through 3, The clustering method used in the above cluster classification step is the K-Means method or the Hierarchical method, Anomaly detection method.

8. In a method for detecting anomalies in a semiconductor manufacturing process, Determining whether there is an abnormality in the above process by comparing pre-stored reference data with target data for detecting abnormalities, An anomaly detection method comprising the above reference data including distribution data for endpoint time and endpoint area.

9. In Paragraph 8, An anomaly detection method in which the above reference data is derived by clustering reference light intensity data, which vary according to the type of object to be removed from the substrate, into multiple clusters, and calculating the endpoint time and endpoint area of ​​the reference light intensity data included in each of the clusters.

10. A program stored on a recording medium that detects abnormalities in a semiconductor manufacturing process by analyzing light inside a process chamber that processes a substrate using plasma, The above program is: Perform an anomaly detection operation to determine whether there is an anomaly in the above process by comparing pre-stored reference data with target data for detecting anomalies, The above reference data includes distribution data for endpoint time and endpoint area, A program stored on a recording medium.

11. In Paragraph 10, The above program is: A wavelength selection operation further performed to select at least one wavelength satisfying a reference condition among the collectible wavelengths collectible within the process chamber as selected wavelengths. A program stored on a recording medium.

12. In Paragraph 11, The above program is: A clustering operation for clustering reference light intensity data for the above-mentioned selected wavelength into multiple clusters; and For each of the above plurality of clusters, a reference data generation operation is further performed to generate reference data including a distribution of abnormal judgment values. A program stored on a recording medium.

13. In Paragraph 12, The above anomaly detection operation is: A cluster determination operation for determining which cluster the target data belongs to among the clusters distinguished in the cluster distinction operation; and A numerical operation for determining an anomaly, comprising calculating the endpoint time and endpoint area of ​​the target data. A program stored on a recording medium.

14. In Paragraph 13, The above anomaly detection operation is: The method further includes an abnormality determination operation for determining whether there is an abnormality in the semiconductor manufacturing process when collecting the target data by comparing the endpoint time and the endpoint area of ​​the target data with the endpoint time and the endpoint area of ​​the reference data. A program stored on a recording medium.

Images