Performance Management of Semiconductor Substrate Tools

Abstract

Preventive management of semiconductor substrate tools. Machine learning models are used to predict future performance characteristics of such tools. In some examples, the model can diagnose problems related to ambient conditions of the tool or the tool's environment. In some examples, the model can recommend one or more corrective actions to maintain proper performance of the substrate tool.

Description

Technical Field

[0001] This application was filed as a PCT international patent application on May 24, 2023, and claims the benefit and priority of U.S. Provisional Patent Application No. 63 / 346,358, filed on May 27, 2022, the entire disclosure of which is incorporated herein by reference.

[0002] This disclosure is directed to the manufacture of semiconductor components and the management of the performance of tools involved in semiconductor component manufacturing processes.

Background Art

[0003] As part of the formation of semiconductor chips or other types of integrated circuits (ICs), semiconductor substrates are manufactured or fabricated. Final IC components can be incorporated into the substrate by a series of fabrication steps. The fabrication steps may include a deposition step in which a thin film layer is added onto the substrate. The substrate can then be coated with a photoresist and the circuit pattern of a reticle can be projected onto the substrate using lithography techniques. An etching process can then be performed using an etching tool.

[0004] To enable the use of the completed substrate (e.g., a wafer or a wafer lot), each tool involved in the substrate fabrication process must operate within a predefined acceptable operating tolerance with respect to the aspects of the substrate to which the tool is involved. If even a single tool in the fabrication process is operating outside of its tolerance, defects of a size sufficient to require the scrapping of all wafers in that run or subsequent fabrication runs may occur on the substrate. The operating performance of a given fabrication tool can degrade or drift over time and through use (e.g., in each successive fabrication run), such that the tool will ultimately require recalibration to operate within specifications.

Summary of the Invention

[0005] Generally speaking, the present disclosure is directed to predictive maintenance of semiconductor substrate tools (also simply referred to as tools herein). Substrate tools can include, for example, but are not limited to, fabrication tools (such as deposition tools, lithography tools, oxidation tools, epitaxial reactors, diffusion tools, ion implantation tools, etching tools, and chemical mechanical planarization (CMP) tools, etc.), inspection tools, dicing tools, grinding tools, polishing tools, metrology and other measurement tools. Although the present disclosure may refer to a particular substrate fabrication tool (such as a deposition tool) in certain examples, the features of the present disclosure can be readily applied to other substrate tools that do not perform fabrication functions.

[0006] More generally speaking, the present disclosure is directed to improving the performance consistency across substrate tools of the same type.

[0007] The features of the present disclosure can be implemented as a system, as a method executed by a computer, and as instructions stored in a non-transitory computer-readable storage device.

[0008] According to certain aspects, the present disclosure is directed to predicting the future state of a substrate tool using a machine learning model.

[0009] According to certain aspects, the present disclosure is directed to predicting performance degradation of a substrate tool using a machine learning model.

[0010] According to further aspects, the present disclosure is directed to diagnosing the cause of performance degradation or predicted performance degradation of a substrate tool using a machine learning model.

[0011] According to further aspects, the present disclosure is directed to diagnosing the cause of performance discrepancy or predicted performance discrepancy between substrate tools of the same type using a machine learning model.

[0012] According to a further aspect, the present disclosure is directed to using a machine learning model to determine adjustments to a substrate tool and correct a predicted performance degradation of the substrate tool.

[0013] According to a further aspect, the present disclosure is directed to using a machine learning model to determine adjustments to a substrate tool and reduce a performance discrepancy between the substrate tool and another substrate tool of the same type.

[0014] According to certain aspects, a method executed by a computer to predict a future performance state of a semiconductor substrate tool following a future use thereof (e.g., a future fabrication operation, a future inspection operation, a future metrology operation) includes providing a current performance state of the fabrication tool to a trained machine learning model, providing operation data about the tool to the trained machine learning model, and receiving a predicted future performance state and a recommended recalibration of the tool, the predicted future performance state and the recommended recalibration being determined by the trained machine learning model based on the current performance state and the operation data.

[0015] According to a further specific aspect, a method executed by a computer to predict a future performance state of a semiconductor substrate fabrication tool following a future use thereof (e.g., a future fabrication operation, a future inspection operation, a future metrology operation) includes receiving, by the trained machine learning model, a current performance state of the fabrication tool, receiving, by the trained machine learning model, operation data about the tool, and generating, by the machine learning model, a predicted future performance state and a recommended recalibration of the tool, the predicted future performance state and the recommended recalibration being determined based on the current performance state and the operation data.

[0016] According to a further specific aspect, a method executed by a computer includes training a machine learning model to generate a predicted future performance state of a semiconductor substrate manufacturing tool, and generating a recommended recalibration of the semiconductor substrate manufacturing tool based on the current performance state of the tool and the operation data associated with the tool, where the operation data includes data generated by sensors associated with the tool or the surrounding environment of the tool, and data generated by automatic tests executed by the tool.

[0017] According to a further specific aspect, a method executed by a computer for predicting a predicted future performance state of a substrate tool following a future use of the substrate tool (e.g., a future manufacturing operation, a future inspection operation, a future measurement operation) includes providing the current performance state of the substrate tool to a trained machine learning model, providing the operation data of the substrate tool to the trained machine learning model, and receiving a predicted future performance state and a recommended recalibration of the substrate tool, where the predicted future performance state and the recommended recalibration are determined by the trained machine learning model based on the current performance state and the operation data.

[0018] According to a further specific aspect, a method for predicting a future performance state of a substrate tool following a future use of the substrate tool (e.g., a future manufacturing operation, a future inspection operation, a future measurement operation) includes receiving, by a trained machine learning model, the current performance state of the substrate tool, receiving, by the trained machine learning model, the operation data about the substrate tool, and generating, by the trained machine learning model, a predicted future performance state and a recommended recalibration of the substrate tool, where the predicted future performance state and the recommended recalibration are determined based on the current performance state and the operation data.

[0019] According to a further specific aspect, a method executed by a computer includes training a machine learning model to generate a predicted future performance state of a substrate tool, and generating a recommended recalibration of the substrate tool based on the current performance state of the substrate tool and the operation data associated with the substrate tool, where the operation data includes data generated by sensors associated with the substrate tool or sensors associated with the surrounding environment of the substrate tool, and data generated by an automatic test executed by the substrate tool.

[0020] This summary is provided to introduce, in a simplified form, a selection of concepts that are further described in the detailed description below. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Additional aspects, features, and / or advantages of examples will be described in part in the description below, become apparent in part from the description, or can be learned by practice of the disclosure.

[0021] With reference to the following figures, non-limiting and non-exhaustive examples will be described.

Brief Description of the Drawings

[0022]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Modes for Carrying Out the Invention

[0023] Examples of the present disclosure describe systems, methods, and computer-readable products for improving substrate fabrication. More specifically, examples of the present disclosure describe systems, methods, and computer-readable products for improving the performance of substrate tools and maintaining the improved performance.

[0024] An example of such a substrate is a semiconductor wafer composed of dies. A given wafer has a yield, which can refer to the percentage of dies on the substrate that meet one or more defined operating criteria, quality criteria, or other acceptability criteria. Defects in the wafer caused by fabrication tools (e.g., deposition tools, lithography tools, etching tools, CMP tools, and other tools described above) operating outside of specifications (also referred to herein as out-of-spec) can reduce the yield of the wafer. Depending on the yield and / or type, number and / or severity of defects in the wafer, it may be necessary to discard the wafer or an entire lot of wafers from a given fabrication run, which is costly.

[0025] Substrate inspection and measurement tools (e.g., metrology tools) that are out of spec can cause similar problems, for example, by failing to identify defects in non-existent wafers and / or by failing to identify defects in existing wafers.

[0026] The present disclosure relates to a systematic approach for substrate tool maintenance. The systematic approach can be easily applied to many different types of substrate tools, such as fabrication tools, inspection tools, and metrology tools. Advantageously, regardless of the type of substrate tool, the systematic approach can reduce the time that the substrate tool is down for production and / or can reduce the number of wafers that must be discarded.

[0027] In addition, the systematic approach of the present disclosure can improve the performance consistency across different tools of the same type. For example, in relation to fabrication tools, a given fabrication facility may have multiple tools that perform the same fabrication process. Even if all of these tools are within specifications (also referred to as within spec in this specification), there may be performance inconsistencies from tool to tool, which can disadvantageously result in inconsistencies across the finished substrates produced by the facility. Further, such inconsistencies can be exacerbated by performance inconsistencies of other tools later included in the fabrication process.

[0028] During substrate fabrication, the substrate goes through many process steps performed by various tools. Such tools can include, but are not limited to, deposition tools, etching tools, lithography tools, oxidation tools, epitaxial reactors, diffusion tools, ion implantation tools, and coating tools. Calibration or tuning of each tool is important for how the tool functions. For example, a deposition tool within spec can deposit a 300 angstrom layer on a wafer within a tolerance of 10 angstroms. For each operation of the tool (an operation can correspond to the fabrication of a wafer lot or any other iteration of use for a given type of tool), the calibration parameters from the tool can vary due to any of several different factors. Inspection and measurement tools are used to inspect and measure aspects of the wafer.

[0029] In addition, there may be a discrepancy even though tools of the same type are both operating within spec. For example, a particular deposition tool may deposit a 300 angstrom layer within a spec of 305 angstroms, while another deposition tool may deposit the same layer at 291 angstroms, resulting in a discrepancy in the final product produced by the facility or fabricator using both tools.

[0030] According to current methods of tool management and maintenance, each tool is periodically taken out of production for testing and maintenance. In some cases, wafers produced by the tool are periodically inspected to confirm whether the tool is operating within specifications, and if it is not operating within specifications, the tool is taken out of production for maintenance. These methods are time-consuming and costly, and while some tools are taken out of production prematurely, other tools are only taken out of production after they have already operated out of spec, requiring the wafers to be discarded, so they can be either overly inclusive or insufficiently inclusive when identifying tools that require maintenance. Further, if a tool begins operating out of spec, the cause and thus the remedy must be manually identified, which can prolong the time the tool is taken out of production.

[0031] According to the present disclosure, a trained machine learning model is used to predict whether a given tool will perform future (or upcoming) use within or out of spec prior to use of the tool (e.g., prior to a production run). Similarly, the model can predict when maintenance will be required for a given tool. The model can also determine the cause of performance discrepancies across tools. In this way, tool maintenance and management can be performed at a timing that minimizes the cost of performing maintenance either too frequently or not frequently enough.

[0032] In addition, the machine learning model is configured to determine deviating or beginning-to-deviate tool parameter(s), thereby identifying and recommending tool parameter(s) that require recalibration or other adjustment to keep the tool within spec.

[0033] In addition, the machine learning model is configured to perform predictive maintenance modeling for different tool types by considering input data specific to each tool type. Specifically, the model is trained to know how the input data correlates with future tool performance and how the input data correlates with the need for specific tuning / recalibration for a particular tool.

[0034] In addition, the machine learning model is configured to perform predictive maintenance modeling for different substrate processes (e.g., different deposition processes, different inspection processes) that can be performed by the same tool by considering input data specific to each tool type and specific to the functions of each process (e.g., fabrication, inspection, measurement). Specifically, the model is trained to know how the input data correlates with future tool performance and how the input data correlates with the need for specific tuning / recalibration for a particular fabrication process and a particular tool performing other substrate functions.

[0035] In addition to improving efficiency in tool maintenance, the present invention can provide tool-to-tool alignment. That is, for a given type of tool (e.g., a tool that performs a particular fabrication process or other substrate function), the use of the present technology can ensure that all of the tools are operating and continue to operate within a given established tolerance of each other, thereby improving fabrication consistency across tools of the same type. Accordingly, the machine learning model can take into account current state performance data associated with one or more other tools to determine an appropriate maintenance plan for a given tool of the same type.

[0036] According to some examples, the machine learning model is a recurrent neural network (RNN) model. The RNN model may be particularly suitable for the purposes of the present invention in that the RNN model is configured to obtain its previous output together with further input as part of its input and generate a next output. As will be described in more detail herein, according to the present disclosure, the machine learning model is configured to provide a predicted future performance state of a tool based on various inputs for a given tool. These inputs include previously predicted future states of the tool generated by the model, as well as additional data that may not have been available for previous model executions of the tool. Thus, by using an RNN for tool management according to the present disclosure, for a given tool, each prediction can be based in part on one or more previous predictions of that tool taken prior to the previous operation of the tool.

[0037] In some examples, the RNN model may be used in combination with one or more convolutional layers.

[0038] FIG. 1 schematically shows an exemplary system 100 for managing the performance of a semiconductor substrate tool in accordance with the present disclosure.

[0039] System 100 includes a computing device 202. The computing device 202 may be a server and / or other computing device that executes operations described herein, such as the tool management operations described herein. The computing device 202 may include computing components 206. The computing components 206 include at least one processor 208 and a memory 204. The memory 204 can include a non-transitory computer-readable medium. Depending on the exact configuration, the memory 204 (especially, storing tool management instructions and other instructions for performing other operations disclosed herein) may be volatile (such as RAM), non-volatile (such as ROM, flash memory, etc.), or some combination of the two. The computing device 202 may include one or more graphics processing units (GPUs) configured to facilitate model training and / or model prediction. Further, the computing device 202 can also include a storage device (removable 210 and / or non-removable 212) including, but not limited to, a solid-state device, a magnetic disk, or an optical disk, or a tape. Further, the computing device 202 can also have input devices (singular or plural) 216 such as a touch screen, a keyboard, a mouse, a pen, a voice input, etc., and / or output devices (singular or plural) 214 such as a display, a speaker, a printer, etc. One or more communication connections 218 such as a local-area network (LAN), a wide-area network (WAN), a point-to-point, Bluetooth, RF, etc. can also be incorporated into the computing device 202.

[0040] System 100 can include one or more monitoring and / or inspection devices 102 that are operably communicating with computing device 202, such as being linked via a network and communication connection(s) 218. Non-limiting examples of such monitoring and / or inspection devices 102 are shown in FIG. 2. Referring to FIG. 2, the monitoring and / or inspection device 102 can include, for example, a substrate inspection device 110 and a tool sensor 112.

[0041] Non-limiting examples of the substrate inspection device 110 can measure the intensity of light or other waves (e.g., sound waves) reflected from the substrate at different wavelengths, generate spectral data, and therefrom determine substrate characteristics such as the thickness of the layer. The inspection data can be supplied as an input that can be partially used by a model to a machine learning model executed on computing device 202 to determine the future state of the tool involved in the measured characteristic(s).

[0042] Non-limiting examples of the tool sensor(s) 112 can include a temperature sensor, a vacuum sensor, a light intensity sensor, a vibration sensor, an ambient light sensor, a humidity sensor, an optical temperature sensor, a fan speed sensor, a pressure sensor, a flow sensor, a current sensor, a voltage sensor, etc. Data from such sensors can be related to specific tool parameters such as the heat or light intensity of the tool's lamp, or the quality of the vacuum chamber generated by the tool. In some examples, data from such sensors can be related to environmental conditions such as ambient temperature, ambient humidity, ambient light, ambient noise, etc. The tool sensor data can be supplied as an input that can be partially used by a model to a machine learning model executed on computing device 202 to determine the future state of the tool affected by the sensed condition(s).

[0043] System 100 includes one or more substrate tools 104. The substrate tools 104 can include, for example, fabrication tools that perform different substrate fabrication processes such as deposition, lithography, oxidation, diffusion, ion implantation, CMP, and etching. The substrate tools 104 can additionally or alternatively include inspection and / or measurement (or gauging) tools. The substrate tool(s) 104 can be linked to a computing device 202 via a network and communication connection(s) 218. In this way, the substrate tool(s) 104 can supply, for example, tool calibration data as an input that can be partially used by a machine learning model executed on the computing device 202 to determine a future state of the tool.

[0044] The substrate tool 104 itself can also provide data related to the predicted future performance. Alternatively or additionally, such data regarding the substrate tool 104 can be obtained by one or more tools or other tools. Such data can include tool auto-test data, calibration data, and runtime data. For example, the substrate tool can monitor or periodically perform monitoring tests regarding various calibrations and other aspects of the substrate tool, such as alignment of the tool's fabrication stage, sway of the tool's fabrication stage, intensity of the tool's lamp, reproducibility of the tool's aperture (e.g., material deposition aperture or lens aperture), video focus of the tool. In some examples, the substrate tool 104 itself can supply this auto-monitoring and / or auto-test data as an input that can be partially used by a machine learning model executed on the computing device 202 to determine a future state of the tool that is affected by the sensed condition(s).

[0045] Runtime data is captured for each operation of the tool and can thus be useful in identifying and tracking small changes in tool performance and when they occur precisely. Examples of runtime data include alignment data and autofocus data of the tool for each use of the tool (e.g., each fabrication operation). In contrast, automated test data and calibration data can be captured periodically, for example, as part of a tool health check process. In some examples, the automated test data and calibration data are not monitored or do not exist in each tool operation and can thus include performance data of a type that would not be included in runtime data. An example of calibration data includes data regarding calibrations performed by the tool.

[0046] In an alternative configuration, one or more components of the computing device 202 are present locally on one or more monitoring and / or inspection devices 102 or substrate tools 104. For example, one or more monitoring and / or inspection devices 102 or one or more substrate tools 104 can be configured to perform one or more of the machine learning model tool management operations described herein themselves. That is, the machine learning model tool management instructions can be executed directly on one or more monitoring and / or inspection devices 102 or one or more substrate tools 104.

[0047] FIG. 3 shows an example of using a machine learning model to manage the performance of a semiconductor substrate tool according to an example of the present disclosure using the system 100 of FIG. 1.

[0048] In some examples, one or more of the operations of FIG. 3 can be performed by one or more of the computing device 202 of FIG. 1 and / or the monitoring and / or inspection device 102 and / or the substrate tool 104.

[0049] Referring to FIG. 3 as a whole, an input 302 is provided to a machine learning model 308. In some examples, the machine learning model 308 is an RNN model. The machine learning model 308 is trained to be configured to analyze the input 302 to generate an output 310 that can be provided to a technician using the computing device 202 of FIG. 1 by the machine learning model 308 (e.g., can be displayed on a display device). The input 302 can be provided by any of several different sources. Examples of such sources can include the substrate tool 104 or the monitoring and / or inspection device 102 (FIG. 1). The input can also be provided from data stored on the computing device 202.

[0050] The input 302 can include the current state data 304 of the tool and additional data 306. The current state data 304 of the tool can include the previously predicted state for a given tool predicted by the machine learning model 308. For example, for a given deposition tool, modeling can be performed to predict the future state of the tool after each successive use of the tool (e.g., substrate fabrication run). For each modeling, the output is the predicted state of the tool following the next use of the tool (e.g., the next fabrication run). For each modeling, one of the inputs 302 to the machine learning model 308 is the previously predicted state of the tool predicted by the machine learning model 308 prior to the previous use of the tool (e.g., the previous fabrication run). This input reflects the current state of the tool. That is, this input reflects the current performance level or quality of the tool (e.g., within spec, out of spec, within spec but having a deviation from another tool of the same type, etc.).

[0051] In certain examples of layer deposition tools, this input can reflect, for a particular substrate layer, the thickness of the deposited layer deposited by the tool in a previous use of the tool (e.g., a previous fabrication run). In certain examples of inspection tools, this input can reflect, for a particular substrate, whether a defect was detected and / or the nature and severity of the defect. In certain examples of metrology tools, this input can reflect, for a particular feature such as a transistor or thin film, whether the critical dimensions are within specification. In certain examples of lithography tools, this input can reflect, for a given substrate, the alignment of two layers of the substrate with respect to each other. For many different types of substrate tools, this input can reflect, among other things, the critical dimensions of one or more structural or functional features of the substrate. Regardless of the substrate tool in question or the performance aspect of the substrate tool, this input data is supplied to the machine learning model 308, and as a result, the machine learning model 308 can use it as a baseline set of data for predicting the current state of the tool to the next future state of the tool (e.g., after the next fabrication run, or after the next metrology run, etc., the state of the tool after the next use of the tool).

[0052] The current state data 304 of the tool can also include performance data of the current state of other tools, based on which the machine learning model 308 can perform inter-tool performance matching for tools of the same type by comparing the predicted performance of a given tool with the current or predicted performance of another tool of the same type.

[0053] The input 302 can also include additional data 306. The additional data can include operation data of the tool.

[0054] Referring to FIG. 4, the additional data 306 can include current data obtained by the monitoring and / or inspection device 102 and / or the substrate tool 104 itself (FIG. 1). In some examples, the current data, or some of the current data, is obtained from the most recent fabrication operation of the tool. For example, the additional data 306 can include sensor data 322 that includes both tool-specific sensor data and ambient condition sensor data. The additional data 306 can also include auto test data 320 generated by the tool itself. The additional data 306 can also include runtime data 326 obtained from each use (e.g., each fabrication run) of the substrate tool 104. The additional data 306 can also include calibration data 328 (e.g., data regarding calibration performed by the substrate tool 104).

[0055] Referring to FIG. 4, the additional data 306 can also include tool event data 324. The tool event data 324 can be provided to the machine learning model 308 by the tool itself, or via another means, or from another device. The event data 324 can include, for example, exceptions that occurred during tool use (e.g., during a fabrication run) regarding a problematic tool, such as the occurrence of an error during operation, or the removal of the tool by the tool owner during or after operation. For example, if the tool is removed from fabrication and recalibrated, the relevance of the current state data 304 of the tool for predicting the future state of the tool can be ignored by the machine learning model 308 depending on how the model was trained. Another example of an event can be the replacement of a part or component of the tool, such as the replacement of a tool lamp, a tool lens, a tool stage, etc.

[0056] The machine learning model 308 can be a trained machine learning model that processes the input 302 and generates an output 310 based on the input 302. The output 310 can be presented via an interface of a computing device such as the output device 214 (FIG. 1). For example, the output 310 can be presented as a tool maintenance or management report indicating the predicted maintenance of the tool. The output 310 can include one or more alerts or alarms. For example, it can generate a warning that the tool is predicted to be out of spec during the next use of the tool so that performance recalibration is ensured, or that the tool deviates in performance from another tool of the same type by more than a predetermined threshold magnitude so that inter-tool recalibration is ensured.

[0057] The output 310 can include the predicted future state 312 of the tool. The predicted future state 312 of the tool can include the predicted performance metrics of a given tool following the next fabrication or other function of the tool. In a specific example of a layer deposition tool, this output can reflect a prediction of the thickness of the layer to be deposited by the tool during the next use of the tool (e.g., during the next fabrication operation of the tool) for a particular substrate layer in which the tool is involved in fabrication. Additionally, or alternatively, this output can include an indication that the predicted performance of the tool for the next operation is within or out of spec. Additionally, or alternatively, this output can include an indication that the predicted performance of the tool for the next operation is less than or greater than a predetermined maximum threshold deviation from the performance of another tool of the same type.

[0058] The output 310 can include the predicted future performance state of the substrate tool. The predicted future performance state output can then be provided as an input to the trained machine learning model 308 as the subsequent performance state of the substrate tool. The predicted future performance state of the substrate tool following a subsequent future use (e.g., a future fabrication operation) of the substrate tool is determined by the trained machine learning model 308 based on the subsequent performance state of the substrate tool.

[0059] Output 310 can include tool diagnostics 313. The tool diagnostics 313 can indicate the cause that the predicted performance of the tool is out of specification or the cause that the predicted performance of the tool is greater than the minimum threshold deviation. Non-limiting examples of the diagnostics can include misalignment stages of the tool, tool optical system temperatures that are too high, stages with excessive vibration, lamp intensities that are too high or too low, fan speeds that are too low, ambient temperatures that are too high or too low, variations in aperture reproducibility that are too high / low, variations in video focus that are too high, etc. Thus, the diagnostics 313 can be related to the parameters of the tool itself or the ambient conditions in which the tool is placed.

[0060] Output 310 can include one or more repair recommendations 314. For example, if it is determined based on the predicted future state 312 of the tool and the tool diagnostics 313 that some maintenance of the tool or the ambient conditions surrounding the tool is warranted, the machine learning model 308 is configured to generate maintenance recommendations. For example, the recommendations 314 can include recommendations such as reducing or increasing the ambient temperature or ambient light, replacing the fan, adjusting the alignment of the tool's stage by a specified amount, adjusting the intensity of the tool's lamp by a specified amount, replacing the tool's optical system, tightening the tool's stage by a specified amount, etc.

[0061] Advantageously, the machine learning model 308 identifies the diagnostics and the remedies specific to the diagnostics, so that the indicated maintenance is targeted, and thus can be performed more quickly and with less production interruption than regular maintenance where more aspects of the tool are checked and tested to see if tuning or other recalibration is needed for each.

[0062] In addition, the machine learning model 308 is configured to identify when performance deviations or insufficient performance are due to the surrounding environment rather than the parameters of the tool or the result of calibration. For example, the remedial measure determined by the machine learning model 308 can be to reduce the ambient temperature of the environment around the tool by a specified number of degrees, which can be an easy and simple modification that does not require removing the tool from the product for any length of time.

[0063] FIG. 5 shows a method 400 for managing the performance of a semiconductor substrate tool in accordance with an example of the present disclosure using the system 100 of FIG. 1. It will be understood that different embodiments of the present disclosure can include different combinations of subsets of the steps of method 400, non-limiting examples of which are described herein.

[0064] In step 402 of method 400, a machine learning model (e.g., machine learning model 308 (FIG. 3)) is trained for each type of substrate tool using known tool performance data. For a given tool or type of tool, performance data from continuous use (e.g., continuous fabrication operations), as well as automated test data, sensor data, and event data associated with each fabrication are processed. From this information, the model 308 learns how different factors (tool-specific factors, environmental factors, etc.) individually and collectively affect the performance of a given tool or tool type over time and cause changes in performance over continuous use of the tool (e.g., over continuous fabrication operations). For example, the machine learning model 308 can learn that for a particular type of layer deposition tool, a lamp intensity of a particular magnitude can change the performance of the tool in a particular way (e.g., the deposited layer of the tool becomes thicker) depending on other tools and surrounding factors.

[0065] In operation 402, by processing training data along with tool performance data known from many operations of the tool, the machine learning model 308 learns how to predict the future state of the tool, diagnose problems related to the tool, and recommend improvement measures to solve the problems. For example, the model learns to match the relevant data of the target tool with the corresponding data of the training tools having known results.

[0066] In operation 404 of method 400, the machine learning model 308 obtains the current performance state of the tool. In some examples, the current performance state of the tool is the predicted future state 312 (FIG. 3) of the tool from previous operations of the model of that tool. In other examples, for example, when the tool is new or has been recalibrated since the last model operation, the current performance state of the tool can be determined using a substrate inspection device such as a metrology device. For example, with respect to a layer deposition tool, the current performance state of the tool can be determined by a spectrometer that measures the thickness of the substrate deposition layer from the most recent (or first) operation of the tool. However, once the current performance state of the tool is determined, data is supplied to model 308 in operation 408.

[0067] In operation 406 of method 400, additional data is obtained. The additional data can include the additional data 306 (FIG. 4) described above. This data is obtained from the substrate tool 104 and the monitoring and inspection device(s) 102 (FIG. 1) and can be supplied to the machine learning model 308 in operation 408 as described above.

[0068] In operation 410 of method 400, the machine learning model 308 generates a model output. The model output can include any of the outputs 310 (FIG. 3) described above. The model output can be provided via any suitably configured output device as described above.

[0069] The machine learning model 308 (Figure 3) can provide known performance data even after the initial training phase. For example, the output prediction of the model after the operation of method 400 can be empirically tested (e.g., using a substrate inspection device), and the results of the test can be input into the model as additional training data to improve the model's ability to predict the need for recalibration for a given tool.

[0070] As described above, the machine learning model 308 can be trained for each tool type. Calibration parameters, environmental factors, and event occurrences can affect the performance of different types of tools in different ways and to different extents. Thus, in step 402 (Figure 5), the machine learning model 308 (Figure 3) can be trained separately for each tool. Additionally, the current state data 304 (Figure 3) of the tool can include the type of the tool. The type of the tool is used by the machine learning model 308 to select an appropriate modeling path to generate the output 310 of that tool.

[0071] Furthermore, the machine learning model 308 can be trained separately for each manufacturing process or other tool function. Calibration parameters, environmental factors, and event occurrences can affect the performance of different types of manufacturing processes and other tool functions in different ways and to different extents. Thus, in step 402 (Figure 5), the machine learning model 308 (Figure 3) can be trained separately for each type of manufacturing process or other tool function. Additionally, the current state data 304 (Figure 3) of the tool can include the manufacturing process or tool function type. The manufacturing process or tool function type is used by the machine learning model 308 to select an appropriate modeling path to generate the output 310 for that tool.

[0072] The embodiments described in this specification may be adopted using software, hardware, or a combination of software and hardware to implement and execute the systems and methods disclosed herein. Throughout this disclosure, specific devices are enumerated as performing specific functions, but those skilled in the art will understand that these devices are provided for illustrative purposes and that other devices may be used to perform the functions disclosed herein without departing from the scope of this disclosure. Additionally, some aspects of this disclosure have been described above with reference to block diagrams and / or operational diagrams of systems and methods according to aspects of this disclosure. The functions, operations, and / or acts described in the blocks may be performed in an order different from the order shown in any respective flowchart. For example, two blocks shown consecutively may actually be performed or implemented substantially simultaneously or in reverse order, depending on the relevant functions and implementations.

[0073] This disclosure describes some embodiments of the technology with reference to the accompanying drawings in which only some of the possible embodiments are shown. However, other aspects may be embodied in many different forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the possible embodiments to those skilled in the art. Further, as used in this specification and the claims, the phrase "at least one of element A, element B, or element C" is intended to convey any of element A, element B, element C, element A and B, element A and C, element B and C, and element A, B, and C. Additionally, those skilled in the art will understand the degree conveyed by terms such as "about" or "substantially" in light of the measurement techniques utilized herein. Unless such terms can be clearly defined or understood by those skilled in the art, the term "about" shall mean plus or minus 10 percent.

[0074] This specification describes specific embodiments, but the scope of the technology is not limited to those specific embodiments. Further, different examples and embodiments can be described separately, but such embodiments and examples can be combined with each other when implementing the technology described in this specification. Those skilled in the art will recognize other embodiments or improvements within the scope and spirit of the technology. Therefore, a specific structure, operation, or medium is disclosed only as an exemplary embodiment. The scope of the technology is defined by the following claims and any equivalents thereof.

Claims

1. A method executed by a computer for determining a predicted future performance state of a substrate tool, comprising: providing a current performance state of the substrate tool to a trained machine learning model; providing operation data of the substrate tool to the trained machine learning model; outputting a predicted future performance state of the substrate tool, wherein the predicted future performance state is determined by the trained machine learning model based on the current performance state and the operation data.

2. The method executed by a computer according to claim 1, further comprising outputting a recommended recalibration of the substrate tool determined by the trained machine learning model based on the current performance state.

3. The method executed by a computer according to claim 1 or 2, wherein the operation data includes data generated by sensors associated with the substrate tool or associated with the surrounding environment of the substrate tool.

4. The method executed by a computer according to any one of claims 1 to 3, wherein the operation data includes data generated by an automatic test executed by the substrate tool.

5. The method executed by a computer according to any one of claims 1 to 4, wherein the operation data includes data generated by the occurrence of an error associated with the substrate tool.

6. The method executed by a computer according to any one of claims 1 to 5, wherein the current performance state includes the type of the substrate tool.

7. The method executed by a computer according to any one of claims 1 to 6, wherein the current performance state includes the type of manufacturing process executed by the substrate tool or the type of another substrate function.

8. The method executed by a computer according to any one of claims 1 to 7, wherein the trained machine learning model includes a recurrent neural network.

9. Outputting the predicted future performance state as the output predicted future performance state; Providing the output predicted future performance state as a subsequent performance state of the substrate tool as an input to the trained machine learning model; Receiving a subsequent predicted future performance state of the substrate tool subsequent to a subsequent future use of the substrate tool, wherein the subsequent predicted future performance state is determined by the trained machine learning model based on the subsequent performance state of the substrate tool, the method executed by a computer according to any one of claims 1 to 8 further comprising.

10. The current performance state includes the thickness of the substrate layer, The predicted future performance state includes a predicted thickness of the substrate layer, and the thickness and the predicted thickness are different, the method executed by a computer according to any one of claims 1 to 9.

11. The current performance state is determined by a substrate inspection tool, the method executed by a computer according to any one of claims 1 to 10.

12. The recommended recalibration includes recalibrating the identified parameters of the substrate tool prior to a future use of the substrate tool, the method executed by a computer according to any one of claims 1 to 11.

13. The method executed by the computer according to claim 12, wherein the identified parameter is lamp intensity.

14. Further comprising receiving a recommendation for adjusting conditions of the ambient environment around the substrate tool, the recommendation being generated by the trained machine learning model based on the current performance state and the operation data, the method executed by the computer according to any one of claims 1 to 13.

15. The method executed by the computer according to any one of claims 1 to 14, wherein the predicted future performance state includes an indication that the performance of the substrate tool will fall outside a predefined performance specification for future use of the substrate tool.

16. The method executed by the computer according to any one of claims 1 to 15, wherein the predicted future performance state includes an indication that the performance of the substrate tool will deviate from the performance of another tool beyond a predefined maximum deviation in future use of the tool.

17. A method for predicting a future performance state of a substrate tool, comprising: means for receiving, by a trained machine learning model, a current performance state of the substrate tool; means for receiving, by the trained machine learning model, operation data of the substrate tool; means for generating, by the trained machine learning model, a predicted future performance state of the substrate tool, wherein the predicted future performance state is determined based on the current performance state and the operation data.

18. The method according to claim 17, further comprising means for generating, by the trained machine learning model, a recommended recalibration of the substrate tool based on the current performance state and the operation data.

19. The operation data is Sensors associated with the substrate tool or associated with the surrounding environment of the substrate tool, Automatic tests performed by the substrate tool, Runtime data for the substrate tool, including data associated with alignment or autofocus of the substrate tool, Calibrations performed by the substrate tool, and The method according to claim 17 or 18, comprising data generated by one or more of the events including replacement of components of the substrate tool.

20. The method according to any one of claims 17 to 19, wherein the operation data includes data generated by the occurrence of an error associated with the substrate tool.

21. A system for determining a predicted future performance state of a substrate tool, One or more processors, A non-transitory computer-readable storage device storing instructions, which when executed by the one or more processors cause the system to, Provide the current performance state of the substrate tool to a trained machine learning model, Provide the operation data of the substrate tool to the trained machine learning model, Output the predicted future performance state of the substrate tool, the predicted future performance state being determined by the trained machine learning model based on the current performance state and the operation data.

22. The system according to claim 21, wherein the operation data includes data generated by the occurrence of an error associated with the substrate tool.

Images