Determination of epitaxial film defects

Abstract

The method includes acquiring first image data of a substrate containing an epitaxial film using a processing device. The method further includes applying a frequency domain filter to the first image data to acquire filtered image data. The method further includes determining the number of epitaxial defects represented in the first image data by performing feature detection on the filtered image data. The method further includes performing corrective actions taking into account the number of epitaxial defects.

Description

Technical Field

[0001]

[0001] This disclosure relates to methods related to epitaxial films of manufactured devices. More specifically, this disclosure relates to the determination of epitaxial film defects.

Background Art

[0002]

[0002] Products can be manufactured by performing one or more manufacturing processes using manufacturing equipment. For example, semiconductor manufacturing equipment can be used to manufacture substrates through semiconductor manufacturing processes. Products are manufactured with specific characteristics suitable for target applications. Images of products (e.g., manufactured devices) can be taken, which can be used to deepen the understanding of the device's function, faults, and performance and for measurement or inspection, etc. Products can include defects that can be characterized based on images of the products.

Summary of the Invention

[0003]

[0003] The following is a simplified summary of this disclosure, provided to offer a basic understanding of some aspects of this disclosure. This summary is not an exhaustive overview of this disclosure. It is not intended to identify the main or important elements of this disclosure, nor to define any scope of specific embodiments of this disclosure or any scope of the claims. The sole purpose of this summary is to present, in a simplified form, some concepts of this disclosure as a prelude to the more detailed description that follows.

[0004]

[0004] In one aspect of this disclosure, the method includes obtaining first image data of a substrate including an epitaxial film by a processing device. The method further includes applying a frequency domain filter to the first image data to obtain filtered image data. The method further includes determining the number of epitaxial defects represented in the first image data by performing feature detection on the filtered image data. The method further includes performing corrective measures considering the number of epitaxial defects.

[0005]

[0005] In another aspect of the present disclosure, a non-temporary machine-readable storage medium stores an instruction which, when executed, causes a processing device to perform a process. The process includes acquiring first image data of a substrate including an epitaxial film. The process includes applying a frequency-domain filter to the first image data. The process further includes determining the number of epitaxial defects represented in the first image data by performing feature detection. The process further includes taking the number of epitaxial defects into account and performing corrective actions.

[0006]

[0006] In another aspect of the present disclosure, the system includes a memory and a processing device coupled to the memory. The processing device is configured to acquire first image data of a substrate including an epitaxial film. The processing device is further configured to apply a frequency domain filter to the first image data. The processing device is further configured to determine the number of epitaxial defects represented in the first image data by performing feature detection. The processing device is further configured to perform corrective actions taking into account the number of epitaxial defects.

[0007]

[0007] The present disclosure is illustrated, not limited, in the drawings of the attached drawings. [Brief explanation of the drawing]

[0008] [Figure 1] This block diagram shows an exemplary system architecture according to several embodiments. [Figure 2A] A block diagram of a system including an exemplary dataset generator for creating datasets for one or more supervised models, according to several embodiments, is shown. [Figure 2B] A block diagram of an exemplary dataset generator for creating a dataset for a supervised model configured to generate an index of predicted epitaxial defect characteristics of a substrate, according to several embodiments, is shown. [Figure 3] Block diagram of a system for generating output data according to several embodiments. [Figure 4A] This is a flowchart of a method for generating a dataset for a machine learning model according to several embodiments. [Figure 4B] This is a flowchart of a method for performing corrective measures that take into account the determination of epitaxial film defects, according to several embodiments. [Figure 4C] This is a flowchart illustrating a method for training and utilizing machine learning models related to epitaxial defect detection, according to several embodiments. [Figure 5A] This is a block diagram of a flow for generating epitaxial film defect data according to several embodiments. [Figure 5B] This is an exemplary epitaxial defect map of a substrate according to several embodiments. [Figure 6] This is a block diagram showing a computer system according to several embodiments. [Modes for carrying out the invention]

[0009]

[0018] This specification describes techniques for identifying, determining, and classifying defects in epitaxial films on substrates. Manufacturing equipment is used to manufacture products such as substrates (e.g., wafers, semiconductors). Manufacturing equipment may include a manufacturing chamber or processing chamber for isolating the substrate from the environment. The characteristics of the manufactured substrate are to meet target values ​​in order to promote a particular function. Manufacturing parameters are selected to manufacture a substrate that meets target characteristic values. Target characteristic values ​​may include target defect parameters (e.g., maximum number of target defects, maximum density of target defects, target count of density of two or more types of defects). Many manufacturing parameters (e.g., hardware parameters, process parameters, etc.) contribute to the characteristics of the processed substrate. The manufacturing system can control the parameters by specifying setpoints for the characteristic values, receiving data from sensors placed in the manufacturing chamber, and adjusting the manufacturing equipment until the sensor readings match the setpoints. In some embodiments, trained machine learning models are used to improve the performance of the manufacturing equipment and / or the performance of the manufactured substrate.

[0010]

[0019] Classifying, counting, and determining the location of defects induced by the substrate manufacturing process enables improvements to the manufacturing process, manufacturing equipment, and substrate design. The substrate may contain an epitaxial film. The epitaxial film is a crystalline layer of material deposited on the substrate, and is also crystalline. In some cases, the arrangement of the epitaxial film (for example, based on the lattice properties of the deposited material) is aligned with the arrangement of the underlying substrate. In some cases, the deposited material may have different lattice properties than the underlying material, which can lead to the formation of defects in the epitaxial film. The epitaxial film may have other types of deposition defects, including dislocations (e.g., deviations from a regular atomic arrangement along a line), point defects (e.g., related to defects or extra atoms in the crystal lattice), stacking faults (e.g., related to interruptions in the regular stacking arrangement of the crystal plane), impurities that disrupt the regular lattice structure, and lower lattice defects that cause defects in the arrangement of subsequent crystal layers. Common symptoms of epitaxial defects include pits (e.g., small, nearly circular defects present in a substrate image) and cross-hatching (e.g., a series of parallel lines present in a substrate image).

[0011]

[0020] In some cases, the detection of epitaxial defects can be performed manually using the coupon method. In this process, a portion of the substrate (e.g., a semiconductor wafer) may be designed or designated to be tested for epitaxial defects. The coupon portion may be removed from the substrate (e.g., cut). The coupon portion may be checked for epitaxial defects. In some cases, the coupon is relatively small compared to the substrate. Separating the coupon from the rest of the substrate, testing, etc., and generating it can be costly. For example, classifying epitaxial defects may involve manual review of one or more images of coupons by a subject expert. This is time-consuming, costly, and potentially inaccurate. Furthermore, in some cases, it may not be possible to remove coupons from a particular substrate, reducing the applicability of coupon technology in epitaxial film defect detection.

[0012]

[0021] Epitaxial defects can be as small as, for example, on the order of nanometers to micrometers. Determining the location of epitaxial defects may involve subject expert review of high-resolution images to make these defects visible. In some cases, the substrate containing the epitaxial film may be orders of magnitude larger than the epitaxial defects (e.g., related to defects in the film's crystal lattice). For example, the substrate may have a diameter of 300 mm. Under such conditions, providing a mapping of epitaxial defects for a sufficiently large portion of the substrate to confidently infer the defect characteristics of the substrate may be extremely expensive, time-consuming, inconvenient, or impossible. For example, it may be possible to image the entire surface of the substrate using thousands of images, which may involve hours of defect classification, defect counting, etc., by trained users to generate any practical insights related to epitaxial defects. The time, expertise, and cost commitments associated with improving manufacturing systems based on epitaxial defects can be exorbitant.

[0013]

[0022] The systems, devices, and methods of this disclosure can address one or more shortcomings of conventional systems for epitaxial film defects. In some embodiments, image data of a substrate containing an epitaxial film is generated. The image data may be generated through several different imaging techniques. Imaging techniques may include optical microscopy, scanning electron microscopy, transmission electron microscopy, or other imaging techniques. In some embodiments, the substrate may be silicon. The epitaxial film may be made of a material having different lattice properties than the underlying substrate.

[0014]

[0023] In some embodiments, the epitaxial defects can be decorated by performing an etching process. An epitaxial defect can be a location where the crystal structure of the epitaxial film is interrupted, disrupted, or otherwise disturbed. Epitaxial defects can be more sensitive to etching conditions (e.g., plasma etching, chemical etching, etc.) than other locations of the epitaxial film. By performing the etching process, the epitaxial defects can be imaged more clearly by one or more selected imaging techniques.

[0015]

[0024] In some embodiments, one or more preprocessing steps can be performed to adjust the image data of the substrate. For example, artifact removal, image smoothing or sharpening processes, noise reduction techniques such as Gaussian blur, etc. can be performed on the image data.

[0016]

[0025] In some embodiments, the spatial image can be converted to the frequency domain. The conversion to the frequency domain can be performed by executing a Fourier transform, a Fourier series decomposition, an application of a fast Fourier transform algorithm, etc. The frequency domain image data can be applied with a frequency domain filter, for example, to improve the detection of target types of defects.

[0017]

[0026] In some embodiments, the number of epitaxial defects can be determined by a processing device. The determination of the number of epitaxial defects can be performed by one or more feature detection methods. For example, a Hough transform process may be performed to detect approximately circular etching pits, and a contour detection process may be performed to detect cross-hatching. Further, the density of various types of defects can be determined based on the number of defects. The distribution of target types of defects can be determined based on the imaged position, the number of defects, the defect density, etc.

[0018]

[0027] The system of the present disclosure provides technical advantages over conventional solutions. Imaging can be performed by non-destructive methods, such as optical microscopy, which offers advantages over conventional bonding methods. The image can be obtained by imaging the substrate using optical microscopy. Additionally, automatic detection, classification, and analysis of defects by a processing device can be performed, which can be done more quickly, with less user input, fewer opportunities for user error, and less input by subject experts, etc., compared to manual classification, counting, and / or further analysis of epitaxial defects.

[0019]

[0028] In one aspect of the present disclosure, the method includes obtaining first image data of a substrate including an epitaxial film by a processing device. The method further includes applying a frequency domain filter to the first image data to obtain filtered image data. The method further includes determining the number of epitaxial defects represented in the first image data by performing feature detection on the filtered image data. The method further includes performing corrective measures in consideration of the number of epitaxial defects.

[0020]

[0029] In another aspect of the present disclosure, a non-transitory machine-readable storage medium stores instructions that, when executed, cause a processing device to perform steps. The steps include obtaining first image data of a substrate including an epitaxial film. The steps include applying a frequency domain filter to the first image data. The steps further include determining the number of epitaxial defects represented in the first image data by performing feature detection. The steps further include performing corrective measures in consideration of the number of epitaxial defects.

[0021]

[0030] In another aspect of this disclosure, the system includes a memory and a processing device coupled to the memory. The processing device is configured to acquire first image data of a substrate including an epitaxial film. The processing device is further configured to apply a frequency-domain filter to the first image data. The processing device is further configured to determine the number of epitaxial defects represented in the first image data by performing feature detection. The processing device is further configured to perform corrective actions taking into account the number of epitaxial defects.

[0022]

[0031] Figure 1 is a block diagram showing an exemplary system 100 (exemplary system architecture) in several embodiments. System 100 includes a client device 120, manufacturing equipment 124, sensors 126, measuring equipment 128, a prediction server 112, and a data store 140. The prediction server 112 may be part of the prediction system 110. The prediction system 110 may further include server machines 170 and 180.

[0023]

[0032] Sensor 126 can provide sensor data 142 related to manufacturing equipment 124 (for example, related to manufacturing corresponding products such as substrates by manufacturing equipment 124). Sensor data 142 can be used to verify the health of the equipment and / or the health of the product (e.g., product quality). Manufacturing equipment 124 can manufacture products according to a recipe or perform processing over a period of time. In some embodiments, sensor data 142 may include one or more values ​​from optical sensor data, spectral data, temperature (e.g., heater temperature), spacing (SP), pressure, high-frequency radio frequency (HFRF), high-frequency (RF) matching voltage, RF matching current, RF matching capacitor position, electrostatic chuck (ESC) voltage, actuator position, current, flow, power, voltage, etc. Sensor data 142 may include historical sensor data and current sensor data. Current sensor data may be related to the product currently being processed, recently processed product, number of recently processed products, etc. Historical sensor data may include data stored in relation to previously manufactured products. Past and / or current sensor data may include attribute data, such as a manufacturing equipment ID or design label, sensor ID, type, and / or location, and a manufacturing equipment status label, such as current failure or service life.

[0024]

[0033] The sensor data 142 may be associated with or indicate manufacturing parameters such as hardware parameters of the manufacturing equipment 124 (e.g., hardware settings or installed components, e.g., size, type, etc.) or process parameters of the manufacturing equipment 124 (e.g., heater settings, gas flow, etc.). Some data associated with hardware parameters and / or process parameters may be stored instead or additionally as manufacturing parameters 150, which may include past and current manufacturing parameters (e.g., associated with past processing runs). Manufacturing parameters 150 may indicate input settings to the manufacturing device (e.g., heater power, gas flow, etc.). The sensor data 142 and / or manufacturing parameters 150 may be provided while the manufacturing equipment 124 is performing a manufacturing process (e.g., readings from the equipment while processing a product). The sensor data 142 may differ for each product (e.g., for each substrate). The substrate may have characteristic values ​​(e.g., film thickness, film strain) measured by the measuring instrument 128 (e.g., measured in a standalone measuring facility). The measurement data 160 may be a component of the data store 140. The measurement data 160 may include past measurement data 164 (e.g., measurement data related to previously processed products). The measurement data 160 may include current measurement data 166 (e.g., measurement data of one or more target substrates, one or more subject substrates, etc.). In some embodiments, the current measurement data 166 may be provided to one or more processing devices for performing epitaxial film defect determination. In some embodiments, the current measurement data 166 may be provided to one or more trained machine learning models for determining epitaxial film defects.

[0025]

[0034] In some embodiments, the measurement data 160 may be provided without using a standalone measurement facility, for example, in-situ measurement data (e.g., measurements or proxies for measurements collected during processing), integrated measurement data (e.g., measurements or proxies for measurements collected when the product is in the chamber or under vacuum but not during the processing steps), in-line measurement data (e.g., data after the substrate has been removed from the vacuum), etc. The measurement data 160 may include current measurement data 166 (e.g., measurement data related to the current or recently processed product).

[0026]

[0035] The measuring instrument 128 may include, for example, an imaging device for performing substrate imaging technology. The measurement data 160 may include image data generated by the imaging device of the measuring instrument 128. The imaging technology performed by the measuring instrument 128 may include one or more optical microscopes, scanning electron microscopes, transmission electron microscopes, etc.

[0027]

[0036] In some embodiments, sensor data 142, measurement data 160, or manufacturing parameters 150 may be processed (e.g., by a client device 120 and / or a prediction server 112). Processing of sensor data 142 may include generating and / or detecting features. In some embodiments, features are patterns of sensor data 142, measurement data 160, and / or manufacturing parameters 150 (e.g., slope, width, height, peak, etc.), or combinations of values ​​from sensor data 142, measurement data, and / or manufacturing parameters (e.g., power derived from voltage and current). In some embodiments, features are shapes or relationships between parts of image data, e.g., circles indicating epitaxial etching pits, contour lines indicating epitaxial cross-hatching, etc. Sensor data 142, manufacturing parameters 150, and / or measurement data 160 may include features, which may be used by a prediction component 114 to perform signal processing and / or obtain prediction data 168 for performing corrective actions.

[0028]

[0037] Each instance (e.g., set) of sensor data 142 may correspond to a product (e.g., a circuit board), a set of manufacturing equipment, or the type of circuit board manufactured by the manufacturing equipment. Similarly, each instance of measurement data 160 and manufacturing parameter 150 may correspond to a product, a set of manufacturing equipment, or the type of circuit board manufactured by the manufacturing equipment. The data store can further store information that associates sets of different data types, such as information indicating that sets of sensor data, sets of measurement data, and sets of manufacturing parameters are all associated with the same product, manufacturing equipment, type of circuit board, etc.

[0029]

[0038] The prediction system 110 may be used to generate prediction data 168. Prediction data 168 may be generated by performing an image processing step on image data of one or more substrates. Prediction data 168 may be data indicating epitaxial defects in the substrate. Prediction data 168 may indicate the classification, count, location, density, severity, etc., of epitaxial defects in one or more substrates. Prediction data 168 may be generated by performing a feature detection step on image data associated with one or more substrates. The process for determining epitaxial defects in the substrate may be performed by a prediction server 112, a prediction component 114, etc.

[0030]

[0039] In some embodiments, the prediction system 110 can be used to generate prediction data 168 using one or more trained machine learning models. In some embodiments, the prediction system 110 can generate prediction data 168 using supervised machine learning. Supervised machine learning refers to the process associated with a machine learning model provided with labeled training data, such as image data labeled by the number of epitaxial defects. In some embodiments, the prediction system 110 can generate prediction data 168 using unsupervised machine learning. Unsupervised machine learning refers to the process associated with a machine learning model trained using unlabeled inputs. Unsupervised machine learning processes may include clustering results, principal component analysis, anomaly detection, etc. In some embodiments, the prediction system 110 can generate prediction data 168 using semi-supervised learning (for example, the training data may include a combination of labeled and unlabeled data).

[0031]

[0040] The client device 120, manufacturing equipment 124, sensor 126, measuring equipment 128, prediction server 112, data store 140, server machine 170, and server machine 180 may be connected to each other via network 130 to generate prediction data 168 and perform corrective actions. In some embodiments, network 130 may provide access to cloud-based services. The processing performed by the client device 120, prediction system 110, data store 140, etc., may be performed by a virtual cloud-based device.

[0032]

[0041] In some embodiments, network 130 is a public network providing client device 120 with access to prediction server 112, data store 140, and other publicly available computing devices. In some embodiments, network 130 is a private network providing client device 120 with access to manufacturing equipment 124, sensors 126, measuring instruments 128, data store 140, and other privately available computing devices. Network 130 may include one or more wide area networks (WANs), local area networks (LANs), wired networks (e.g., Ethernet networks), wireless networks (e.g., 802.11 networks or Wi-Fi networks), cellular networks (e.g., Long-Term Evolution (LTE) networks), routers, hubs, switches, server computers, cloud computing networks, and / or combinations thereof.

[0033]

[0042] The client device 120 may include computing devices such as personal computers (PCs), laptops, mobile phones, smartphones, tablet computers, netbooks, network-connected televisions ("smart TVs"), network-connected media players (e.g., Blu-ray players), set-top boxes, over-the-top (OTT) streaming devices, and operator boxes. The client device 120 may include a corrective action component 122. The corrective action component 122 may receive user input from a display associated with the manufacturing equipment 124 (e.g., through a graphical user interface (GUI) displayed via the client device 120). In some embodiments, the corrective action component 122 sends instructions to a prediction system 110, receives output from the prediction system 110 (e.g., prediction data 168), determines corrective action based on the output, and has the corrective action performed. In some embodiments, the corrective action component 122 retrieves sensor data 142 associated with the manufacturing equipment 124 (e.g., current sensor data 146) (e.g., from a data store 140, etc.) and provides the sensor data 142 associated with the manufacturing equipment 124 (e.g., current sensor data 146) to the prediction system 110.

[0034]

[0043] In some embodiments, the corrective action component 122 can acquire current measurement data 166 (e.g., one or more images of the substrate of interest that may contain signs of an epitaxial defect). The corrective action component 122 can provide the current measurement data 166 to the prediction system 110 to determine the presence of an epitaxial defect in the substrate. The prediction system 110 may utilize image processing, feature detection processing, image filtering, frequency domain transformation, and other techniques to make a determination in relation to the measurement data and the epitaxial defect. The prediction system 110 can provide one or more images of the substrate to a trained machine learning model (e.g., model 190) to make a determination in relation to the epitaxial defect in the substrate. In some embodiments, model 190 may represent one or more physically based models, image processing models, or machine learning models. In some embodiments, model 190 may be configured to make predictions indicating portions of the substrate that were not imaged (e.g., generate prediction data 168). For example, Model 190 may be configured to predict the characteristics of a portion of the substrate that was not imaged (e.g., epitaxial defect location and / or classification) based on image data of a portion of the substrate that was imaged.

[0035]

[0044] In some embodiments, the prediction data 168 generated by the prediction system 110 may include instructions for corrective actions to be taken. In some embodiments, a corrective action component 122 receives instructions for corrective actions from the prediction system 110 and causes the corrective actions to be taken. Each client device 120 may include an operating system that enables the user to generate, view, or edit one or more of the data (e.g., displays associated with the manufacturing equipment 124 and corrective actions associated with the manufacturing equipment 124).

[0036]

[0045] In some embodiments, measurement data 160 (e.g., historical measurement data 164) corresponds to historical characteristic data of a product, and predictive data 168 is associated with predictive characteristic data. Historical measurement data 164 may correspond to products processed using manufacturing parameters and / or recipes related to historical sensor data and historical manufacturing parameters. Predictive data 168 may include analysis performed on the measurement data to generate additional insights into the performance of the manufacturing system, chamber, recipe, substrate, etc. In some embodiments, predictive data 168 is or includes predictive measurement data (e.g., virtual measurement data, virtual synthetic microscope image) of a product manufactured or a product manufactured according to conditions recorded as current sensor data, current measurement data, current measurement data, and / or current manufacturing parameters. In some embodiments, predictive data 168 is or includes any anomalies (e.g., an abnormal product, an abnormal component, an abnormal manufacturing equipment 124, abnormal energy usage, etc.) and optionally an indication of one or more causes of the anomaly. In some embodiments, the predictive data 168 is an indicator of change or drift over time in several components, such as manufacturing equipment 124, sensors 126, and measuring instruments 128. In some embodiments, the predictive data 168 is an indication of the end of life of components, such as manufacturing equipment 124, sensors 126, and measuring instruments 128. In some embodiments, the predictive data 168 is an indication of the progress of a processing step being performed (e.g., used for process control).

[0037]

[0046] Executing a manufacturing process that results in defective products can be costly in terms of time, energy, products, parts, manufacturing equipment 124, the cost of identifying defects and discarding defective products, and additional environmental impacts arising from the manufacturing process, characterization, and / or disposal process. By inputting current measurement data 166 indicating epitaxial defects into a prediction system 110, receiving output of prediction data 168, and performing corrective actions based on the prediction data 168, the system 100 can have the technical advantage of avoiding the costs of manufacturing, identifying, and discarding defective products. For example, after the deposition of an epitaxial film but before additional manufacturing processes for producing a finished product, imaging data can be used to screen a substrate to determine whether the epitaxial defects in the substrate fall within a target threshold (e.g., frequency, density, location, classification, etc.). Substrates that do not meet the target performance threshold can be excluded from further processing, thus avoiding the costs associated with generating defective products.

[0038]

[0047] Executing a manufacturing process that results in the failure of components of manufacturing equipment 124 can incur costs such as downtime, product damage, equipment damage, and urgent order replacement components. By inputting current measurement data 166 into the predictive system 110, receiving the output of predictive data 168, and performing corrective actions (e.g., predictive operational maintenance such as component replacement, processing, and cleaning) based on the predictive data 168, system 100 can have the technical advantage of avoiding one or more costs such as unexpected component failure, unexpected downtime, productivity loss, unexpected equipment failure, and product disposal. For example, by monitoring the performance of components such as manufacturing equipment 124, sensors 126, and measuring instruments 128 over time, signs of component degradation, drift, and / or aging can be obtained.

[0039]

[0048] Manufacturing parameters may not be optimal for manufacturing products that have costly consequences such as increased consumption of resources (e.g., energy, coolant, gas, etc.), increased product manufacturing time, increased component failures, increased quantity of defective products, and increased environmental impact. By inputting measurement instructions into the prediction system 110, receiving the output of prediction data 168, and performing corrective actions to update the manufacturing parameters (e.g., setting optimal manufacturing parameters), system 100 may have the technical advantage of using optimal manufacturing parameters (e.g., hardware parameters, process parameters, optimal design) to avoid the costly consequences of suboptimal manufacturing parameters.

[0040]

[0049] Corrective actions may be associated with one or more of the following: computational processing control (CPC), statistical processing control (SPC) (e.g., SPC on electronic components to determine the process under control, SPC to predict the useful life of components, etc.), advanced processing control (APC), model-based processing control, preventive processing maintenance, design optimization, manufacturing parameter updates, manufacturing recipe updates, feedback control, machine learning modifications, etc.).

[0041]

[0050] In some embodiments, the corrective action includes providing an alarm (e.g., an alarm to stop or not perform the manufacturing process if the predictive data 168 indicates a predicted anomaly, such as a product, part, or manufacturing equipment 124). In some embodiments, a machine learning model is trained to monitor the progress of the process execution (e.g., monitor in-sensor data to predict whether the manufacturing process is complete). In some embodiments, the machine learning model may send an instruction to terminate the process execution when the model determines that the process is complete. In some embodiments, the corrective action includes providing feedback control (e.g., modifying manufacturing parameters in response to the predictive data 168 indicating a predicted anomaly). In some embodiments, the implementation of the corrective action includes ensuring that one or more manufacturing parameters are updated. In some embodiments, the corrective action includes excluding one or more substrates from further processing, flagging one or more substrates for additional measurement or other investigation, or similar actions (e.g., substrate screening). In some embodiments, the execution of corrective action may include scheduling and / or performing one or more maintenance processes, such as cleaning or replacing components, chamber cleaning, and chamber seasoning. In some embodiments, the execution of corrective action may include retraining a machine learning model associated with the manufacturing equipment 124. In some embodiments, the execution of corrective action may include training a new machine learning model associated with the manufacturing equipment 124.

[0042]

[0051] The manufacturing parameters 150 may include hardware parameters (e.g., information indicating which components are installed in the manufacturing equipment 124, information indicating component replacement, information indicating the elapsed time of components, information indicating software version or update, etc.) and / or processing parameters (e.g., temperature, pressure, flow rate, current, voltage, gas flow, lift speed, etc.). In some embodiments, corrective actions include ensuring that preventive operational maintenance (e.g., replacement, processing, cleaning, etc. of components of the manufacturing equipment 124) is performed. In some embodiments, corrective actions include ensuring that design optimization is performed (e.g., updating manufacturing parameters, manufacturing processes, manufacturing equipment 124, etc., for an optimized product). In some embodiments, corrective actions include updating the recipe (e.g., changing the timing of manufacturing subsystems that enter idle mode or active mode, changing the setpoints for various characteristic values, etc.).

[0043]

[0052] Each of the prediction server 112, server machine 170, and server machine 180 may include one or more computing devices, such as rack-mount servers, router computers, server computers, personal computers, mainframe computers, laptop computers, tablet computers, desktop computers, graphics processing units (GPUs), and application-specific integrated circuits (ASICs) (e.g., Tensor Processing Units (TPUs)). Processing by the prediction server 112, server machine 170, server machine 180, data store 140, etc., may be performed by cloud computing services, cloud data storage services, etc.

[0044]

[0053] The prediction server 112 may include a prediction component 114. In some embodiments, the prediction component 114 may receive current measurement data 166 and generate output (e.g., prediction data 168) for performing corrective actions related to the manufacturing equipment 124 based on the current data. Receiving current measurement data 166 may include obtaining data from a client device 120. Receiving current measurement data 166 may include obtaining data from a data store 140. In some embodiments, the prediction data 168 may include one or more prediction performance measurements of the processed product. In some embodiments, the prediction data 168 may include a count, classification, density, and / or location map of epitaxial defects represented in substrate image data, prediction characteristics of epitaxial defects based on substrate image data, etc.

[0045]

[0054] In some embodiments, the prediction component 114 can determine epitaxial defects represented in image data (e.g., measurement data 160) based on image processing steps. For example, the prediction component 114 can receive substrate image data. The prediction component 114 can perform preprocessing to improve the usability of the image, such as Gaussian blur and sharpening operations. The prediction component 114 can perform transformation operations (e.g., Fourier transform operations) to transform the image data from the spatial domain to the frequency domain. The prediction component 114 can apply filtering to the frequency domain image data. The filters can be designed to highlight or emphasize one or more target epitaxial defects. In some embodiments, several filters can be applied to the image of the substrate to generate a number of filtered image data for determining various types of epitaxial defects. The prediction component 114 can perform additional image processing techniques, such as returning the image data to the spatial domain and performing image thresholding. Image thresholding may include increasing the contrast of the image to improve feature detection, defect detection, defect classification, etc. The image threshold may include adjusting each pixel with a brightness above the threshold to its maximum brightness (e.g., brightness of 1) and adjusting each pixel with a brightness below the threshold to its minimum brightness (e.g., brightness of 0). The prediction component 114 can perform feature detection processes such as contour detection and circle detection to determine evidence of the presence of epitaxial defects in the image data.

[0046]

[0055] In some embodiments, the performance of epitaxial film defect detection can be adjusted / improved by adjusting one or more processes of epitaxial film defect detection. In some embodiments, defect label data 162 may be provided for adjusting the parameters of epitaxial film defect detection. The defect label data 162 may include information indicating defects, such as one or more substrates or one or more images, generated by a method different from the method used to adjust the defect label data 162. For example, the defect label data 162 may be provided by a subject expert who manually determines the characteristics of defects in epitaxial films. The defect label data 162 may be used to adjust the parameters of the epitaxial film defect detection system to improve the performance of the film defect detection system. Exemplary parameters that can be adjusted include image preprocessing parameters such as Gaussian Blur standard deviation, image filter parameters such as frequency domain filter shape, image threshold parameters such as adjustment thresholds, and feature thresholds where detected features are included in the defect classification.

[0047]

[0056] In some embodiments, the output of the epitaxial film defect detection process may be used for further process tuning, calibration, and / or training. For example, if the epitaxial film defect detection process meets a target performance threshold (e.g., one or more target precision thresholds), the output epitaxial film defect data may be used as defect label data 162 for tuning the parameters of other models. In some embodiments, the defect characterization output by the epitaxial film defect detection system may be used as training input and / or target output for training a machine learning model.

[0048]

[0057] The manufacturing equipment 124 may be associated with one or more machine cleaning models, for example, model 190. The machine learning models associated with the manufacturing equipment 124 can perform many tasks, including process control, classification, and performance prediction. Model 190 may be trained using data related to the manufacturing equipment 124 or the products processed by the manufacturing equipment 124, such as sensor data 142 (e.g., collected by sensor 126), manufacturing parameters 150 (e.g., related to process control of the manufacturing equipment 124), and measurement data 160 (e.g., generated by measurement equipment 128).

[0049]

[0058] In some embodiments, the prediction system 110 may include a machine learning model for predicting unmeasured defect characteristics of the substrate. For example, an imaging technique may be performed to generate images of a subset of the substrate surface. The machine learning model (e.g., performed by the prediction component 114) may be used to predict defect characteristics of unmeasured portions of the substrate, for example, based on training data of other substrates having similar epitaxial film defect characteristics.

[0050]

[0059] One type of machine learning model that can be used to perform some or all of the above tasks is an artificial neural network, such as a deep neural network. An artificial neural network generally includes a feature representation component, which includes a classifier or a regression layer that maps features to a desired output space. A convolutional neural network (CNN) hosts, for example, multiple layers of convolutional filters. In the lower layers, pooling is performed and nonlinearity can be dealt with, and a multilayer perceptron is usually added on top, with the top layer mapping the features extracted by the convolutional layers to produce a decision (e.g., a classification output). CNNs have found particular applicability in the field of image processing, for example, processing substrate image data for epitaxial defect detection.

[0051]

[0060] A recurrent neural network (RNN) is another type of machine learning model. Recurrent neural network models are designed to interpret a set of inputs that are inherently related to each other, such as time trace data or sequential data. The output of the perceptron in an RNN is fed back into the perceptron as input, generating the next output.

[0052]

[0061] Deep learning is a class of machine learning algorithms that use a cascade of multiple layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. Deep neural networks can be trained in supervised (e.g., classification) and / or unsupervised (e.g., pattern analysis) forms. A deep neural network contains a hierarchy of layers, with different layers learning different levels of representation corresponding to different levels of abstraction. In deep learning, each level learns to transform its own input data into a slightly more abstract composite representation. For example, in an image recognition application, the raw input is a matrix of pixels, The first representation layer can abstract pixels and encode edges, The second layer can synthesize and encode the edge arrangement. A third layer can encode higher-order shapes (e.g., teeth, lips, gums), and a fourth layer can recognize the role of scanning. Notably, the deep learning process can learn on its own which features are best placed at which levels. The "deep" in "deep learning" refers to the number of layers transforming the data. More precisely, deep learning systems have a fairly deep CAP (credit assignment path). The CAP is a chain of transformations from input to output. It describes the latent causal relationships between input and output. In the case of a feedforward neural network, the CAP depth can be the network depth, which can be the number of hidden layers plus one. For recurrent neural networks where a signal can propagate through a layer multiple times, the CAP depth is potentially unlimited.

[0053]

[0062] In some embodiments, the prediction component 114 receives current sensor data, current measurement data 166, and / or current manufacturing parameters, performs signal processing to decompose the current data into a set of current data, provides the set of current data as input to a trained model 190, and obtains an output from the trained model 190 indicating prediction data 168. In some embodiments, the prediction component 114 receives measurement data of the substrate (e.g., prediction measurement data based on sensor data) and provides the measurement data to the trained model 190. For example, the current sensor data may include sensor data indicating measurements of the substrate (e.g., shape dimensions). The model 190 may be configured to accept data indicating substrate measurements and generate it as an output prediction characteristic for epitaxial defects in the substrate. In some embodiments, the prediction data indicates measurement data (e.g., prediction of substrate quality). In some embodiments, the prediction data indicates the health of the components. In some embodiments, the prediction data indicates the progress of the process (e.g., used to terminate the process).

[0054]

[0063] In some embodiments, the various models described in relation to Model 190 (e.g., supervised machine learning models, unsupervised machine learning models, etc.) may be combined into a single model (e.g., an ensemble model) or they may be separate models.

[0055]

[0064] Data may travel back and forth between several different models included in Model 190 and the prediction component 114. In some embodiments, some or all of these processes may instead be performed by different devices (e.g., client device 120, server machine 170, server machine 180, etc.). Those skilled in the art will understand that variations in data flow, such as which components perform which processes and which models are provided with which data, are within the scope of this disclosure.

[0056]

[0065] The data store 140 may be memory (e.g., random access memory), drives (e.g., hard drives, flash drives), a database system, a cloud-accessible memory system, or another type of component or device capable of storing data. The data store 140 may include multiple storage components (e.g., multiple drives or multiple databases), and such storage components may be located across multiple computing devices (e.g., multiple server computers). The data store 140 may store sensor data 142, manufacturing parameters 150, measurement data 160, composite data 162, and prediction data 168.

[0057]

[0066] In some embodiments, the prediction system 110 further includes server machines 170 and 180. Server machine 170 includes a dataset generator 172 that can generate datasets (e.g., sets of data inputs and sets of target outputs) for training, validating, and / or testing one or more models 190, which include one or more machine learning models. Some of the operations of the dataset generator 172 are described below in detail with reference to Figures 2A-2B and 4A. In some embodiments, the dataset generator 172 can partition historical data (e.g., historical measurement data 164) into a training set (e.g., 60% of the historical data), a validation set (e.g., 20% of the historical data), and a test set (e.g., 20% of the historical data).

[0058]

[0067] In some embodiments, the prediction system 110 generates multiple sets of features (for example, via a prediction component 114). For example, a first set of features may correspond to a first type of sensor dataset corresponding to each of the datasets (e.g., a training set, a validation set, and a test set) (e.g., a first combination of values ​​from the sensors of the first set, from the measuring instruments of the first set, a first type of substrate imaging technique, etc.), and a second set of features may correspond to a second type of sensor dataset corresponding to each of the datasets (e.g., a second sensor set different from the first sensor set, a second combination of values ​​different from the first combination, a second pattern different from the first pattern).

[0059]

[0068] The server machine 180 includes a training engine 182, a validation engine 184, a selection engine 185, and / or a test engine 186. The engines (e.g., training engine 182, validation engine 184, selection engine 185, and test engine 186) may refer to hardware (e.g., a network, dedicated logic, programmable logic, microcode, processing devices, etc.), software (e.g., instructions executed on a processing device, a general-purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof. The training engine 182 may be capable of training a model 190 using one or more sets of features associated with a training set from a dataset generator 172. The training engine 182 may produce multiple trained models 190, where each trained model 190 corresponds to a distinct set of features from the training set (e.g., sensor data from a distinct set of sensors). For example, the first trained model may be trained using all features (e.g., X1-X5), the second trained model may be trained using a first subset of features (e.g., X1, X2, X4), and the third trained model may be trained using a second subset of features that may partially overlap with the first subset (e.g., X1, X3, X4, and 5). The dataset generator 172 receives the output of the trained models, collects that data into training, validation, and test datasets, and can use those datasets to train a second model (e.g., a machine learning model configured to output predictive data, corrective actions, etc.).

[0060]

[0069] The validation engine 184 may be able to validate the trained models 190 using the features of the corresponding set of validation sets from the dataset generator 172. For example, a first trained machine learning model 190, trained using a first set of features from the training set, may be validated using a first set of features from the validation set. The validation engine 184 can determine the accuracy of each trained model 190 based on the corresponding set of features from the validation set. The validation engine 184 may discard trained models 190 that have an accuracy that does not meet a threshold accuracy. In some embodiments, the selection engine 185 may be able to select one or more trained models 190 that have an accuracy that meets a threshold accuracy. In some embodiments, the selection engine 185 may be able to select the trained model 190 with the highest accuracy.

[0061]

[0070] The test engine 186 can test the trained model 190 using the corresponding set of features from the test set generated by the dataset generator 172. For example, a first trained machine learning model 190, trained using a first set of features from the training set, can be tested using a first set of features from the test set. Based on the test set, the test engine 186 can determine which trained model 190 has the highest accuracy among all the trained models.

[0062]

[0071] In the case of a machine learning model, model 190 may refer to a model artifact created by the training engine 182 using a training set containing data inputs and corresponding target outputs (ground truth for each training input). Patterns in the dataset can be found that map data inputs to target outputs (ground truth), and machine learning model 190 is provided with mappings that capture these patterns. Machine learning model 190 may use one or more of the following: support vector machines (SVMs), radial basis functions (RBFs), clustering, supervised machine learning, semi-supervised machine learning, unsupervised machine learning, k-nearest neighbor algorithm (k-NN), linear regression, random forests, and neural networks (e.g., artificial neural networks, regressive neural networks).

[0063]

[0072] The prediction component 114 can provide current data to the model 190 and run the model 190 on the input to obtain one or more outputs. For example, the prediction component 114 can provide current sensor data 146 to the model 190 and run the model 190 on the input to obtain one or more outputs. The prediction component 114 may be able to determine (e.g., extract) prediction data 168 from the output of the model 190. The prediction component 114 may be able to determine (e.g., extract) confidence data from the output that indicates the confidence that the current sensor data 146 and / or input data of the product manufactured or produced using the manufacturing equipment 124 are accurate predictors of the process related to the input data of the product manufactured or produced using the manufacturing equipment 124 with the current manufacturing parameters. The prediction component 114 or the corrective action component 122 may use the confidence data to determine whether to take corrective action related to the manufacturing equipment 124 based on the prediction data 168.

[0064]

[0073] Confidence data may include or indicate confidence that the prediction data 168 is an accurate prediction of a product or component related to at least a portion of the input data. In one embodiment, the confidence level is a real number from 0 to 1, where 0 indicates no certainty that the prediction data 168 is an accurate prediction of a product processed according to the input data or component health of the components of the manufacturing equipment 124, and 1 indicates absolute confidence that the prediction data 168 is an accurate prediction of a product characteristic processed according to the input data or component health of the components of the manufacturing equipment 124. In response to confidence data indicating confidence below a threshold level for a predetermined number of instances (e.g., percentage of instances, frequency of instances, total number of instances), the prediction component 114 may allow the trained model 190 to be retrained (e.g., based on current sensor data 146, current manufacturing parameters, etc.). In some embodiments, retraining may include generating one or more datasets using historical data (e.g., via a dataset generator 172).

[0065]

[0074] For illustrative purposes, aspects of this disclosure describe, without limitation, training one or more machine learning models 190 using historical data (e.g., historical sensor data, historical manufacturing parameters) and inputting current data (e.g., current sensor data, current manufacturing parameters, and current measurement data) into one or more trained machine learning models to determine predictive data 168. In other embodiments, a heuristic model, a physics-based model, or a rule-based model is used to determine predictive data 168 (e.g., without using a trained machine learning model). In some embodiments, such models may be trained using historical data. The predictive component 114 may monitor historical sensor data 144, historical manufacturing parameters, and measurement data 160. Any of the information described with respect to the data inputs 210A to 210B in Figures 2A to 2B may be monitored or otherwise used in a heuristic model, a physics-based model, or a rule-based model.

[0066]

[0075] In some embodiments, the functions of client device 120, prediction server 112, server machine 170, and server machine 180 may be provided by fewer machines. For example, in some embodiments, server machines 170 and 180 may be integrated into a single machine, while in some other embodiments, server machine 170, server machine 180, and prediction server 112 may be integrated into a single machine. In some embodiments, client device 120 and prediction server 112 may be integrated into a single machine. In some embodiments, the functions of client device 120, prediction server 112, server machine 170, server machine 180, and data store 140 may be performed by a cloud-based service.

[0067]

[0076] In general, functions described in one embodiment as being performed by the client device 120, prediction server 112, server machine 170, and server machine 180 can, where appropriate, also be performed on the prediction server 112 in other embodiments. Furthermore, functions belonging to a particular component may be performed by different components or by multiple components working together. For example, in some embodiments, the prediction server 112 may determine corrective actions based on the prediction data 168. In another example, the client device 120 may determine the prediction data 168 based on the output from a trained machine learning model.

[0068]

[0077] Furthermore, the functionality of a particular component can be performed by different or multiple components working together. One or more of the prediction server 112, server machine 170, or server machine 180 may be accessed as a service provided to other systems or devices via an appropriate application programming interface (API).

[0069]

[0078] In some embodiments, “User” may be represented as a single individual. However, other embodiments of the present disclosure include cases where “User” is an entity controlled by multiple users and / or automated sources. For example, a collection of individual users formed as a group of administrators may be considered “User.”

[0070]

[0079] Figures 2A and 2B show block diagrams of exemplary dataset generators 272A and 272B (e.g., dataset generator 172 in Figure 1) for creating datasets for training, testing, validating, etc., models (e.g., model 190 in Figure 1) according to several embodiments. Each dataset generator 272 may be part of the server machine 170 in Figure 1. In some embodiments, several machine learning models associated with manufacturing equipment 124 may be trained, used, and maintained (e.g., within the manufacturing facility). Each model may be associated with one dataset generator 272, and multiple models may share dataset generators 272, etc.

[0071]

[0080] Figure 2A shows a system 200A that includes a dataset generator 272A for creating datasets for one or more supervised models (e.g., model 190 in Figure 1). The dataset generator 272A can use historical data to create datasets (e.g., data input 210A, target output 220A). In some embodiments, an unsupervised machine learning model can be trained using a dataset generator similar to the dataset generator 272A. For example, the target output 220A may not be generated by the dataset generator 272A.

[0072]

[0081] The dataset generator 272A can generate datasets for training, testing, and validating a model. In some embodiments, the dataset generator 272A can generate datasets for a machine learning model. In some embodiments, the dataset generator 272A can generate datasets for training, testing, and / or validating a model configured to generate characterizations of epitaxial defects in a substrate. The machine learning model is provided with a set of historical measurement data 264A as data input 210A. The machine learning model may be configured to accept the measurement data as input data and generate characterizations of epitaxial film defects as output.

[0073]

[0082] The dataset generator 272A can be used to generate data for any type of machine learning model that takes input measurement data. The dataset generator 272A can be used to generate data for a machine learning model that generates predictive measurement data of a substrate. The dataset generator 272A can be used to generate data for a machine learning model configured to provide substrate screening. The dataset generator 272A can be used to generate data for a machine learning model configured to identify product anomalies and / or failures of processing equipment. The dataset generator 272A can be used to generate data for a machine learning model configured to predict epitaxial defect characteristics of a substrate.

[0074]

[0083] In some embodiments, the dataset generator 272A generates a dataset (e.g., a training set, a validation set, a test set) containing one or more data inputs 210A (e.g., a training input, a validation input, a test input). The data inputs 210A may be provided to the training engine 182, the validation engine 184, or the test engine 186. The dataset may be used to train, validate, or test a model.

[0075]

[0084] In some embodiments, data input 210A may include one or more datasets. For example, system 200A may include one or more datasets, such as measurement data from one or more types of measuring instruments, combinations of data from one or more types of measuring instruments, measurement data from one or more locations on the associated substrate, and patterns from measurement data from one or more types of imaging.

[0076]

[0085] In some embodiments, data input 210A may include one or more datasets. For example, system 200A may generate a set of historical measurement data which may include one or more of the following: measurement data of a group of device dimensions (e.g., including device height and width, but not optical data or surface roughness, etc.), measurement data obtained from one or more types of sensors, combinations of measurement data obtained from one or more types of sensors, patterns from measurement data, measurement data preprocessed in various ways, measurement data generated by various instruments and / or techniques, etc. The set of data input 210A may include data describing various aspects of manufacturing (e.g., combinations of measurement data and sensor data, combinations of measurement data and manufacturing parameters, combinations of some measurement data, some manufacturing parameter data, and some sensor data, etc.).

[0077]

[0086] In some embodiments, the dataset generator 272A may generate a first data input corresponding to a first set of historical measurement data 264A for training, validating, or testing a first machine learning model. The dataset generator 272A may generate a second data input corresponding to a second set of historical measurement data (e.g., a set of historical measurement data 264B, not shown) for training, validating, or testing a second machine learning model. Further sets of historical measurement data may be used to generate further machine learning models. Any number of sets of historical measurement data may be used up to the last set, the set of historical measurement data 264N (where N represents any target quantity such as a dataset or model), when generating any number of machine learning models.

[0078]

[0087] In some embodiments, the dataset generator 272A generates a dataset (e.g., a training set, a validation set, a test set) which may include one or more data inputs 210A (e.g., a training input, a validation input, a test input) and one or more target outputs 220A corresponding to the data inputs 210A. The dataset may further include mapping data that maps the data inputs 210A to the target outputs 220A. In some embodiments, the dataset generator 272A can generate data for training a machine learning model configured to characterize epitaxial defects contained in the input image data, for example, as output defect data 268. In some embodiments, the output defect data 268 may be generated by another system for determining epitaxial defects in a substrate, and the target outputs 220A may be output by another model, algorithm, prediction system, etc. The output defect data 268 may be generated by the prediction system 110 in Figure 1. The data inputs 210A may also be referred to as “features,” “attributes,” or “information.” In some embodiments, the dataset generator 272A can provide the dataset to the training engine 182, validation engine 184, or test engine 186, and the dataset is used to train, validate, or test a machine learning model (e.g., one of the machine learning models included in the synthetic data generator 174, model 190, ensemble model 190, etc.).

[0079]

[0088] Figure 2B shows a block diagram of an exemplary dataset generator 272B for creating datasets for supervised models configured to generate an index of predicted epitaxial defect characteristics of a substrate, according to several embodiments. System 200B, including dataset generator 272B (e.g., dataset generator 172 in Figure 1), creates datasets for one or more machine learning models (e.g., model 190 in Figure 1). Dataset generator 272B may create datasets (e.g., data input 210B) using historical data.

[0080]

[0089] The exemplary dataset generator 272B is configured to generate a dataset for a machine learning model that takes a portion of the substrate as input image data and outputs a second portion of the substrate, for example, the entire surface of the substrate, as output predicted defect data. Similar dataset generators (or similar operations of dataset generator 272B) can be used for machine learning models configured to perform different functions, such as a machine learning model that receives as input measurement image data and outputs predicted root causes of epitaxial defects, or a machine learning model that receives as input image data of the substrate and outputs recommended corrective actions. Dataset generator 272B may share features and / or functions with dataset generator 272A.

[0081]

[0090] The dataset generator 272B can generate datasets for training, testing, and validating machine learning models. The machine learning model includes a set of substrate image data 262A as a data input 210B (e.g., image data of one or more parts of a substrate, location data associated with the image data identifying the relevant substrate locations, etc.). The machine learning model may include two or more separate models (e.g., the machine learning model may be an ensemble model). The machine learning model may be configured to generate output data that includes predicted substrate defect data (e.g., for parts of the substrate not included in the image data provided as input). In some embodiments, training may not include providing a target output to the machine learning model. The dataset generator 272B can generate datasets for training an unsupervised machine learning model (e.g., a model configured to generate output clustering data, outlier detection data, anomaly detection data, etc.). The dataset generator 272B can generate an accompanying target output 220B for the data input 210B. The target output 220B may include output substrate defect data 269 (e.g., an output showing defects in parts of the substrate not included in the input data) for configuring the machine learning model to predict epitaxial defect characteristics (e.g., defect classification, count, density, location, etc.) in parts of the substrate not associated with the image data provided to the machine learning model as input.

[0082]

[0091] In some embodiments, the dataset generator 272B generates a dataset (e.g., a training set, a validation set, a test set) containing one or more data inputs 210B (e.g., a training input, a validation input, a test input). The data inputs 210B may also be referred to as “features,” “attributes,” or “information.” In some embodiments, the dataset generator 272B can provide the dataset to the training engine 182, the validation engine 184, or the test engine 186, and the dataset is used to train, validate, or test a machine learning model (e.g., model 190 in Figure 1). Several embodiments of generating the training set are described further with reference to Figure 4A.

[0083]

[0092] In some embodiments, the dataset generator 272B can generate a first data input corresponding to a first set of substrate image data 262A for training, validating, or testing a first machine learning model, and the dataset generator 272A can generate a second data input corresponding to a second dataset (e.g., a set of substrate image data 262B, not shown) for training, validating, or testing a second machine learning model. Further sets of substrate image data may be used when generating further machine learning models. Any number of sets of substrate image data may be used up to the final set, a set of substrate image data 262N (where N represents any target quantity such as a dataset or model), when generating any number of machine learning models.

[0084]

[0093] The data input 210B for training, validating, or testing a machine learning model may include information for a specific manufacturing chamber (e.g., for a particular substrate manufacturing machine). In some embodiments, the data input 210B may include information for a specific type of manufacturing machine, e.g., manufacturing machine that shares specific characteristics. The data input 210B may include data related to a specific type of device (e.g., manufactured with an intended function, design, or specific recipe). By training a machine learning model based on types such as equipment, devices, and recipes, the trained model may be able to generate plausible synthetic sensor data in several settings (e.g., for several different facilities, products, etc.).

[0085]

[0094] In some embodiments, a dataset can be generated, and a machine learning model can be trained, validated, or tested using the dataset, followed by further training, validation, testing, or tuning of the model (e.g., tuning weights or parameters related to the model's input data, such as connection weights in a neural network).

[0086]

[0095] Figure 3 is a block diagram showing a system 300 for generating output data (e.g., prediction data 168 in Figure 1) according to several embodiments. In some embodiments, the system 300 may be used in conjunction with a machine learning model (e.g., model 190 in Figure 1) configured to generate predictions of epitaxial film defects. In some embodiments, the system 300 may be used in conjunction with a machine learning model to determine corrective actions related to manufacturing equipment. In some embodiments, the system 300 may be used in conjunction with a machine learning model to determine failures of manufacturing equipment. In some embodiments, the system 300 may be used in conjunction with a machine learning model to predict epitaxial film defect characteristics of a substrate. The system 300 may be used in conjunction with a machine learning model associated with a manufacturing system that has different functions than those listed.

[0087]

[0096] In block 310, system 300 (for example, a component of prediction system 110 in Figure 1) performs data partitioning (for example, via a dataset generator 172 on server machine 170 in Figure 1) of data used for training, validating, and / or testing machine learning models. In some embodiments, the training epitaxial film data 364 includes historical data such as historical measurement data, historical classification data (for example, classification of epitaxial defects associated with the same substrate represented by the measurement data), and historical microscopic image data. In some embodiments, the classification data may be provided by an epitaxial film defect system, for example, prediction system 110 in Figure 1. The training epitaxial film data 364 can be partitioned in block 310 to generate a training set 302, a validation set 304, and a test set 306. For example, the training set may be 60% of the training data, the validation set may be 20% of the training data, and the test set may be 20% of the training data.

[0088]

[0097] The generation of the training set 302, validation set 304, and test set 306 can be tailored for specific applications. For example, the training set may consist of 60% of the training data, the validation set 20% of the training data, and the test set 20% of the training data. The system 300 can generate multiple sets of features for each of the training set, validation set, and test set. For example, if the training epitaxial membrane data 364 includes sensor data containing features derived from sensor data from 20 sensors (e.g., sensor 126 in Figure 1) and 10 manufacturing parameters (e.g., manufacturing parameters corresponding to the same processing execution as the sensor data from the 20 sensors), the sensor data may be split into a first set of features containing sensors 1-10 and a second set of features containing sensors 1-10. The manufacturing parameters may also be split into sets such as a first set of manufacturing parameters containing parameters 1-5 and a second set of manufacturing parameters containing parameters 6-10. Target inputs, target outputs, both, or both may be split into sets. Multiple models can be trained on different datasets.

[0089]

[0098] In block 312, system 300 performs model training (e.g., via training engine 182 in Figure 1) using training set 302. Training of machine learning models and / or physically based models (e.g., digital twins) can be achieved in a supervised learning manner. A supervised learning manner involves providing a training dataset containing inputs labeled through the model, observing its output, defining the error (by measuring the difference between the output and the label value), and using techniques such as depth gradient structuring to adjust the weighting of the model so that the error is minimized. In many applications, repeating this process across many labeled inputs in the training dataset yields a model that can produce the correct output when presented with inputs different from those present in the training dataset. In some embodiments, training of a machine learning model can be achieved in an unsupervised manner. For example, no labels or classifications may be provided during training. The unsupervised model may be configured to perform anomaly detection, result clustering, etc.

[0090]

[0099] For each training data item in the training dataset, the training data item can be input into a model (e.g., a machine learning model). The model can then process the input training data item (e.g., an image of a substrate) to generate an output. The output may include, for example, a characterization of epitaxial defects in the substrate. The output can be compared to the labels of the training data item (e.g., epitaxial film defect data classified by another reliable method).

[0091]

[0100] The processing logic may then compare the generated output (e.g., predicted epitaxial defect data) with the labels included in the training data items (e.g., labeled epitaxial defect data). The processing logic determines the error (i.e., classification error) based on the difference between the output and the labels. The processing logic adjusts one or more weights and / or values ​​of the model based on the error.

[0092]

[0101] When training a neural network, an error term or delta may be determined for each node in the artificial neural network. Based on the above error, the artificial neural network adjusts one or more of its own parameters (weights for one or more inputs of a node) for one or more of its own nodes. Parameters may be updated in a backpropagation manner, such that the top layer nodes are updated first, followed by the nodes of the next layer, and so on. An artificial neural network contains multiple layers of "neurons," each layer receiving values ​​as input from the neurons of the previous layer. The parameters for each neuron include weights associated with the values ​​received from each neuron of the previous layer. Therefore, adjusting parameters may involve adjusting the weights assigned to each input for one or more neurons in one or more layers of the artificial neural network.

[0093]

[0102] System 300 can train multiple models using multiple sets of features from the training set 302 (e.g., a first set of features from the training set 302, a second set of features from the training set 302, etc.). For example, System 300 can generate a first trained model using a first set of features in the training set (e.g., sensor data from sensors 1-10, measured values ​​1-10, etc.) and generate a second trained model using a second set of features in the training set (e.g., sensor data from sensors 11-20, measured values ​​11-20, etc.). In some embodiments, the first and second trained models can be combined to generate a third trained model (e.g., which may be a better predictor than the first or second trained model itself). In some embodiments, the sets of features used for comparing the models may overlap (e.g., a first set of features which are sensor data from sensors 1-15 and a second set of features which are sensors 5-20). In some embodiments, hundreds of models can be generated, including models with various feature substitutions and model combinations.

[0094]

[0103] In block 314, system 300 performs model validation (e.g., via validation engine 184 in Figure 1) using validation set 304. System 300 can validate each of the trained models using the corresponding set of features in validation set 304. For example, system 300 may validate a first trained model using a first set of features in the validation set (e.g., sensor data from sensors 1-10 or measured values ​​1-10) and a second trained model using a second set of features in the validation set (e.g., sensor data from sensors 11-20 or measured values ​​11-20). In some embodiments, system 300 can validate hundreds of models generated in block 312 (e.g., models with various permutations of features, combinations of models, etc.). In block 314, system 300 can determine the accuracy of each of one or more trained models (e.g., via model validation) and determine whether one or more of the trained models have an accuracy that meets a threshold accuracy. In response to the determination that none of the trained models have the accuracy to meet the threshold, the flow returns to block 312, where system 300 trains models using different sets of features from the training set. In response to the determination that one or more of the trained models have the accuracy to meet the threshold, the flow continues to block 316. System 300 may discard trained models that have accuracy below the threshold (for example, based on the validation set).

[0095]

[0104] In block 316, system 300 performs model selection (e.g., via selection engine 185 in Figure 1) to determine which of the one or more trained models that satisfy the threshold precision has the highest precision (e.g., the selected model 308 based on the validation in block 314). In response to the determination that two or more trained models that satisfy the threshold precision have the same precision, the flow can return to block 312, where system 300 performs model training using a further refined training set corresponding to a further refined set of features to determine the trained model with the highest precision.

[0096]

[0105] In block 318, system 300 performs a model test (e.g., via the test engine 186 in Figure 1) using test set 306 to test the selected model 308. System 300 can test a first trained model using a first set of features in the test set (e.g., sensor data from sensors 1-10) to determine if the first trained model meets the threshold accuracy. Determining whether the first trained model meets the threshold accuracy can be obtained based on the first set of features in test set 306. In response that the accuracy of the selected model 308 does not meet the threshold accuracy, the flow proceeds to block 312, where system 300 performs model training (e.g., retraining) using a different training set corresponding to a different set of features. If the selected model 308 is overfitted to training set 302 and / or validation set 304, the accuracy of the selected model 308 may not meet the threshold accuracy. The accuracy of the selected model 308 may not meet the threshold accuracy if the selected model 308 is not applicable to other datasets, including the test set 306. Training using different features may include training using data from different sensors, different manufacturing parameters, etc. In response to the determination that the selected model 308 has accuracy that meets the threshold accuracy based on the test set 306, the flow proceeds to block 320. In at least block 312, the model can learn patterns in the training data in order to make predictions. In block 318, the system 300 can apply the model to the remaining data (e.g., the test set 306) to test the predictions.

[0097]

[0106] In block 320, system 300 receives current data 322 and uses a trained model (e.g., selected model 308) to determine (e.g., extract) predictive data 324 from the output of the trained model. Current data 322 may be image data related to a process, step, substrate, or operation of the object. Current data 322 may be image data of at least a portion of the substrate in question. Current data 322 may be measurement data of the substrate in question. Current data 322 may be manufacturing parameters related to a process such as development, redevelopment, or investigation. Corrective actions related to the manufacturing equipment 124 in Figure 1 may be performed taking predictive data 324 into consideration. In some embodiments, current data 322 may correspond to the same type of features in historical data used to train the machine learning model. In some embodiments, current data 322 may correspond to a subset of the types of features in historical data used to train the selected model 308. For example, a machine learning model may be trained using several manufacturing parameters and configured to produce an output based on a subset of the manufacturing parameters.

[0098]

[0107] In some embodiments, the performance of a machine learning model trained, validated, and tested by system 300 may degrade. For example, the manufacturing system associated with the trained machine learning model may undergo gradual or abrupt changes. Changes in the manufacturing system may degrade the performance of the trained machine learning model. A new model may be generated to replace the degraded machine learning model. The new model may be generated by modifying the old model through retraining.

[0099]

[0108] Generating a new model may include providing additional training data 346. Generating a new model may further include providing current data 322, for example, data used by the model to make predictions. In some embodiments, the current data 322 provided for generating a new model may be marked with an indication of the accuracy of the predictions generated by the model based on the current data 322. The additional training data 346 may be provided in block 312 for model training, such as generating one or more new machine learning models, updating, retraining, and / or improving a selected model 308.

[0100]

[0109] In some embodiments, one or more of steps 310-320 may occur in various orders and / or in conjunction with other steps not presented and described herein. In some embodiments, one or more of steps 310-320 may not be performed. For example, in some embodiments, one or more of the data partitioning of block 310, model verification of block 314, model selection of block 316, or model testing of block 318 may not be performed.

[0101]

[0110] Figure 3 shows a system configured to train, validate, test, and use one or more machine learning models. The machine learning models are configured to accept data as input (e.g., setpoints provided for manufacturing equipment, sensor data, measurement data, etc.) and to provide data as output (e.g., predictive data, corrective action data, classification data, etc.). Block splitting, training, validation, selection, testing, and use of the system 300 can be performed in a similar manner to training a second model using different types of data. Retraining can also be performed using the current data 322 and / or additional training data 346.

[0102]

[0111] Figures 4A to 4C are flowcharts of methods 400A to 400C. Figures 4A to 4C relate to performing the determination of epitaxial film defects in a substrate, including defect classification, defect counting, defect prediction, and recommendation of corrective actions based on epitaxial film defects. One or more of Figures 4A to 4C may be associated with the training and use of a machine learning model, depending on the particular embodiment. Methods 400A to 400C may be performed by processing logic which may include hardware (e.g., circuits, dedicated logic, programmable logic, microcode, processing devices, etc.), software (e.g., instructions executed on a processing device, general-purpose computer system, or dedicated machine), firmware, microcode, or a combination thereof. In some embodiments, methods 400A to 400C may be performed in part by a prediction system 110. Method 400A may be performed in part by a prediction system 110 (e.g., server machine 170 and dataset generator 172 in Figure 1, dataset generators 272A-B in Figures 2A-B). The prediction system 110 may use Method 400A to generate datasets for at least one of training, validating, or testing a machine learning model, according to embodiments of the present disclosure. Methods 400B-C may be performed by a prediction server 112 (e.g., prediction component 114) and / or server machine 180 (e.g., the training, validation, and testing steps may be performed by server machine 180). In some embodiments, a non-temporary machine-readable storage medium stores instructions, which, when executed by a processing device (e.g., prediction system 110, server machine 180, prediction server 112, etc.), cause the processing device to perform one or more of Methods 400A-400C.

[0103]

[0112] For the sake of simplicity, methods 400A–400C are illustrated and described as a series of steps. However, the steps according to this disclosure may be performed in various orders and / or simultaneously with other steps not presented and described herein. Furthermore, not all illustrated steps are performed in order to carry out methods 400A–400C according to the disclosed subject matter. In addition, those skilled in the art will understand and recognize that methods 400A–400C may alternatively be represented as a series of interrelated states via a state diagram or events.

[0104]

[0113] Figure 4A is a flowchart of Method 400A for generating a dataset for a machine learning model, according to several embodiments. Referring to Figure 4A, in some embodiments, in block 401, the processing logic implementing Method 400A initializes the training set T to an empty set.

[0105]

[0114] In block 402, the processing logic generates a first data input (e.g., a first training input, a first validation input) which may include one or more of the following: sensors, manufacturing parameters, measurement data, etc. In some embodiments, the first data input may include a first set of features for the type of data, and the second data input may include a second set of features for the type of data (as described, for example, with respect to Figure 3). In some embodiments, the data input includes historical measurement data. In some embodiments, the data input includes historical image data including signs of epitaxial film defects.

[0106]

[0115] In some embodiments, in block 403, the processing logic optionally generates a first target output for one or more of the data inputs (e.g., a first data input). In some embodiments, the inputs include one or more measurement values, and the target output is one or more characteristics of an epitaxial defect associated with the measurement values. In some embodiments, the first target output of a machine learning model is predictive data. In some embodiments, the input data may be in the form of data indicating an epitaxial defect, and the target output may be a list of components that are likely to be defective, as in the case of a machine learning model configured to identify a defective manufacturing system. In some embodiments, no target output is generated (e.g., an unsupervised machine learning model that does not need to provide a target output and can group or find correlations in the input data).

[0107]

[0116] In block 404, the processing logic optionally generates mapping data indicating input / output mappings. The input / output mapping (or mapping data) may refer to data inputs (e.g., one or more of the data inputs described herein), target outputs for the data inputs, and associations between one or more data inputs and target outputs. In some embodiments, such as those relating to machine learning models for which no target outputs are provided, block 404 may not be executed.

[0108]

[0117] In block 405, the processing logic adds the mapping data generated in block 404 to the dataset T, in some embodiments.

[0109]

[0118] In block 406, the processing logic branches based on whether the dataset T is sufficient for at least one of the training, validation, and / or testing of a machine learning model (e.g., synthetic data generator 174 or model 190 in Figure 1). If it is sufficient, the process execution proceeds to block 407; otherwise, it returns to block 402 and the process execution continues. Note that in some embodiments, the sufficiency of dataset T may be determined simply based on the number of inputs, in some embodiments it may be mapped to the outputs in the dataset, while in some other embodiments the sufficiency of dataset T may be determined based on one or more other criteria (e.g., a measure of the diversity of data examples, precision, etc.) in addition to the number of inputs.

[0110]

[0119] In block 407, the processing logic provides a dataset T (for example, to a server machine 180) to train, validate, and / or test a machine learning model 190. In some embodiments, dataset T is a training set and is provided to the training engine 182 of the server machine 180 to perform training. In some embodiments, dataset T is a validation set and is provided to the validation engine 184 of the server machine 180 to perform validation. In some embodiments, dataset T is a test set and is provided to the test engine 186 of the server machine 180 to perform testing. For a neural network, for example, input values ​​of a given input / output mapping (e.g., numerical values ​​associated with data input 210A) are input to the neural network, and output values ​​of the input / output mapping (e.g., numerical values ​​associated with target output 220A) are stored in the output nodes of the neural network. The connection weights in the neural network are then adjusted according to a learning algorithm (e.g., backpropagation), and this procedure is repeated for other input / output mappings in dataset T. After block 407, the model (e.g., model 190) may be trained using the training engine 182 of server machine 180, validated using the validation engine 184 of server machine 180, or tested using the test engine 186 of server machine 180. The trained model can be implemented by the prediction component 114 (of the prediction server 112) to generate prediction data 168 for performing signal processing or for performing corrective actions related to the manufacturing equipment 124.

[0111]

[0120] Figure 4B is a flowchart of method 400B for performing corrective actions considering the determination of epitaxial film defects, according to several embodiments. In block 410, the processing logic optionally generates first image data of the substrate. Generating first image data of the substrate may include depositing an epitaxial film on the substrate. Generating first image data of the substrate may include performing etching of the epitaxial film, for example, decorating epitaxial defects by plasma etching or chemical etching. Decorating epitaxial defects may make them more clearly visible in images of the substrate, for example, microscopic images. Generating first image data may include capturing one or more images of the substrate using optical microscopy, scanning electron microscopy, or transmission electron microscopy.

[0112]

[0121] In block 412, the processing logic acquires a first image data of the substrate including the epitaxial film. The epitaxial film may be a film deposited on a semiconductor substrate. The epitaxial film may also be a film deposited on a silicon substrate. The epitaxial film may contain one or more defects. The epitaxial film may contain etching pit defects, cross-hatch defects, etc. The epitaxial film may consist of a single layer (e.g., a single epitaxial layer) or multiple layers. The epitaxial film may be doped. The epitaxial film may not be doped. The epitaxial film may contain silicon (e.g., n-type doped silicon, p-type doped silicon), silicon germanium, silicon carbide, gallium nitride, aluminum nitride, gallium arsenide, aluminum gallium nitride, or silicon arsenide. The epitaxial film may contain carbon, boron, arsenic, antimony, tin, indium, phosphorus, etc. The epitaxial film may contain a silicon / silicon-germanium superlattice.

[0113]

[0122] In block 414, the processing logic applies a frequency domain filter to the first image data to obtain filtered image data. The frequency domain filter may be designed to highlight one or more types of epitaxial defects in the image data. Multiple frequency domain filters may be applied (for example, to different copies of the first image data) to determine different classes of epitaxial defects. The frequency filter may include a two-dimensional high-pass filter. The frequency filter may include a low-pass filter. The frequency filter may be tuned to improve epitaxial defect detection of the processing device.

[0114]

[0123] Further processing of image data may be performed. For example, various image processing techniques to improve feature detection may be applied to the image data, such as sharpening, smoothing, Gaussian blurring, and artifact removal. Image data may be provided as spatial domain data and may be converted to frequency domain data (e.g., via Fourier transform, fast Fourier transform, etc.). Filtered image data may be converted back to spatial domain. Image thresholding may be performed to produce high-contrast images, for example, by adjusting all pixels of an image above a threshold brightness to maximum brightness and all pixels of an image below a threshold brightness to minimum brightness.

[0115]

[0124] In block 416, the processing logic determines the number of epitaxial defects represented in the first image by performing feature detection on the filtered image data. Feature detection may be used to determine the portion of the image data that exhibits epitaxial defects. Feature detection may include the Hough transform. The Hough transform is used to identify shapes or patterns in an image by transforming the pixel data of the image into another representation called Hough space, where the target shape in the original image is represented by a set of parameters in Hough space. Accumulators are used to find related points corresponding to the target shape. The Hough transform can be used to detect circles in the image data, which can be used to determine the presence of etching pits in the epitaxial film. Feature detection may include contour counting. Contour counting may involve using techniques such as Canny edge detection to find connected regions of pixels in the image that satisfy threshold similarity conditions such as luminance similarity or color similarity. Contour counting may be used to determine the location and number of cross-hatchings in the image data of the epitaxial film.

[0116]

[0125] In block 418, the processing logic takes into account the number of epitaxial defects and performs corrective actions. The corrective actions may include screening the substrate for additional processing. The corrective actions may include screening the substrate for the execution of additional processing steps. The corrective actions may include scheduling maintenance of the processing chamber associated with the substrate (e.g., the chamber used to deposit the epitaxial film on the substrate). The corrective actions may include performing or initiating maintenance on the processing chamber. The corrective actions may include updating the processing recipe. The corrective actions may include updating one or more parameters of the processing chamber (e.g., manufacturing parameters of the processing chamber). The corrective actions may include providing a warning to the user.

[0117]

[0126] Figure 4C is a flowchart of Method 400C for training and utilizing a machine learning model related to epitaxial defect detection of a target substrate, according to several embodiments. In block 420, the processing logic acquires a first set of multiple image data of the substrate, including the epitaxial film. The image data may include images of the substrate collected by measuring instruments, such as microscope data, scanning electron microscope data, transmission electron microscope data, etc.

[0118]

[0127] In block 422, the processing logic determines whether the substrate has epitaxial defects. Determining whether the substrate has epitaxial defects may involve applying one or more of the processes described in relation to Figure 4B. For example, determining whether the substrate has epitaxial defects may involve converting the image data to the frequency domain, applying one or more frequency domain filters, and performing feature detection on the filtered frequency domain data.

[0119]

[0128] Block 424 optionally includes a step for determining epitaxial defects. Determining epitaxial defects may include classifying epitaxial defects. Classifying epitaxial defects may include labeling defects as etching pits or cross-hatch defects. Classifying epitaxial defects may be performed as part of a step for training a machine learning model for epitaxial defect classification.

[0120]

[0129] Block 426 optionally includes a step for determining epitaxial defects. Determining epitaxial defects may include determining the density of epitaxial defects. Determining the density of epitaxial defects may be performed as part of a step for training a machine learning model for predicting epitaxial defect density. Further steps related to epitaxial defects may be performed, for example, for later use when training a machine learning model for a target function. These steps may include locating the epitaxial defects, identifying the root cause of the epitaxial defects, and determining corrective actions in consideration of the epitaxial defects.

[0121]

[0130] In block 428, the processing logic trains a machine learning model to determine epitaxial defects in a target substrate. Training the machine learning model includes providing a first set of image data as training inputs. Training the machine learning model also includes providing data indicating epitaxial defects in the substrate as a target output. In some embodiments, for example, to train a machine learning model to predict defect density in unimaged portions of the substrate, the training inputs may include image data of one or more first portions of the substrate, and the target output data may include image data of one or more second portions of the substrate that are different from the first portions.

[0122]

[0131] In blocks 430 and 432, the processing logic optionally performs inference operations on the trained machine learning model. In block 430, the processing logic optionally provides the trained machine learning model with image data showing a first portion of the target substrate.

[0123]

[0132] In block 432, the processing logic obtains output from a trained machine learning model. The output shows one or more of the following: epitaxial defect classification, number of epitaxial defects, epitaxial defect density, epitaxial defect location, epitaxial defects in the second part of the target substrate, root cause of the epitaxial defect, or recommended corrective actions related to the epitaxial defect.

[0124]

[0133] Figure 5A is a block diagram of a flow 500A for generating epitaxial film defect data according to several embodiments. The flow may begin in an epitaxial chamber 502, for example, in a processing chamber of a manufacturing system configured to deposit an epitaxial film onto a substrate (e.g., a semiconductor wafer). In the epitaxial chamber 502, one or more layers of an epitaxial film may be deposited on the substrate. One or more materials may be deposited on the substrate. In some embodiments, the substrate having the epitaxial film may be further etched, for example, to decorate the substrate. In some embodiments (for example, if the substrate is designed to be etched after one or more epitaxial film deposition steps before being offered for additional substrate processing), the substrate may be a substrate manufactured as a final device, a usable / commercial semiconductor, etc. In some embodiments, the substrate may be a test substrate for characterizing, for example, manufacturing equipment, manufacturing recipes, substrate designs, substrate materials, etc.

[0125]

[0134] The flow is followed by one or more imaging tools 504. The imaging tools may be part of a measurement system. The imaging tools may be part of a standalone measurement system, an integrated measurement system, an inline measurement system, etc. The imaging tools may include optical imaging tools, electron microscopy tools, etc. The imaging tools 504 may be used to generate one or more images of the substrate. The images may include an image of the entire surface of the substrate. The images may include an image of a portion of the surface of the substrate. The images may include images of a target portion of the substrate having different characteristics (e.g., different distances from the center) to improve image-based predictions.

[0126]

[0135] The flow then proceeds to image processing 506. Image processing may proceed through an image processing flow that includes preprocessing 510, feature detection 512, batch processing 514, and defect counting 516. Alternatively, or additionally, image processing may proceed through machine learning processing 518. In some embodiments, the output of the image processing flow may be used to improve machine learning processing 518, such as by providing labeled training data for machine learning processing 518.

[0127]

[0136] Preprocessing 510 may include optional processing to adjust image data for improved feature detection, epitaxial film defect detection, epitaxial defect prediction, etc. Preprocessing may include, for example, noise reduction, one or more domain transformations, filtering, smoothing, thresholding, etc., to improve image quality and / or improve the determination of epitaxial defects. Feature detection 512 may include processing to identify features in the image data that indicate epitaxial defects, such as etching pits and cross-hatching.

[0128]

[0137] Batch processing 514 may involve processing a large number of images. A large number of images may be associated with a single substrate. For example, imaging a single substrate at a resolution suitable for epitaxial film defect detection may involve generating thousands of images. A large number of images may be associated with multiple substrates.

[0129]

[0138] The defect count of 516 may include calculating the number of defects detected, classifying the detected defects by type, and calculating the local or global defect density.

[0130]

[0139] The machine learning process 518 can be used to generate results similar to those of the image processing flow. The machine learning process 518 can be used in addition to or instead of the image processing flow. Using the machine learning process 518, output data can be generated based on the image generated by the imaging tool 504. The machine learning process 518 can generate prediction results such as predicted defect density, location, and map for parts of the substrate not represented in the imaging data. The machine learning process 518 can generate predictions of the causes of epitaxial defects, recommended corrective actions, etc.

[0131]

[0140] The flow may then proceed to user interface 508. User interface 508 may include, for example, a graphical user interface for presenting the results of epitaxial defect detection to the user. User interface 508 may include one or more interactive elements. User interface 508 may include data presentation. User interface 508 may include one or more visualizations, such as visualizations of detected and / or predicted epitaxial defects. User interface 508 may include the ability to provide the user with one or more alerts, such as the number of epitaxial defects, global density, and / or local density that meets threshold conditions, and recommended corrective actions.

[0132]

[0141] Figure 5B shows an exemplary epitaxial defect map 500B of a substrate according to several embodiments. The epitaxial defect map 500B may be provided to the user via a user interface. The epitaxial defect map 500B can efficiently provide the user with information such as the density, location, severity, and classification of epitaxial defects. Further information, such as numerical values ​​for the number and density of defects and the coordinate locations of defects, may also be presented via the user interface.

[0133]

[0142] The epitaxial defect map 500B includes information indicating the location of defects on the target substrate. The etching pit map 520 includes markings of etching pits detected by the epitaxial defect detection system. The crosshatch map 522 includes indications of crosshatching detected by the epitaxial defect detection system. One or more characteristics of the determined epitaxial defect may be presented by the epitaxial defect map 500B. For example, the severity or size of the defect may be indicated by the color, pattern, shape, size, orientation, etc., of the elements indicating the defect.

[0134]

[0143] The etching pit map 520 and the crosshatch map 522 may include a visual representation of the target substrate. The etching pit map 520 and the crosshatch map 522 may further include a visual representation of each type of epitaxial defect. In some embodiments, the location of epitaxial defects represented in the epitaxial defect map 500B may be determined, for example, by an epitaxial defect detection and / or classification procedure. In some embodiments, one or more of the epitaxial defect representations may be predictions based, for example, on the measured location of defects in other areas of the substrate.

[0135]

[0144] The crosshatch map 522 includes a set of indicators for the determined crosshatching on the target substrate. In some embodiments, one or more characteristics of the indicators may relate to the orientation of the determined crosshatching. For example, a vertical orientation of an indicator may indicate vertically aligned crosshatching, and a horizontal orientation of an indicator may indicate horizontally aligned crosshatching. In some embodiments, one or more characteristics of the indicators may be associated with frequency-domain filtering performed to determine defects. In some embodiments, one or more characteristics of the visual indicators in the epitaxial defect map 500B may be associated with feature detection techniques. For example, as shown in the epitaxial defect map 500B, a circular defect visual indicator may be associated with defects determined using a frequency-domain filter configured to detect etching pits, or a feature detection technique (e.g., Hog transform) for detecting etching pits. As a further example, a line visual indicator may indicate epitaxial defects detected using a frequency-domain filter and / or feature detection technique related to crosshatching. The orientation of a linear visual indicator may represent horizontal or vertical frequency domain filtering, horizontal or vertical cross-hatching, or horizontal or vertical feature detection (e.g., contour counting).

[0136]

[0145] The epitaxial defect map 500B can provide users with easily accessible information such as the types of defects found on the substrate, the concentration of defects in specific areas of the substrate, differences in the distribution of different classes of defects on the substrate, a display of defect density across the entire substrate or across one or more target areas of the substrate, and the radial and / or angular distribution of defects.

[0137]

[0146] In some embodiments, the user interface can provide a numerical, color, pattern, or other representation of epitaxial defects in the substrate. In some embodiments, the user interface can provide spatial breakdown of defect locations, for example, by providing the density and / or number of defects in various zones of the substrate (e.g., central zone, edge zone, intermediate zone, etc.). The user interface can further provide the angular distribution of epitaxial defects in the substrate.

[0138]

[0147] Figure 6 is a block diagram showing computer system 600 in several embodiments. In some embodiments, computer system 600 may be connected to other computer systems (for example, via a network such as a local area network (LAN), intranet, extranet, or internet). Computer system 600 may operate as a server or client computer in a client-server environment, or as a peer computer in a peer-to-peer or distributed network environment. Computer system 600 may be provided by a personal computer (PC), tablet PC, set-top box (STB), personal digital assistant (PDA), mobile phone, web appliance, server, network router, switch or bridge, or any device capable of executing a set of instructions (sequential or otherwise) that specify the actions that the device should perform. Furthermore, the term “computer” includes any collection of computers that individually or collectively execute a set (or set) of instructions to perform any one or more of the methods described herein.

[0139]

[0148] In a further embodiment, the computer system 600 may include processing devices 602, volatile memory 604 (e.g., random access memory (RAM)), non-volatile memory 606 (e.g., read-only memory (ROM) or electrically erasable programmable ROM (EEPROM)), and data storage devices 618, all of which can communicate with each other via a bus 608.

[0140]

[0149] The processing device 602 may be provided by one or more processors, such as general-purpose processors (e.g., composite instruction set arithmetic (CISC) microprocessors, reduced instruction set arithmetic (RISC) microprocessors, very long instruction word (VLIW) microprocessors, microprocessors implementing other types of instruction sets, or microprocessors implementing combinations of types) or dedicated processors (e.g., application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), or network processors).

[0141]

[0150] The computer system 600 may further include a network interface device 622 (for example, connected to network 674). The computer system 600 may also include a video display unit 610 (e.g., an LCD), an alphanumeric input device 612 (e.g., a keyboard), a cursor control device 614 (e.g., a mouse), and a signal generation device 620.

[0142]

[0151] In some embodiments, the data storage device 618 may include a non-temporary computer-readable storage medium 624 (e.g., a non-temporary machine-readable medium) capable of storing instructions 626 that encode any one or more of the methods or functions described herein, including instructions for encoding the components of Figure 1 (e.g., the prediction component 114, the corrective action component 122, the model 190, etc.) and for implementing the methods described herein.

[0143]

[0152] Instruction 626 may reside entirely or partially in the volatile memory 604 and / or the processing device 602 while being executed by the computer system 600, so that the volatile memory 604 and the processing device 602 may also constitute a machine-readable storage medium.

[0144]

[0153] In illustrative examples, computer-readable storage medium 624 is shown as a single medium, but the term “computer-readable storage medium” includes single or multiple mediums that store one or more sets of executable instructions (e.g., centralized or distributed databases, and / or associated caches and servers). The term “computer-readable storage medium” also includes any tangible medium capable of storing or encoding a set of instructions that a computer executes, causing a computer to perform any one or more of the methods described herein. The term “computer-readable storage medium” includes, but is not limited to, solid-state memory, optical media, and magnetic media.

[0145]

[0154] The methods, components, and features described herein may be implemented by separate hardware components or incorporated into the functionality of other hardware components such as ASICS, FPGAs, DSPs, or similar devices. In addition, the methods, components, and features may be implemented by firmware modules or functional networks within hardware devices. Furthermore, the methods, components, and features may be implemented in any combination of hardware devices and computer program components, or in computer programs.

[0146]

[0155] Unless otherwise specified, terms such as “receive,” “execute,” “provide,” “acquire,” “cause,” “access,” “determine,” “add,” “use,” “train,” “reduce,” “generate,” and “correct” refer to computer system execution or implementation processes and operations that manipulate and transform data represented as physical (electronic) quantities in the registers and memory of a computer system into other data similarly represented as physical quantities in the memory or registers of a computer system, or in other information storage, transmission, or display devices. Furthermore, terms such as “first,” “second,” “third,” and “fourth” used in this document are symbols used to distinguish various elements from one another and do not necessarily have an orderly meaning according to their numerical designations.

[0147]

[0156] The embodiments described herein also relate to apparatus for carrying out the methods described herein. This apparatus may include a general-purpose computer system that is specifically constructed for carrying out the methods described herein or is selectively programmed by a computer program stored in the computer system. Such a computer program may be stored in a computer-readable tangible storage medium.

[0148]

[0157] The methods and exemplary embodiments described herein are not inherently related to any particular computer or other device. A wide variety of general-purpose systems may be used in accordance with the teachings described herein, or it may be convenient to construct more specialized devices to perform the methods and / or each of their individual functions, routines, subroutines, or processes described herein. Examples of the construction of such a wide variety of systems are explicitly shown above.

[0149]

[0158] The above description is illustrative and not limiting. While this disclosure has been described with reference to certain exemplary examples and embodiments, it should be recognized that this disclosure is not limited to the examples and embodiments described. The scope of this disclosure should be determined in reference to the entire scope of equivalents to which these claims are granted, together with the subsequent claims.

Claims

1. The processing device acquires first image data of the substrate including the epitaxial film, The first image data is subjected to a frequency domain filter to obtain filtered image data, The number of epitaxial defects represented in the first image data is determined by performing feature detection on the filtered image data. Corrective measures shall be taken taking into consideration the number of epitaxial defects, Methods that include...

2. The method according to claim 1, wherein performing the feature detection includes performing one or more of the following: Hough transform or contour counting.

3. The method according to claim 1, wherein the epitaxial defect includes one or more etching pits or cross-hatching.

4. The epitaxial film is Silicon, carbon, Boron, Arsenic, Antimony, Suzu, Lin, Silicon germanium, Silicon carbide, Gallium nitride, Aluminum nitride, Gallium arsenide, Gallium aluminum nitride, Indium, or silicon arsenide The method according to claim 1, comprising one or more of the above.

5. The method according to claim 4, wherein the epitaxial film comprises a silicon / silicon germanium superlattice.

6. The first image data mentioned above is generated by Depositing the epitaxial film onto the substrate, The process involves etching the epitaxial film, The process involves capturing an image of the substrate using optical microscopy, scanning electron microscopy, or transmission electron microscopy. The method according to claim 1, including the method described in claim 1.

7. The method according to claim 1, wherein the corrective action includes screening the substrate for additional processing.

8. The aforementioned corrective measures are, To schedule maintenance of the processing chamber associated with the epitaxial film, Update the processing recipe. Updating one or more manufacturing parameters of the processing chamber, or Provide warnings to users The method according to claim 1, comprising one or more of the above.

9. The frequency domain filter, Two-dimensional high-pass filter, or Low-pass filter The method according to claim 1, comprising one of the following.

10. The method according to claim 1, further comprising one or more of classifying the epitaxial defects or determining the density of the epitaxial defects.

11. A non-temporary machine-readable storage medium for storing instructions, wherein when the instruction is executed, To acquire first image data of a substrate containing an epitaxial film, Applying a frequency domain filter to the first image data to generate filtered image data, The number of epitaxial defects represented in the first image data is determined by performing feature detection on the filtered image data, Corrective measures shall be taken taking into consideration the number of epitaxial defects, A non-temporary, machine-readable storage medium that causes a processing device to perform operations including those mentioned above.

12. The non-temporary machine-readable storage medium according to claim 11, wherein the epitaxial defect includes one or more etching pits or cross-hatching.

13. The epitaxial film is Silicon, carbon, Boron, Arsenic, Antimony, Suzu, Lin, Silicon germanium, Silicon carbide, Gallium nitride, Aluminum nitride, Gallium arsenide, Gallium aluminum nitride, Indium, or silicon arsenide A non-temporary machine-readable storage medium according to claim 11, comprising one or more of the above.

14. The first image data mentioned above is generated by Depositing the epitaxial film onto the substrate, The process involves etching the epitaxial film, The process involves capturing an image of the substrate using optical microscopy, scanning electron microscopy, or transmission electron microscopy, A non-temporary machine-readable storage medium according to claim 11, including the following:

15. The aforementioned corrective measures are, The additional process steps involve screening the substrate, Scheduling maintenance for the processing chamber, Updating the processing recipe, Updating one or more manufacturing parameters of the processing chamber, To provide alerts to users, A non-temporary machine-readable storage medium according to claim 11, comprising one or more of the above.

16. The frequency domain filter, Two-dimensional high-pass filter or Low-pass filter A non-temporary machine-readable storage medium according to claim 11, comprising one of the above.

17. A system including memory and a processing device coupled to the memory, wherein the processing device is First image data of a substrate containing an epitaxial film is acquired. A frequency domain filter is applied to the first image data to obtain filtered image data. By performing feature detection on the filtered image data, the number of epitaxial defects represented in the first image data is determined. Corrective measures will be implemented taking into account the number of epitaxial defects. A system that is configured in such a way.

18. The aforementioned corrective measures are, Screening the substrate for the execution of additional processing steps, To schedule maintenance of the processing chamber, Update the processing recipe. Updating one or more manufacturing parameters of the processing chamber, or Provide warnings to users The system according to claim 17, comprising one or more of the above.

19. The system according to claim 17, wherein the processing device further trains a machine learning model for determining the number of epitaxial defects in a target substrate by providing the first image data as a training input and the number of epitaxial defects as a target output.

20. The epitaxial film is Silicon, carbon, Boron, Arsenic, Antimony, Suzu, Lin, Silicon germanium, Silicon carbide, Gallium nitride, Aluminum nitride, Gallium arsenide, Gallium aluminum nitride, Indium, or silicon arsenide The system according to claim 17, comprising one or more of the above.

Images