Process characterization and compensation using optical wall process sensors (OWPS)

Abstract

The method includes receiving, by a processing device, first sensor data indicative of a condition of a wall corresponding to a first processing chamber. The first sensor data includes optical spectrum data. The method further includes determining, by the processing device, a first value based on the first sensor data. The first value corresponds to a first amount of a product disposed along a surface of the wall at a first time. The method further includes determining, by the processing device, a first update of a first process operation associated with the first processing chamber based on the first value. The method further includes performing, by the processing device, one or more of: (i) preparing a notice indicative of the first update for presentation on a graphical user interface (GUI), or (ii) causing the first process operation to be performed in accordance with the first update.

Description

[Technical field]

[0001] SUMMARY OF THE DISCLOSURE Embodiments of the present disclosure relate to process control and process learning using optical sensors. [Background technology]

[0002] Changes in conditions on the surfaces of the processing chamber affect various processing parameters. For example, redeposition of etch by-products on the chamber walls can change the etch rate of a given process. Thus, as substrates are processed in the chamber, the etch rate (or other process parameters or conditions), among other conditions, can change, resulting in non-uniform processing from substrate to substrate.

[0003] Conventional systems lack a reliable method for monitoring surface conditions within the processing chamber, including surfaces such as the liner (or inner wall), lid, electrostatic chuck (ESC), and process ring. For example, the chemical, physical, and thermal conditions of the processing chamber liner are known to affect the plasma process by influencing the recombination of outgassing gases near the liner. Several monitoring methods have been developed, such as capacitance or resonant frequency monitoring, but these methods are subject to thermal or high frequency noise, especially during processing.

[0004] Furthermore, substrate processing can be inefficient and / or inaccurate due to lack of information, such as optical information, related to the condition of the processing chamber walls, especially as processing equipment ages. For example, process changes can cause changes in the amount of deposition (e.g., deposition thickness) or changes in the composition of the deposits, both of which can result in a large number of unacceptable processed substrates, which are often discarded. In another example, the determination of when to perform a cleaning process in a processing chamber can be based solely on the time in process, which is merely guesswork. However, performing a cleaning process too frequently can affect substrate throughput, and performing a cleaning process too infrequently can affect product quality. Other processing inefficiencies or inaccuracies exist, which will be discussed in more detail below. Summary of the Invention

[0005] The following is a simplified summary of the disclosure intended to provide a basic understanding of some aspects of the disclosure. This summary is not an extensive overview of the disclosure. It is not intended to identify key or critical elements of the disclosure, nor to delineate the scope or claims of particular embodiments of the disclosure. The sole purpose of this summary is to present some concepts of the disclosure in a simplified form as a prelude to the more detailed description presented later.

[0006] Some of the embodiments described herein encompass a method that includes receiving, by a processing device, first sensor data indicative of a condition of a wall corresponding to a first processing chamber. The first sensor data includes optical spectrum data. The method further includes determining, by the processing device, a first value based on the first sensor data. The first value corresponds to a first amount of a product disposed along a surface of the wall at a first time. The method further includes determining, by the processing device, a first update of a first process operation associated with the first processing chamber based on the first value. The method further includes performing, by the processing device, one or more of: (i) preparing a notice indicating the first update for presentation on a graphical user interface (GUI), or (ii) causing the first process operation to be performed in accordance with the first update.

[0007] Additional or related embodiments described herein encompass a system including a processing chamber including a first wall, an inner surface of the first wall positioned within the processing chamber. The system further includes an optical sensor coupled to the processing chamber. The optical sensor is configured to acquire spectral data associated with the inner surface. The system further includes a memory and a processing device coupled to the memory. The processing device is for receiving first sensor data indicative of a condition of the wall corresponding to the first processing chamber. The first sensor data includes optical spectral data. The processing device is further for determining a first value based on the first sensor data. The first value corresponds to a first amount of product disposed along a surface of the wall at a first time. The processing device is further for determining a first update of a first process operation associated with the first processing chamber based on the first value. The processing device is further for performing one or more of (i) preparing a notification indicative of the first update for presentation on a GUI, or (ii) causing the first process operation to be performed in accordance with the first update.

[0008] Additional or related embodiments described herein encompass a non-transitory machine-readable storage medium storing instructions that, when executed, cause a processing device to receive first sensor data indicative of a condition of a wall corresponding to a first processing chamber. The first sensor data includes optical spectrum data. The processing device is further for determining a first value based on the first sensor data. The first value corresponds to a first amount of a product disposed along a surface of the wall at a first time. The processing device is further for determining a first update of a first process operation associated with the first processing chamber based on the first value. The processing device is further for performing one or more of (i) preparing a notification indicative of the first update for presentation on a GUI, or (ii) causing the first process operation to be performed according to the first update.

[0009] These and other aspects of the present disclosure provide numerous other features, which will become more fully apparent from the following detailed description, the claims, and the accompanying drawings.

[0010] The present disclosure is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings in which like references indicate similar elements. It should be noted that in this disclosure, various references to "an" or "one" embodiment do not necessarily refer to the same embodiment, but such references mean at least one. [Brief description of the drawings]

[0011] [Figure 1] 1 is a top schematic view of an exemplary processing system, according to some embodiments. [Diagram 2] 1 illustrates a simplified side view of a system for monitoring a processing chamber, according to some embodiments. [Diagram 3] 3 is a close-up of a side view of the system of FIG. 2 showing transparent crystals and transparent deposition thin films embedded in the walls and liner of the processing chamber according to some embodiments. [Figure 4] FIG. 1 is a block diagram illustrating an example system architecture in which embodiments of the present disclosure can operate. [Figure 5A] 4 is a flow diagram of a method for determining an updated process operation based on chamber wall values ​​according to some embodiments. [Figure 5B] 1 is a flow diagram of a method for using light sensor data in triggering the execution of corrective actions, according to some embodiments. [Figure 5C] 1 is a flow diagram of a method for maintaining consistency of a processing chamber, according to some embodiments. [Figure 5D] 1 is a flow diagram of a method for detecting and correcting anomalies and / or inconsistencies in a bank of processing chambers, according to some embodiments. [Figure 5E] 1 is a flow diagram of a method for utilizing light sensor data generated over time, according to some embodiments. [Figures 6A-6D] 1 is a flow diagram of a process associated with determining a process operation update based on a chamber wall value in accordance with some embodiments. [Figure 7] 1 is a flow diagram of a method for improving processing in a processing chamber using one or more optical film properties of a deposited thin film on a transparent thin film, according to some embodiments. [Figure 8] FIG. 1 illustrates a model training workflow and a model application workflow for prediction and simulation of chamber operation updates in accordance with some embodiments. [Figure 9] FIG. 1 is a block diagram of an exemplary computing device, according to some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] The embodiments described herein relate to systems and methods for process control and process learning using optical wall process sensors (OWPS). For example, in embodiments, data from an OWPS can be used to update process parameters of one or more processes. These systems and methods can use one or more sensors (e.g., OWPS) that can measure one or more properties of a coating on a chamber wall, and can optionally use machine learning techniques to analyze data from such sensors. In an embodiment, these systems and methods can use interference reflectance measurements on a liner (or inner wall) of a processing chamber to determine the condition and / or process state of the processing chamber. For example, reflectance measurements can be used to determine one or more optical thin film properties of a thin film layer (e.g., process film layer) deposited on a liner surface or on a reflector surface adjacent to and substantially corresponding to the liner. The optical properties of the film can be determined before, during, and / or after processes such as deposition processes, etching processes, and cleaning processes. The optical properties of the film (e.g., optical thin film properties) can include, but are not limited to, the film thickness, the refractive index (n) and extinction coefficient (k) values, chemical characteristics, and the like.

[0013] In some systems, characteristics (e.g., one or more property values) of a film deposited on a surface in a processing chamber will be monitored. In some embodiments, the deposition or etching of a film on a surface of the processing chamber can be used to infer an etching or deposition rate of a material on a substrate. In some embodiments, the properties of this film can be used to infer the performance of a processing operation. In some embodiments, the processing operation can be a cleaning operation, such as a cleaning operation performed between processing of several substrates.

[0014] In conventional systems, coupons can be utilized to monitor thin films deposited on surfaces within a processing chamber. A removable object (e.g., coupon or thin sheet) can be placed on a surface (e.g., inner wall, liner, etc.) within the processing chamber. After some time has passed, the coupon can be removed from the chamber and the film on the surface of the coupon can be analyzed. This implementation can be very cumbersome in several ways. Placing the coupons in the processing chamber and removing them from the chamber can include exposing the interior of the processing chamber to ambient atmospheric conditions. In some embodiments, returning to processing conditions (e.g., clean interior, vacuum interior, etc.) can be expensive in terms of time consumed, energy consumed, wear of components, etc. In some embodiments, returning to an operational state can include cleaning operations, chamber conditioning operations, chamber coating operations, and / or chamber seasoning operations after exposing the interior of the processing chamber to ambient conditions. Utilizing coupon-based methods can limit the frequency with which the film evolution can be analyzed. Additionally, coupons can be measured between processing operations, but not during the operation of the processing chamber (e.g., during a deposition process, an etch process, or a cleaning process).

[0015] Embodiments of the present disclosure may address at least some of these deficiencies of conventional solutions. In some embodiments, measurements may be performed by an optical sensor configured to detect a signal from a film on an interior surface of the processing chamber. In some embodiments, determining the state of the processing chamber is performed during the course of processing of a substrate in the processing chamber. In some embodiments, determining the state of the processing chamber is performed before, during, and / or after processing of a substrate. In some embodiments, determining the state of the processing chamber is performed before, during, and / or after a cleaning operation or other operation that is performed without a substrate in the processing chamber.

[0016] In some embodiments, determining the condition of the processing chamber can involve adapting the analysis to compensate for the presence of plasma in the processing chamber, which may be a source of electromagnetic radiation. The plasma includes corrosive gases used to etch the substrate. In some embodiments, one or more optical thin film properties are determined before or after processing, and there is no need to compensate for residual plasma during an active process.

[0017] In various embodiments, the partial condition of the liner of the processing chamber is indicative of the condition of the surface of the processing chamber as a whole. For example, the condition can be indicative of process changes (or variations) for which compensation should be made (e.g., via calibration, cleaning, and / or replacement of processing equipment or other structures of the processing chamber). The process changes can affect the performance and yield of processed substrates. By tracking such process changes using OWPS data, the systems and methods described herein can use such process change (e.g., variation) information to compensate for the variations and / or determine when and / or how to perform maintenance (e.g., cleaning operations).

[0018] In another example, the thickness of a deposited film layer on an inner surface of a processing chamber can be correlated with the deposited thin film on a substrate being processed (e.g., measured during processing). Characterizing the deposited thin film on one or more inner surfaces of a processing chamber can be extended to identify and manage deposition rates, deposition amounts, process start and / or end conditions, process frequency, and the like. For example, surface conditions within a processing chamber can be used to ascertain optimal frequency and duration of cleaning procedures. The frequency and duration of other operations, such as chamber conditioning operations, chamber coating operations, chamber seasoning operations, and the like, can also be ascertained. Further advantages include reduced disruption of the plasma volume when performing reflectance measurements (e.g., during substrate processing). These embodiments can include one or more of the above advantages with minimal modifications to existing liner and processing chamber designs.

[0019] Some embodiments of the disclosed systems and methods include a light source (e.g., a broadband light source), a spectrometer (or other apparatus for recording and measuring the incident light as an analytical method), and a light coupling device such as a collimator or mirror. These embodiments, as well as the process chamber that is part of the disclosed system, can further include a transparent material (e.g., a crystal) embedded at least partially in the wall and liner of the process chamber. The transparent material can be made of a transparent ceramic material, or can be made of a transparent material such as, for example, sapphire, diamond, quartz, or silicon carbide. A transparent thin film (or substrate) can be formed on a surface of the transparent material exposed inside the process chamber. The transparent material can act as a window for the process chamber. The transparent thin film can be deposited via deposition using atomic layer deposition (ALD), chemical vapor deposition (CVD), plasma vapor deposition (PVD), or another method of depositing a thin film on a surface. The transparent thin film can include yttrium oxide, aluminum oxide, zirconium oxide, combinations thereof, or another transparent material (e.g., a transparent crystalline material, etc.). In one embodiment, the transparent thin film is approximately flush with the surface of the liner. The transparent thin film can also have chemical properties that substantially match those of the liner.

[0020] In various embodiments, during processing in the chamber, the light coupling device can direct light from the light source through the transparent material, through the transparent thin film, and to the process film layer deposited on the transparent thin film. In this manner, light passing through both the transparent crystal and the transparent thin film is reflected back from their respective surfaces and combines with light reflected from the surface of the process film layer deposited on the transparent thin film. This reflected light can include a first spectrum consistent with the condition of the process film layer deposited on the liner.

[0021] In some embodiments, the optical coupling device directs the reflected light (e.g., a combination of light reflected from various surfaces) to an optical detection device. In some embodiments, the optical coupling device focuses this combination of reflected light into a fiber optic cable coupled to a spectrometer. The spectrometer can detect a first spectrum of the focused light representative of the process film layer, and the first spectrum can be used to determine the optical thin film properties. The spectrometer can also detect a second instance (e.g., spectrum) of the focused light when the light source is turned off during processing in the chamber. This second measurement can be utilized for background subtraction, normalization, etc. In some embodiments, the second optical measurement corresponds to an optical emission spectroscopy (OES) of the plasma in the process chamber being measured, and the second optical measurement can be removed from the first measurement to obtain a reflectance measurement signal that isolates the effect of the process film layer deposited on the surface of the liner (e.g., a transparent material and / or a transparent thin film, optionally below the process film layer).

[0022] Embodiments of the system and method may also include a processing device (or controller) coupled to the spectrometer. The processing device may be adapted to receive the first measurement (e.g., spectrum) and the second measurement and calculate the reflectance measurement data by subtracting (e.g., subtracting) the second data from the first data. The processing device may calculate the reflectance measurement signal by dividing (e.g., dividing) the reflectance measurement data by a reference spectrum that normalizes the reflectance measurement data. The reference spectrum may be obtained under known conditions, such as during initial installation of the system, immediately after a maintenance or seasoning event, etc.

[0023] The processing device can then fit the reflectance measurement signal to a thin film optical model to determine information including one or more properties (e.g., optical thin film properties) of the process film layer. Such properties can include, but are not limited to, thickness (e.g., optical thickness), refractive index (n) and extinction coefficient (k) values, reflectance and / or transmittance (all of which can be wavelength dependent), and / or composition material. The one or more properties can then inform a determination of a particular endpoint associated with the process performed in the processing chamber, including, but not limited to, whether to adjust the deposition rate, when to stop deposition of chemicals or plasma, when to start cleaning the processing chamber, when to stop cleaning the processing chamber, and the amount of process variation or elapsed chamber life. In some embodiments, the one or more film properties can indicate anomalies in the processing equipment, such as abnormal or damaged input materials, hardware failures, aging or variations in the processing equipment.

[0024] In one embodiment, for example, the processing device (or controller) compares the thickness of the process film layer to a baseline measurement, e.g., obtained when the processing chamber was first put into operation. If this value varies beyond a threshold variation (e.g., for the thickness of the process film layer), the processing device can trigger a process in the processing chamber to correct the deposition or etch rate of the process film layer (e.g., a cleaning process, a calibration process, etc.). The processing device can also modify the process to restore the process state and / or alert users of the processing chamber of the process change, among other actions discussed below.

[0025] In some embodiments, the processing device (or controller) determines a dimensionless value corresponding to a characteristic of a process film layer on the transparent thin film and / or on the chamber wall / liner. The film characteristic can include thickness, optical properties, chemical properties, etc. The dimensionless value can indicate a trend of the process film layer characteristic. For example, a higher dimensionless value can correspond to a larger thickness. The dimensionless value can correspond to a ratio and / or a percentage, e.g., a ratio of a maximum allowable value. The dimensionless value can be related to a critical wall thickness (e.g., a critical thickness of the process film layer). The dimensionless value can correlate the process film layer thickness to a maximum and / or minimum allowable thickness (e.g., in a ratio or percentage). The dimensionless value can be an index corresponding to the process film layer characteristic. The processing device can determine an update of the process operation based on the dimensionless value. The processing device can calibrate one or more sensors of the processing chamber based on the dimensionless value. In addition, the processing device can receive sensor data corresponding to a second processing chamber (e.g., another processing chamber) and determine a second dimensionless value corresponding to a characteristic of the process film layer in the second processing chamber. The first and second dimensionless values ​​can be normalized to each other by the processing device. The processing device can normalize the first and second dimensionless values ​​by a data normalization process. The processing device can further calibrate sensors of both processing chambers based on the first and second values. The processing device can further update a recipe operation (e.g., a cleaning operation, a deposition operation, an etching operation, etc.) based on the first and second values.

[0026] The embodiments of the present disclosure address the deficiencies of the current technology by providing a technique for updating process parameters based on optical chamber wall values. The process parameter updates can enable increased optimization of process operations (e.g., substrate yield) associated with a processing chamber of a manufacturing system. The embodiments of the present disclosure provide a technique for updating process operations at an improved measure than conventional processing systems. The techniques provided can result in more efficient operation of the processing chamber, leading to greater throughput. In addition, the embodiments of the present disclosure provide closer monitoring of the conditions of the processing chamber by determining a dimensionless value based on sensor data indicative of the thickness of material on the interior surface of the processing chamber. As chamber quality changes or chamber performance deteriorates, closer monitoring of the processing chamber conditions can reduce downtime of the processing chamber by scheduling maintenance, cleaning, seasoning, tuning, etc. The reduced downtime can increase the overall throughput and efficiency of the processing chamber over conventional systems.

[0027] Additionally, embodiments of the present disclosure provide improved sensor accuracy by calibrating one or more sensors of a processing chamber based on a dimensionless value derived from measurements of the OWPS sensor. Calibrating the sensor based on a dimensionless value can offset sensor variations. In some embodiments, sensor variations can be caused by material buildup on the interior surfaces of the processing chamber. Improved sensor accuracy can improve the precision with which process operations and / or procedures are performed. Overall, the present disclosure allows for greater control over process operations and parameters, resulting in increased process precision, consistency, and throughput. Thus, a system controller can consume fewer computing resources (e.g., processing cycles, memory space, etc.) to perform substrate processes on a manufacturing system, thereby improving overall efficiency and reducing overall latency of computing systems associated with the manufacturing system.

[0028] The disclosed method includes receiving, by a processing device, first data from an optical sensor of the processing chamber. The method further includes processing the first data to obtain second data. The second data includes an indication of a condition of a coating on an interior surface of the processing chamber. The method further includes generating a performance indication of a processing operation of the processing chamber in consideration of the second data. The method further includes causing a corrective action to be taken in consideration of the performance indication of the processing chamber. The corrective action may include adjusting a processing recipe, e.g., adjusting a processing recipe endpoint, scheduling maintenance, alerting a user, etc.

[0029] A further aspect of the present disclosure includes a non-transportable, machine-readable storage medium. The storage medium stores instructions. The instructions, when executed by a processing device, cause the processing device to perform an operation. The operation includes receiving, by the processing device, first data from an optical sensor of the processing chamber, the first data associated with a first operation of the processing chamber. The operation further includes receiving second data from the optical sensor of the processing chamber, the second data associated with a second operation of the processing chamber. The operations stored in the storage medium further include generating first and second performance indications of the processing chamber based on the first and second data. The operations stored in the storage medium further include determining one or more performance differentials between the first and second operations. Based on the one or more differentials, the operations stored in the storage medium further include causing a corrective action to be performed. The corrective action can include adjusting a processing recipe, for example, adjusting an endpoint of the processing recipe. The corrective action can include scheduling maintenance, such as corrective or preventative maintenance. The corrective action can include providing a warning to a user.

[0030] Further aspects of the present disclosure include a method, the method including receiving, by a processing device, first data from an optical sensor of a first processing chamber. The method further includes receiving second data from an optical sensor of a second processing chamber. The method further includes generating a first performance indication of the first processing chamber based on the first data and generating a second performance indication of the second processing chamber based on the second data. The method further includes determining one or more performance differentials between the first processing chamber and the second processing chamber. The method further includes causing a corrective action to be taken based on the one or more differentials. The corrective action may include updating a process recipe, such as updating a process recipe endpoint, scheduling maintenance, providing a warning to a user, etc.

[0031] FIG. 1 is a top schematic view of an exemplary processing system 100 according to one aspect of the disclosure. The processing system 100 includes a transfer chamber robot 101 and a factory interface robot 121, each adapted to pick up and place a substrate 110 (sometimes referred to as a "wafer" or "semiconductor wafer") to and from a destination within an electronic device processing system, such as the processing system 100 shown in FIG. 1. However, any type of electronic device substrate, mask, or other silica-containing substrate (collectively referred to herein as a "substrate") can be carried and transferred by the disclosed robot. For example, the destination of the substrate 110 can be one or more of the processing chambers 103 and / or load lock apparatuses 107A, 107B, which can be distributed around and coupled to the transfer chamber 114. As shown, the transfer of the substrate can be through, for example, a slit valve 111.

[0032] The processing system 100 may further include a main frame 102 that includes a transfer chamber 114 and at least two processing chambers 103. The housing of the main frame 102 includes the transfer chamber 114 therein. The transfer chamber 114 may include a top wall (not shown), a bottom wall (floor) 139, and side walls, and in some embodiments may be maintained, for example, by a vacuum. In the embodiment shown, the transfer chamber robot 101 is attached to the bottom wall (floor) 139. However, the transfer chamber robot 101 may also be attached elsewhere, such as to the top wall.

[0033] In various embodiments, the processing chamber 103 can be adapted to perform any number of processes on the substrate 110. These processes can include deposition, oxidation, nitridation, etching, polishing, cleaning, lithography, metrology (e.g., integrated metrology), and the like. Other processes can be performed as well. The load lock apparatus 107A, 107B can be adapted to interface with a factory interface 117 or other system components, which can receive the substrate 110 from a substrate carrier 119 (e.g., a front opening integrated pod (FOUP)) that can be docked, for example, to a load port of the factory interface 117. A factory interface robot 121 (shown in dashed lines) can be used to transfer the substrate 110 between the substrate carrier 119 and each of the load lock apparatuses 107A, 107B. The transfer of the substrate 110 can be performed in any order or direction. Factory interface robot 121, in some embodiments, may be the same as (or similar to) transfer chamber robot 101, but may further include mechanisms to allow the factory interface robot to move in either lateral direction as indicated by arrow 123. Any other suitable robot may also be used as factory interface robot 121. In some embodiments, system 100 may be coupled (e.g., interfaced) with a metrology system, for example an integrated metrology system.

[0034] In an embodiment, as an exemplary illustration of a robot, the transfer chamber robot 101 includes at least one arm 113 (e.g., a robot arm) and at least one end effector 115 coupled to the arm 113. The end effector 115 is controllable by the transfer chamber robot 101 to lift the substrate 110 from the load lock apparatus 107A or 107B, guide the substrate 110 through one of the slit valves 111 of the processing chambers 103, and precisely place the substrate 110 on a substrate support of the processing chamber 103.

[0035] In various embodiments, one or more of the processing chambers 103 can include a transparent window 120 (e.g., a transparent material such as a transparent crystal), at least a portion of which is embedded within the walls and liner 124 (e.g., an inner wall) of the processing chamber 103. In the disclosed embodiments, light can be collimated and directed to the transparent window 120 to generate reflected light. The reflected light can then travel back through the transparent window 120, as discussed in more detail with reference to FIGS. 2-3. The reflected light can then be collected for further analysis. In some embodiments, the reflected light can be focused into a fiber optic cable coupled to a spectrometer 125 for analysis, e.g., spectral analysis. The spectrometer 125 can perform a reflectance measurement to determine one or more spectra of the focused light, and such one or more spectra can be used to determine at least one optical property of a process film layer deposited on the transparent window 120 (e.g., on a transparent thin film of the transparent window), whether during or after substrate processing. The transparent window 120 and the spectrometer 125 may comprise an optical sensor associated with the process chamber 103. The transparent window 120 and the spectrometer 125 may comprise an optical wall process sensor (OWPS). The optical sensor may detect at least one characteristic of a material on an inner surface of the process chamber 103 (e.g., on the transparent window). In some embodiments, through optical calculations of thin films by a processing device (e.g., processing device 130), the optical sensor may provide sensor data indicative of a thickness of a material on an inner surface of the process chamber 103.

[0036] A controller 109 (e.g., a tool and equipment controller) can control various aspects of the processing system 100, such as the gas pressure in the processing chamber 103, the individual gas flows, spatial flow ratios, the temperature of various chamber parts, and the radio frequency (RF) or electrical conditions of the processing chamber 103. The controller 109 can receive signals from and send commands to the factory interface robot 121, the transfer chamber robot 101, one or more sensors, and / or other processing components of the processing system 100. Thus, the controller 109 can control the start and stop of processing, adjust the deposition rate, type, or mix of deposition compositions, etc. The controller 109 can further receive and process sensed data from the various sensors.

[0037] In various embodiments, the controller 109 includes (or is coupled to) a processing device 130 and is coupled to the spectrometer 125. The processing device 130 can be configured to receive and process sensor data, including results of reflectance measurements performed by the spectrometer 125. The processing device 130 can be configured to accept multiple measurements and perform signal processing, for example, to separate the effects of the thin film (e.g., from the optical effects of plasma generation) by receiving data collected with and without an external light source. The processing device 130 can calculate a reflectance measurement signal by subtracting a second measurement from a first measurement, for example, the second spectrum from the first spectrum. The processing device can then fit the reflectance measurement signal to a thin film optical model to determine information including one or more optical thin film properties of the process film layer. From the thin film optical model, the processing device 130 can determine one or more characteristics of the process product deposited on the wall of the processing chamber 103. The characteristics including thickness (e.g., optical thickness), reflectance, chemical characteristics, etc. can be calculated by the processing device 130. Depending on the results of analyzing the one or more optical film characteristics, the processing device 130 (e.g., controller 109) can perform one or more actions. For example, the processing device 130 can guide the processing chamber 103 in a process change or adjustment. For example, the controller 109 can adjust process parameters or settings, such as, for example, the deposition rate, the type or mix of deposition compositions, the timing of performing a cleaning process in the processing chamber, and other actions discussed in more detail with reference to FIG. 7. Additionally, in some embodiments, the controller 109 can trigger a calibration of one or more sensors of the processing chamber 103 based on the results of analyzing the one or more optical film characteristics, can perform chamber matching based on the results of analyzing the one or more optical film characteristics, or can identify chamber variations based on the results of analyzing the one or more optical film characteristics.

[0038] In some embodiments, the processing device 130 can determine a dimensionless value based on the results of analyzing one or more optical film properties. In some embodiments, the dimensionless value corresponds to a first amount of a process product deposited on the transparent thin film (e.g., a thickness of a process film layer). The dimensionless value can indicate an amount of a process product deposited on an inner surface of the processing chamber 103. The processing device 130 can update a recipe (e.g., a cleaning recipe, an etching recipe, a deposition recipe, etc.) based on the dimensionless value. In addition, the processing device 130 can normalize the dimensionless value with a dimensionless value corresponding to one or more other processing chambers 103 to further update the recipe or calibrate one or more sensors. Based on the normalized dimensionless value, the recipe of the other processing chamber 103 can be updated and the sensor can be calibrated.

[0039] The controller 109 and / or the processing device 130 may be and / or include a computing device such as a personal computer, a server computer, a programmable logic controller (PLC), a microcontroller, etc. The controller 109 and / or the processing device 130 may include (or be) one or more processing devices that may be general-purpose processing devices such as a microprocessor, a central processing unit, etc. More specifically, the processing device may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, or a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. The processing device may also be one or more special-purpose processing devices such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, etc. The controller 109 and / or the processing device 130 may include a data storage device (e.g., one or more disk drives and / or solid state drives), a main memory, a static memory, a network interface, and / or other components. The processing device 130 may execute instructions to perform any one or more of the methods and / or embodiments described herein. These instructions may be stored on a computer-readable storage medium, which may include a main memory, a static memory, a secondary storage, and / or a processing device during execution of the instructions.

[0040] FIG. 2 illustrates a simplified side view of a system 200 for monitoring a processing chamber, according to one aspect of the disclosure. The system 200 can include an optical sensor that produces optical data indicative of one or more characteristics of a material on an interior surface of the processing chamber. In some embodiments, the optical sensor can collect spectral data. In some embodiments, the optical sensor can be configured to generate data indicative of an amount of material deposited on a transparent window (e.g., a transparent material or a transparent crystal) in proximity to the optical sensor. In some embodiments, the optical sensor can be configured to generate data indicative of a thickness of a film deposited on an interior surface of the processing chamber, which relates to a measurement indicative of, for example, an amount of material.

[0041] The system 200 can include, for example, a process chamber 103, which includes a liner 124 as shown in FIG. 1. The process chamber 103 can include a wall 222 to which the liner 124 is attached. The liner 124 is specifically designed to have high chemical resistance to the chemistry of the process carried out in the chamber 103. The chemical resistance of the liner 124 can minimize physical or chemical changes to the liner 124. The chemical resistance of the liner 124 can increase the life (e.g., time between maintenance, time between replacement) of the liner 124. Additionally, as shown, at least a portion of the transparent window 120 can be embedded within the wall 222 and the liner 124. The transparent window 120 can be made of a transparent ceramic material or can be made of a durable transparent material such as sapphire, diamond, quartz, silicon carbide, or a combination thereof.

[0042] In an embodiment, system 200 further includes a light source 201 (e.g., a broadband light source or other source of electromagnetic radiation), an optical coupling device 204 (e.g., a collimator or mirror), a spectrometer 225, a controller 109, and a processing device 130. The light source 201 and the spectrometer 225 may be optically coupled to the optical coupling device 204 through one or more couplings 232, which may include fiber optic cables.

[0043] In some embodiments, the light source 201 is a flash lamp. The light source 201 can generate illumination that is predictable, e.g., with respect to power, color, spectrum, etc. In some embodiments, the wavelengths generated by the light source 201 can be limited, e.g., the wavelengths can be limited to a wavelength range of 200-800 nanometers (nm). In some embodiments, the light source 201 can include a pulsed source, such as a pulsed flash lamp. In some embodiments, the light source 201 is a pulsed xenon source. In some embodiments, the power of the light source 201 can be predictable, e.g., the light source 201 can be a flash lamp that exhibits a sigma flash power variation of less than 0.5% over the entire wavelength range in an open loop. In other embodiments, the wavelength range and power variation can be different than those discussed above and still be within the scope of the present disclosure. In some embodiments, the emitted wavelengths can include near infrared wavelengths. For some practical light sources, the variable power control of the light source 201 can be 45%-100% while providing less than 8% power reduction over a life of at least one year, which can provide more than a billion flashes. Additional or different embodiments are contemplated for light source 201 having varying power and wavelengths.

[0044] In various embodiments, the optical coupling device 204 can be adapted to collimate or otherwise transmit light in two directions along the optical path. The first direction can include light from the light source 201 being collimated and transmitted through the transparent window 120 into the process chamber 103. The second direction can be light reflected from the transparent window 120 traveling back into the optical coupling device 204, as discussed in more detail with reference to FIG. 3. The reflected light can be focused into a coupler 232 (e.g., a fiber optic cable) and thus directed along the optical path in a second direction to the spectrometer 225. Additionally, the coupler 232 can be coupled between the spectrometer 225 and the light source 201 to efficiently transmit light from the light source 201 back through the transparent window 120 to the spectrometer 225.

[0045] In one embodiment, the spectrometer 225 also has a detection range similar to the wavelength range of the light source. In some embodiments, the spectrometer 225 can have a detection range (e.g., spanning the range of wavelengths over which the spectrometer operates, e.g., visible, UV, infrared, etc.), a wavelength resolution (e.g., indicating the ability of the spectrometer 225 to separate different wavelengths of light), a sampling rate, etc. The characteristics of the spectrometer 225 can be tailored to the analysis of the chamber type, recipe type, coating type, etc. The spectrometer 225 can be calibrated according to the same schedule as the light source 201, or can be calibrated and / or maintained more or less frequently. The spectrometer 225 can be adapted to detect the spectrum of reflected light received from the optical coupling device 204, e.g., light reflected back from the transparent window 120 and focused by the optical coupling device 204 into the coupler 232.

[0046] In various embodiments, the controller 109 includes or is coupled to the processing device 130 and includes or is coupled to the memory 134 or other computer storage. The controller 109 can be communicatively coupled to the light source 201, the spectrometer 225, and the process chamber 103. The controller 109 can instruct the light source 201 to turn on the flash and then receive a first measurement (e.g., spectrum) from the spectrometer 225. The controller 109 can also leave the light source off and receive a second measurement from the spectrometer 225 when the light source 201 is off. The second measurement can represent the OES of the plasma or another light-emitting chemical process in the process chamber. The processing device 130 can subtract the second measurement from the first measurement to determine a reflectance measurement signal at a given instant. The processing device 130 can then mathematically fit the reflectance measurement signal to one or more thin film models to determine one or more optical thin film properties of the process film layer deposited on the transparent thin film of the transparent crystal 120.

[0047] In some embodiments, the one or more optical thin film properties include a deposited film thickness (e.g., optical thickness) and values ​​of a refractive index (n) and an extinction coefficient (k). The refractive index is the ratio of the speed of light in a vacuum to the speed of light in the process film layer. The extinction coefficient is a measure of how much light is absorbed in the process film layer. In embodiments, the processing device 130 can use the n and k values ​​to determine the composition of the process film layer. The processing device 130 can be further configured to analyze data of the one or more optical film properties and trigger a new process (e.g., a cleaning or maintenance process) or update a current process (e.g., a deposition process, a cleaning process, an etching process, etc.) in the processing chamber 103 based on the analysis. The processing device 130 can determine a dimensionless value based on the analysis data and can trigger or update based on the dimensionless value. Such updates can include warnings, which are discussed in more detail with reference to FIG. 7.

[0048] 3 is an enlarged portion of a side view of the system of FIG. 2 showing the transparent window 120 embedded in the wall 222 and the liner 124 of the processing chamber 303, according to one embodiment. The transparent window 120 can have a proximal end and a distal end, the proximal end having a proximal face 120A that receives light from the optical coupling device 204, and the distal end 120B having a distal face 120B that reflects light back through the proximal face 120A to the optical coupling device 204. In some embodiments, the proximal face 120A can also reflect light. The distal face 120B of the transparent window 120 can be located proximate to the processing chamber 303 and, in one embodiment, can be approximately flush with the inner surface of the wall 222.

[0049] In various embodiments, the transparent window 120 can be shaped like a plug (e.g., plug-shaped) and can include a shaft 301 and a flange 305 integrally formed or attached to the shaft 301. The plug can be, for example, a sapphire plug, a diamond plug, a quartz plug, or a silicon carbide plug. The shaft 301 and / or the flange 305 can be square, rectangular, cylindrical, or other shapes. For ease of illustration, both the shaft 301 and the flange 305 are shown as cylindrical. As discussed, the transparent window 120 can be made from a transparent crystalline material, such as sapphire, diamond, quartz, silicon carbide, etc., to provide a hard material that resists changes and to reflect light. The shaft 301 can be embedded in the wall 222 and the liner 124 such that the distal face is exposed to the interior of the processing chamber 303.

[0050] The flange 305 can abut the outer surface of the wall 222. In one embodiment, a seal 313 is positioned between the flange and the outer surface of the wall 222. The seal 313 can be, for example, an O-ring seal, a rectangular seal or gasket seal, a valve seal, or the like. The material of the seal 313 can be propylene diene monomer, a fluoroelastomer, or the like. In some embodiments, the seal can be arranged in a different manner, for example along the shaft 301, and still be within the scope of the present disclosure. The flange 305 can have an inner surface that physically contacts the seal 313, the inner surface including a surface roughness (Ra) of less than 20 nanometers or a scratch-dig value of at least 80 / 50, the scratch-dig value being an optical specification used for surface roughness. The very smooth inner surface of the flange 305 can help provide a seal between the ambient atmosphere outside the processing system 100, 200 and the very clean filtered air of the processing chamber 303, which can be under vacuum.

[0051] In various embodiments, a transparent thin film 307 can be deposited on the distal surface 120B of the shaft 301 of the transparent window 120, for example via atomic layer deposition (ALD). In some cases, the transparent thin film 307 is deposited on the transparent window 120 before the transparent window 120 is embedded in the wall 222 and the liner 124. The transparent thin film 307 can be exposed inside the processing chamber 303. In some embodiments, the transparent thin film 307 is substantially flush with the liner 124 and has chemical properties that substantially match those of the liner, as previously discussed. In various embodiments, the transparent thin film 307 includes yttrium oxide, aluminum oxide, zirconium oxide, or combinations thereof to create a similar transparent alloy. In one embodiment, the transparent thin film 307 includes yttrium oxide and is between 10 nm and 1 micrometer (μm) thick, for example 270 nm thick, and is deposited to a known thickness, which is useful for analyzing the reflected light.

[0052] In various embodiments, the deposition process (e.g., ALD) of the transparent thin film 307 is carefully controlled to ensure that it mimics the material of the liner 124, but in a different physical state, e.g., smoother and flatter, to facilitate light reflection from its surface. In this way, the transparent thin film 307 reflects light to provide data that can be used to generate an accurate assessment of the chamber conditions. In some embodiments, the thinness and smoothness of the transparent thin film 307 increases the signal-to-noise ratio (SNR) of measurements of light passing through the distal surface, facilitating the ability to test thinner process film layers 309 in the processing chamber 303. In embodiments, the transparent thin film 307 also passivates and protects the material of the transparent window 120, allowing the expensive transparent window 120 to last longer in the corrosive processing environment of the processing chamber 303.

[0053] In some embodiments, a process film layer 309 (e.g., a process thin film layer) is deposited during processing in the processing chamber 303. The process film layer 309 can include a single film or a film stack to be measured by the spectrometer 225. The optical coupling device 204 can collimate the light provided by the light source 201 and direct the collimated light 304A through the proximal surface 120A of the transparent window 120, e.g., the outer surface of the flange 305. The incoming light is then reflected from the transparent window, e.g., the distal surface 120B of the shaft 301, the surface of the transparent thin film 307, and the surface of the process film layer 309. These three reflected light sources return together, shown as reflected light 304B, and re-enter the optical coupling device 204. The optical coupling device 204 can focus the reflected light into the coupler 232 to enter the spectrometer for measurement (see FIG. 2). In some embodiments, additional reflections may be collected and / or compensated for in the analysis, e.g., reflections from proximal surface 120A, reflections from the interface of flange 305 and shaft 301, etc. In some embodiments, one or more reflections may be reduced or eliminated. For example, reflections from surface 120A may be reduced by application of an anti-reflective coating, the design and / or construction of transparent window 120 may eliminate an optical interface between flange 305 and shaft 301, etc.

[0054] The return optical signals to each interface (e.g., the aforementioned surfaces that can reflect light) can all combine and result in optical constructive and destructive interference. The exact nature of this interference depends on the relative thickness of each layer, e.g., the transparent window 120, the transparent thin film 307, and the process film layer 309 (or film stack). The final result measured by the spectrometer 225 (after normalization to an initial reference) can include a spectrum with interference fringes. These fringes can be fitted to a mathematical optical thin film model, which can then determine one or more thin film optical parameters of the process film layer 309, such as thickness (e.g., optical thickness), values ​​of n and k. The values ​​of n and k can be used to determine the composition of the process film layer 309 (e.g., material deposited on the walls of the process chamber). Knowing the exact initial thickness and optical properties of the transparent thin film 307 contributes to fitting the spectrum to the thin film model. The thickness of the process film layer 309 can be incorporated into the calculation of the dimensionless values, and updates to process operations (e.g., cleaning operations, deposition operations, etching operations, etc.) can be made to take this into account.

[0055] 4 is a block diagram illustrating an example system architecture in which an embodiment of the present disclosure can operate. As shown in FIG. 4, the system architecture 400 includes a manufacturing system 402, a metrology system 410, a client device 450, a data store 440, a server 420, and a modeling system 470. The modeling system 470 can be part of the server 420. In some embodiments, one or more components of the modeling system 470 can be integrated in whole or in part into the client device 450. The manufacturing system 402, the metrology system 410, the client device 450, the data store 440, the server 420, and the modeling system 470 can each be hosted by one or more computing devices, including a server computer, a desktop computer, a laptop computer, a tablet computer, a notebook computer, a personal digital assistant (PDA), a mobile communication device, a mobile phone, a handheld computer, or a similar computing device.

[0056] The manufacturing system 402, the metrology system 410, the client device 450, the data store 440, the server 420, and the modeling system 470 can be coupled to each other (e.g., to perform the methods described herein) via a network 460. In some embodiments, the network 460 is a private network that provides each element of the system architecture 400 with access to each other and to other privately available computing devices. The network 460 can 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, and / or any combination thereof. In some embodiments, processing and / or data storage operations can be performed by remote cloud-based devices, virtual devices, etc., and the network 460 can enable access to and communication with these devices. Alternatively or additionally, any of the elements of system architecture 400 may be integrated together or otherwise coupled without the use of network 460.

[0057] The client device 450 can be or include any personal computer (PC), laptop, mobile phone, tablet computer, netbook computer, network-connected television ("smart TV"), network-connected media player (e.g., Blue-ray player), set-top box, over-the-top (OOT) streaming device, operator box, etc. The client device 450 can include a browser 452, applications 454, and / or other tools described and executed by other systems of the system architecture 400. In some embodiments, the client device 450 may be capable of accessing the manufacturing system 402, the metrology system 410, the data store 440, the server 420, and / or the modeling system 470 as described herein to communicate (e.g., transmit and / or receive) sensor data, processed data, data classifications, process result data, process update data, and / or representations of inputs and outputs of various process tools (e.g., the metrology tools 414, the data preparation tools 416, the chamber matching tools 422, the sensor matching tools 424, the recipe development tools 426, the process control tools 428, and / or the chamber value components 494) at various processing stages of the system architecture 400.

[0058] As shown in FIG. 4, the manufacturing system 402 includes process tools 404, process procedures 406, and a process controller 408. The process controller 408 can coordinate the operation of the process tools 404 to perform one or more process procedures 406. For example, the various process tools 404 can include specialized chambers such as etch chambers, deposition chambers (including chambers for atomic layer deposition, chemical vapor deposition, or plasma enhanced chemical vapor deposition), anneal chambers, etc. In another example, a machine can incorporate a sample transport system (e.g., a selective compliant assembly robotic arm (SCARA) robot, transfer chambers, front opening pods (FOUPs), side storage pods (SSPs), etc.) to transport samples between machines and process steps.

[0059] The process actions 406, which may also be referred to as process recipes, process operations, or process steps, may include various specifications for performing an operation by the process tool 404. For example, the process actions 406 may include process specifications such as duration of start-up of the process operation, process tools used for the operation, machine (e.g., chamber) temperatures, flow rates, pressures, etc., deposition sequence, etc. In another example, the process actions may include transferring instructions for transporting a sample to a further process step or to be measured by the metrology system 410. The one or more process actions 406 may include preventive maintenance actions, cleaning actions, adjustment actions, seasoning actions, etc.

[0060] The process controller 408 may include devices designed to manage and / or coordinate the actions of the process tool 404. In some embodiments, the process controller 408 is associated with a process recipe or set of process action instructions that, when applied as designed, results in a desired process outcome of the substrate process. For example, a process recipe may be associated with processing a substrate to produce a target process result (e.g., critical dimension, thickness, uniformity criteria, etc.). The process controller 408 may coordinate the execution of preventive maintenance actions (e.g., chamber cleaning), chamber recovery actions (e.g., chamber seasoning), etc. to return the chamber to a production state, and / or updates to process operations associated with the processing chamber. Additionally, the process controller 408 may cause updates to the calibration of sensors associated with the process tool 404 or the metrology system 410.

[0061] As shown in FIG. 4, the metrology system 410 includes a metrology tool 414 and a data preparation tool 416. The metrology tool 414 can include various sensors for measuring process results (e.g., critical dimensions, thickness, uniformity, etc.) and detecting parameters within the fabrication system 402. In some embodiments, the metrology tool 414 includes one or more optical sensors as shown in FIGS. 2-3. In some embodiments, the metrology tool 414 includes a sensor for measuring a thickness of a material on an inner surface of a processing chamber. The metrology tool 414 can include a tool for performing in-line metrology and can include an OWPS. The metrology tool 414 can include multiple sensors for detecting a condition of an environment of the processing chamber. For example, sensors disposed within one or more processing chambers can be used to measure conditions within the processing chamber, such as temperature, pressure, vacuum conditions, deposition layer thickness, etc. In another example, sensors disposed within the processing chambers can be used to measure a thickness of a process product deposited along a surface of a wall of the processing chamber. The metrology tool 414 can also include devices for measuring process results of a substrate processed using the fabrication system. For example, process results such as critical dimensions, thickness measurements (e.g., film layers from etching, deposition, etc.) of a substrate processed according to the process recipe and / or actions executed by the process controller 408 can be evaluated. In another example, various sensors can measure conditions in the chamber throughout the process operation.

[0062] The data preparation tool 416 may include process methods for extracting features and / or generating synthetic / processed data associated with data measured by the metrology tool 414. In some embodiments, the data preparation tool 416 may identify correlations, patterns, and / or anomalies in the metrology and / or process execution data. For example, the data preparation tool 416 may perform feature extraction, and the data preparation tool 416 may use a combination of measurement data to determine whether a criterion is met. For example, the data preparation tool 416 may analyze multiple data points of associated parameters (e.g., temperature, pressure, deposition layer thickness, vacuum conditions, plasma conditions, process product thickness, etc.) to determine whether rapid changes have occurred during an iteration of a process operation. In some embodiments, the data preparation tool 416 performs normalization across different sensor data associated with different process chamber conditions. Normalization may include processing the incoming sensor data to look similar across different chambers and sensors used to acquire the data. The data preparation tool 416 may create dimensionless values ​​based on the data collected by the metrology tool 414. In particular, the data preparation tool 416 can determine a dimensionless value corresponding to a characteristic (eg, thickness) of a process product film layer deposited on a surface of a wall of a processing chamber.

[0063] In some embodiments, the data preparation tools 416 can perform one or more of a process control analysis, a univariate limit violation analysis, or a multivariate limit violation analysis of the metrology data (e.g., obtained by the metrology tools 414). For example, the data preparation tools 416 can perform statistical process control (SPC) by utilizing statistics-based methods to monitor and control the process controller 408. For example, SPC can increase the efficiency and accuracy of process operation (e.g., by identifying data points that are within and / or outside of control limits).

[0064] In some embodiments, conditions of the processing chamber may be measured throughout a process operation. For example, a thickness of a process product deposited on an inner wall of the processing chamber may be measured during a process operation (e.g., during a process recipe operation, during a cleaning operation, etc.). In some embodiments, an amount of metrology data obtained during a given processing operation is increased. For example, additional sensors may be activated and / or currently activated sensors may obtain additional data during or immediately after processing of a wafer. In some embodiments, the process controller 408 may trigger measurements by the metrology tool 414 based on an operation to be performed by the process tool 404. For example, the process controller 408 may trigger the activation of one or more sensors (e.g., of the metrology tool 414) in response to an update of one or more parameters associated with the process. In another example, the process controller 408 may trigger the activation of the OWPS to measure an optical property of a process film layer that may be indicative of a thickness of the process film layer on an inner surface of a wall of the processing chamber.

[0065] In some embodiments, the extracted features, generated synthetic / processed data, and statistical analysis can be used in connection with the modeling system 470 (e.g., to train, validate, and / or test the model 490). Additionally and / or alternatively, the data preparation tool 416 can output data to the server 420 to be used by any of the chamber matching tool 422, the sensor matching tool 424, the strategy development tool 426, and / or the process control tool 428.

[0066] The data store 440 can be a memory (e.g., random access memory), a drive (e.g., hard drive, flash drive), a database system, or another type of component or device capable of storing data. The data store 440 can store one or more historical data 442 including historical sensor data 444 of previous chamber conditions and process results of substrates processed in the associated chamber conditions, historical chamber value data 446, and / or historical chamber update data 447. In some embodiments, the historical chamber value data 446 includes historical normalized chamber value data. In some embodiments, the historical sensor data 444 includes historical data from an OWPS. The historical sensor data 444 can include optical spectrum data indicative of a thickness of a process film layer deposited along an inner surface of a wall of the processing chamber. In some embodiments, the historical chamber update data 447 can include historical updates of one or more process operations associated with the processing chamber. For example, historical chamber update data 447 may include historical updates (e.g., cleaning recipe updates, etch recipe updates, deposition recipe updates, etc.) made to one or more process operations associated with the processing chamber based on historical chamber value data 446. In some embodiments, historical chamber update data 447 may be mapped to historical chamber value data 446. In some embodiments, historical data 442 may be used to train, validate, and / or test models 490 of modeling system 470 (see, e.g., FIG. 8 for example methods).

[0067] The server 420 may include one or more computing devices, such as a rack mount server, a router computer, a server computer, a personal computer, a mainframe computer, a laptop computer, a tablet computer, a desktop computer, etc. The server 420 may include a chamber matching tool 422, a sensor matching tool 424, a recipe development tool 426, and a process control tool 428.

[0068] Each of the chamber matching tool 422, the sensor matching tool 424, the recipe development tool 426, and the process control tool 428 can receive data from the data preparation tool 416. Each of the chamber matching tool 422, the sensor matching tool 424, the recipe development tool 426, and the process control tool 428 can receive chamber value data from the data preparation tool 416. The chamber value data can be associated with conditions of the processing chamber. In some embodiments, the chamber value data indicates a thickness of a process product deposited on an inner surface of the processing chamber. The chamber value data can correspond to an amount or characteristic of a process product deposited along an inner surface of a wall of the processing chamber. The chamber value data can indicate a net gain or loss of a film thickness of a process product added to or removed from an inner surface of the processing chamber during a processing operation (e.g., a cleaning operation, a deposition operation, an etching operation, etc.).

[0069] In some embodiments, the chamber matching tool 422, the sensor matching tool 424, the recipe development tool 426, and the process control tool 428 each receive chamber data from the process tool 404 and / or processed data output from the data preparation tool 416. In some embodiments, the data preparation tool 416 receives raw sensor data from the metrology tool 414. The raw sensor data can be combined with processed composite data from the data preparation tool 416. The chamber matching tool 422, the sensor matching tool 424, the recipe development tool 426, and the process control tool 428 each can generate updates to one or more process operations 406 and / or generate reports / notifications indicative of chamber matching, chamber variations, anomalies, etc. The updates and / or reports or notifications can be based on the chamber value data.

[0070] In one example, the chamber matching tool 422 can determine an update to a process operation of a first process chamber based on the chamber value data to more closely match the process operation performed on the first process chamber to a corresponding process operation performed on a second process chamber. Further details of the chamber matching tool 422 are discussed with reference to process 600C of FIG.

[0071] In another example, the sensor matching tool 424 can determine a calibration of one or more sensors associated with the process chamber based on the chamber value data. Additionally, the sensor matching tool 424 can determine a sensor calibration that compensates for sensor variation based on the chamber value data. Further details of the sensor matching tool 424 are discussed with reference to FIG. 6B process 600B.

[0072] In a further example, the recipe development tool 426 can determine updates to a process recipe based on the chamber value data. Specifically, the recipe development tool 426 can determine updates to one or more process parameters (e.g., run time, cleaning time, process knobs, etc.) to optimize a process recipe (e.g., process operation) based on the chamber value data. In one embodiment, the recipe development tool 426 determines updates to a cleaning recipe that is periodically performed to clean a process chamber based on the chamber value data. Further details of one example of the recipe development tool 426 are discussed with reference to process 600A of FIG. 6A.

[0073] As an additional example, the process control tool 428 can determine an update to an endpoint of a process operation (e.g., an endpoint of a chamber clean operation) based on the chamber value data. Further details of the process control tool 428 are discussed with reference to process 600D in FIG. 6D.

[0074] The process treatment updates can change the way the process treatment 406 is performed by the process tool 404. For example, the updates can include increasing or decreasing the process time of one or more process operations of the process treatment 406. For example, the updates can increase or decrease the duration of a chamber clean operation. In some embodiments, the chamber matching tool 422, the sensor matching tool 424, the strategy development tool 426, and the process control tool 428 can each use a model (e.g., a model of the modeling system 470) that receives input data, such as chamber data from the process tool 404 and / or processed data output from the data preparation tool 416, and outputs one or more updates to the process treatment 406. As discussed later, the model can be a machine learning model and can include a bootstrap aggregation model, a random forest tree decision tree model, a partial least squares regression (PLS) model, a least absolute shrinkage and selection operator (LASSO) regression model, and / or a ridge regression model, among other models. The models can include ensemble modeling, which includes multiple models (e.g., stacked modeling) to utilize a more reliable model for the final classification of the received data. Additionally, the models can include linear models, dynamic models, digital models, mechanistic models, statistical models, physics-based models, digital twin models, etc.

[0075] As mentioned above, in some embodiments, each of the chamber matching tool 422, the sensor matching tool 424, the strategy development tool 426, and / or the process control tool 428 can use machine learning models to perform the described methods. In some embodiments, each of the chamber matching tool 422, the sensor matching tool 424, the strategy development tool 426, and / or the process control tool 428 can use traditional models and / or physics-based models (e.g., digital twins) to perform the described methods. The modeling system 470 can be used to generate (e.g., train, validate, and / or test) the associated models. The following exemplary description of the modeling system 470 is described in the context of using the modeling system 470 to generate a model 490 associated with the chamber matching tool 422. However, it should be noted that this description is purely exemplary. As will be further discussed in connection with other embodiments, the generation and execution of models associated with each of the chamber matching tool 422, the sensor matching tool 424, the strategy development tool 426, and / or the process control tool 428 can use similar processing hierarchies and methods, individually and / or in combination with each other.

[0076] The modeling system 470 can include one or more computing devices, such as a rack mount server, a router computer, a server computer, a personal computer, a mainframe computer, a laptop computer, a tablet computer, a desktop computer, etc. The modeling system 470 can include a chamber value component 494. In some embodiments, the chamber value component 494 can use the historical data 442 to determine updates (e.g., updates to one or more process parameters of a process action (e.g., process action 406)) based on the chamber value data. In some embodiments, the chamber value component 494 can use a model 490 to determine the updates. In certain embodiments, the model 490 is a trained machine learning model. The model 490 can use the historical data to determine the updates.

[0077] In some embodiments, the model 490 is stored on the server machine 492 and / or the server 420. The server machine 480 and / or the server machine 492 can store one or more recently trained machine learning models (e.g., for performing model execution management). For example, one or more models trained within a duration (e.g., the last six months) can be stored in the server machine 492, and one or more final models for predicting chamber conditions can be stored on the server 420 (e.g., operative with the chamber matching tool 422).

[0078] In some embodiments, modeling system 470 further includes server machine 472 and server machine 480. Server machines 472 and 480 can be one or more computing devices (such as a rack mount server, a router computer, a server computer, a personal computer, a mainframe computer, a laptop computer, a tablet computer, a desktop computer, etc.), a data store (e.g., a hard disk, a memory database), a network, a software component, or a hardware component.

[0079] The server machine 472 can include a dataset generator 474 capable of generating datasets (e.g., a set of data inputs and a set of target outputs) for training, validating, or testing a model. The dataset generator 474 can divide the historical data 442 into a training set (e.g., 60 percent of the historical data, or any other percentage of the historical data), a validation set (e.g., 20 percent of the historical data, or some other percentage of the historical data), and a test set (e.g., 20 percent of the historical data). In some embodiments, the chamber value component 494 generates multiple sets of training data. For example, the one or more sets of training data can include each of the datasets (e.g., a training set, a validation set, and a test set).

[0080] The server machine 480 includes a training engine 482, a validation engine 484, and a testing engine 486. The training engine 482 can train a model 490 using one or more of the historical sensor data 444, the historical chamber value data 446, and / or the historical chamber update data 447 of the historical data 442 (of the data store 440). In some embodiments, the model 490 can be trained using one or more outputs of the data preparation tools 416, the chamber matching tools 422, the sensor matching tools 424, the strategy development tools 426, and / or the process control tools 428. For example, the model 490 can be a hybrid machine learning model using sensor data and / or mechanistic features, such as feature extraction, mechanistic modeling, and / or statistical modeling (e.g., using the data preparation tools 416). The training engine 482 can generate multiple trained models 490, with each trained model 490 corresponding to a separate set of features for each training set.

[0081] The validation engine 484 can determine the accuracy of each of the trained models 490 based on the corresponding set of features of each training set. The validation engine 484 can discard trained models 490 that have an accuracy that does not meet a threshold accuracy. The testing engine 486 can determine the trained model 490 that has the highest accuracy of all trained models based on the testing (and optionally validation) set.

[0082] Model 490 may refer to a model created by training engine 482 using a training set including data inputs and corresponding target outputs (historical results of the process chamber under parameters associated with the target inputs). Patterns may be found in the data set that map the data inputs to the target outputs (e.g., identifying connections between portions of sensor data and / or chamber value data and chamber process updates), and model 490 is provided with mappings that capture these patterns. Model 490 may use one or more of logistic regression, syntactic analysis, decision trees, or support vector machines (SVM). In some embodiments, the machine learning may consist of single-level linear or nonlinear operations (e.g., SVM) and / or may be or include a neural network.

[0083] The chamber value component 494 can provide current data (e.g., current sensor data indicating a thickness of material on an inner surface of a processing chamber) as input to the model 490, which can execute on the input to obtain one or more outputs including a set of values ​​indicative of updates to one or more process operations 406.

[0084] For purposes of illustration and not limitation, aspects of the disclosure describe training and using models using information about historical data 442. In other embodiments, heuristic or rule-based models may be used to determine process operation updates.

[0085] In some embodiments, the functionality of client device 450, server 420, data store 440, and modeling system 470 may be provided by fewer machines than those shown in Figure 4. For example, in some embodiments, server machines 472 and 480 may be combined into a single machine, and in some other embodiments, server machines 472, 480, and 492 may be combined into a single machine. In some embodiments, modeling system 470 may be provided in whole or in part by server 420.

[0086] In general, functions described in one embodiment as being performed by client device 450, data store 440, metrology system 410, manufacturing system 402, and modeling system 470 may also be performed on server 420 in other embodiments, where appropriate. In addition, functions attributed to particular components may be performed jointly by different or multiple component operations.

[0087] In some embodiments, a "user" may be represented as a single individual. However, other embodiments of the present disclosure encompass "users" that are entities and / or automated sources managed by multiple users. For example, a set of individual users grouped together as a group of administrators may be considered a "user."

[0088] 5A-5E are flow diagrams of methods 500A-E associated with using an optical sensor for a processing chamber. FIG. 5A is a flow diagram of a method 500A for determining a process operation update based on a chamber wall value, according to some embodiments. Methods 500A-E can be performed by processing logic, which can include hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (e.g., instructions executed on a processing device, a general-purpose computer system, or a dedicated machine), firmware, microcode, or combinations thereof. In some embodiments, methods 500A-E can be performed in part by modeling system 470 of FIG. 4 (e.g., server machine 472, server machine 480, and server machine 492 of FIG. 4). Modeling system 470 can use method 500A to generate a data set for at least one of training, validating, or testing a machine learning model, according to embodiments of the present disclosure. Methods 500A-E can be performed by server machine 480 (e.g., training engine 482, etc.). Methods 500A-E may be performed by server machine 492 (e.g., chamber value component 494). In some embodiments, methods 500A-E may be performed by server 420 (e.g., one or more of chamber matching tool 422, sensor matching tool 424, strategy development tool 426, and / or process control tool 428). In some embodiments, a non-transitory storage medium stores instructions that, when executed by a processing device (e.g., modeling system 470, server machine 480, server machine 492, server 420, etc.), cause the processing device to perform methods 500A-E.

[0089] For ease of explanation, methods 500A-E are shown and described as a series of operations. However, operations in accordance with the present disclosure may occur in various orders and / or simultaneously, with other operations not shown and described herein. Moreover, not all illustrated operations may be performed to implement methods 500A-E in accordance with the disclosed subject matter. In addition, those skilled in the art will understand and appreciate that methods 500A-E may alternatively be represented as a series of interrelated states via a state diagram or events.

[0090] Referring to FIG. 5A, at block 502 of method 500A, the processing device may receive first sensor data indicative of a condition of a wall corresponding to a first processing chamber. The wall may be an interior wall of the first processing chamber. In some embodiments, the wall is a liner of the first processing chamber. In some embodiments, the first sensor data includes optical spectrum data indicative of a condition of the wall of the first processing chamber. The optical spectrum data may be from an OWPS. The optical spectrum data may include thin film interference data indicative of a thickness and / or other characteristics of a thin film deposited on the first processing chamber wall. In some embodiments, the sensor data is indicative of a thickness of a process film layer deposited on the interior wall (e.g., a liner) of the processing chamber. The process film layer may be a deposition by-product of a first process operation (e.g., a by-product of a deposition operation, an etch operation, etc.). The sensor data may be indicative of a condition (e.g., a cleanliness condition) inside the processing chamber. For example, the sensor data indicative of a thickness of a process film layer on the first processing chamber wall may also be indicative of a thickness of a process film layer deposited on one or more other interior surfaces of the first processing chamber.

[0091] At block 504, the processing device can determine a first value based on the first sensor data. The first value can be a dimensionless value indicative of a thickness of the process film layer. In some embodiments, the first value is a first index indicative of a thickness of the process film layer deposited on the inner wall of the first processing chamber at a first time. In some embodiments, the first value is indicative of a thickness of the process film layer. For example, the first value can correspond to a thickness of the process film layer deposited on the first processing chamber wall prior to a first process operation. In some embodiments, the first value can correspond to a first amount of product (e.g., process film layer) deposited along the surface of the wall. In some embodiments, the first value indicates a trend of the thickness of the process film layer deposited on one or more inner surfaces of the first processing chamber. As an example, a larger first value can indicate a thicker process film layer and a smaller first value can indicate a thinner process film layer. The first value can further indicate a condition of the first processing chamber.

[0092] At block 506, the processing device may determine a first update of a first process operation associated with the first processing chamber based on the first value. In some embodiments, the first process operation is a process action (e.g., process action 406 of FIG. 4). In some embodiments, the first update may be an update of a duration of one or more of a cleaning recipe, a deposition recipe, or an etch recipe, etc. In some embodiments, the first update may be an update of a deposition rate or an etch rate associated with a deposition recipe or an etch recipe, respectively. In some embodiments, the first update may be an update of a process knob of the first process operation. In some embodiments, the first update is made to more closely match a target process action (e.g., of process action 406 of FIG. 4). In some embodiments, the first update is a maintenance action (e.g., a preventive maintenance action, a corrective maintenance action, etc.). In some embodiments, the first update is performed by a user (e.g., replacing a chamber part).

[0093] In some embodiments, the optical sensor data can be used to track the condition of the chamber. In some embodiments, an index indicative of the condition of the coating (e.g., CWI) can be indicative of the condition of the chamber. In some embodiments, an index, e.g., a chamber condition index, can be generated from data including the optical sensor data. The evolution of the index over time can indicate component variations or aging, variations in chamber performance, etc.

[0094] In some embodiments, receiving and storing optical data (and / or data based on the optical data) of a group of processing chambers can enable population-specific outlier detection, tool matching, consistency assessment, etc. In some embodiments, the target processing conditions can be the same across the group of chambers. The optical data can be used to homogenize processing conditions within the separate chambers. In some embodiments, the optical data from the group of chambers can be provided to a trained machine learning model configured to detect outliers. Such a model can have common characteristics to machine learning models configured to detect anomalies, as described above. In some embodiments, detecting an outlier chamber can include comparing the optical data received from the chamber to historical optical data. In some embodiments, an outlier chamber can generate optical data consistent with other chambers, but other data can be different (e.g., a cleaning procedure can be run longer to achieve the same optical results). Corrective action can be initiated, e.g., schedule maintenance, provide a warning to a user, etc.

[0095] In some embodiments, multivariate methods can be used, e.g., for anomaly detection, flagging outliers, etc. In some embodiments, the spectral data can be combined with other data (e.g., other sensor data, such as pressure data, temperature data, and / or wafer data, such as in-line or in-situ metrology data). In some embodiments, the spectral data, one or more indices based on the spectral data, metrology data, and / or sensor data can be fed to a machine learning model. The machine learning model can be configured to receive as input data indicative of process conditions, process quality, chamber performance, etc., and generate as output an indication of whether the chamber is performing similarly to other chambers in the population (e.g., not performing abnormally). In some embodiments, the data based on the spectral data can include operations performed on the spectral data, e.g., derivatives of indices based on the spectral data, slopes of the spectrum at particular wavelengths, averages or medians, etc.

[0096] At block 508, the processing device may prepare a notification indicating the first update for presentation on the GUI. In some embodiments, the notification is prepared for viewing by a user (e.g., a technician, operator, engineer, etc.) via the GUI. At block 508, the processing device may cause a first process operation to be performed according to the first update. For example, the processing device may cause the first process operation to be performed according to the first update. As a specific example, the processing device may cause a cleaning operation to be performed by a processing chamber according to the updated duration.

[0097] Optionally, at block 510, the processing device may determine an update of at least one of a sensor gain value or a sensor offset value corresponding to the OWPS based on the first value. The processing device may determine a calibration value corresponding to the OWPS based on the first value. For example, the processing device may determine that the sensor data from the OWPS has fluctuated over time based on the first value and may determine an update of the OWPS gain value and / or offset value based on the first value. In some embodiments, the processing device may determine an update of at least one of a sensor gain value or a sensor offset value corresponding to one or more sensors of the first processing chamber based on the first value. The processing device may determine one or more calibration values ​​for one or more sensors of the first processing chamber based on the first value.

[0098] In some embodiments, a baselining procedure may be performed to identify and / or correct for light sensor variations. The baselining procedure may be performed periodically, e.g., separated by a fixed amount of time, a fixed number of processed substrates, etc. In some embodiments, the baselining procedure may be performed at a target frequency. In some embodiments, characteristics of the light sensor arrangement may change over time. For example, the transparent window of the OWPS sensor may erode, undergo compositional changes, etc. Baselining may be performed to compensate for changes in the behavior of the sensor assembly. Baselining may be performed by generating spectral data in a clean, idle chamber, e.g., when no processing operations are occurring. The baseline spectrum may be used to zero out spectra collected by the OWPS system during processing of substrates (e.g., by subtracting the baseline spectrum from the collected spectra). The frequency at which the baselining operation is performed may be adjusted based on the intended use of the chamber, the quality or performance of the chamber, the service life of the chamber (e.g., time since chamber installation, run time since chamber installation, etc.), etc. In some embodiments, the baseline measurement may be used to determine a baseline response of the light sensor (e.g., response in the absence of process material). The baseline data can be used to correct the data received from the light sensor.

[0099] Optionally, at block 512, the processing device can determine a second value based on the first sensor data. The second value can correspond to a second amount of product disposed along the surface of the first processing chamber wall at a second time. For example, the second value can correspond to a thickness of a process film layer deposited on the first processing chamber wall after the first process operation. In a more specific example, the second value can correspond to a thickness of a process film layer after a cleaning operation. A difference between the first value and the second value can correspond to a net change in the process film layer thickness during the first process operation.

[0100] Optionally, at block 514, the processing device may determine a process rate associated with the first process operation. The processing device may determine the process rate using the first value and the second value. The processing device may determine the process rate using a difference between the first value and the second value divided by the duration of the first process operation. The first update determined by the processing device may be further determined based on the process rate. As an example, the cleaning operation may be updated based on a cleaning rate (e.g., process rate) determined based on a difference between the thickness of the initial process film layer and the thickness of the final process film layer divided by the duration of the cleaning operation. The first update may correspond to increasing or decreasing the duration of the cleaning operation based on the cleaning rate.

[0101] 5B is a flow diagram of a method 500B of using light sensor data in initiating execution of a corrective action, according to some embodiments. At block 520, processing logic receives first data from a light sensor of a processing chamber. The light sensor may be an OWPS sensor. The first data may include spectral data. The first data may be indicative of a coating on a surface of the OWPS sensor assembly (e.g., a transparent window of the OWPS sensor assembly). In some embodiments, the first data may be generated by the light sensor during a processing operation. In some embodiments, the first data may be generated between processing operations.

[0102] At block 522, the processing logic processes the first data to obtain second data. The second data includes an indication of a condition of the coating on the interior surface of the processing chamber. In some embodiments, the condition of the coating includes a thickness of the coating. In some embodiments, the second data includes a dimensionless parameter related to the condition of the coating, e.g., a CWI index. In some embodiments, the first data can be processed to compensate for optical effects, e.g., baseline drift, in-chamber plasma effects, etc. In some embodiments, a baseline response of the optical sensor assembly (e.g., OWPS) can be obtained. In some embodiments, baselining can include collecting optical data when the processing chamber is in an idle mode. In some embodiments, the effects of a radioactive plasma can be compensated for. In some embodiments, the optical sensor data can be collected while a light source coupled to the optical sensor is turned off and a plasma is generated in the processing chamber. In some embodiments, the plasma signal and the baseline signal can be used to correct the data received from the optical sensor, e.g., by baseline subtraction.

[0103] In some embodiments, data can be generated when the process chamber is at a reference condition, e.g., under conditions different from normal processing conditions. Sensor data can be collected when the chamber is at the reference condition to determine a baseline sensor response. The baseline sensor response can be utilized to correct or update the first and / or second data.

[0104] At block 524, the processing logic generates a performance indication of a processing operation of the processing chamber given the second data. In some embodiments, the performance indication can be based on a value of a dimensionless index, e.g., a CWI index. In some embodiments, the processing operation can be an operation that deposits or removes material from a coating on a surface in the processing chamber, e.g., an etching operation, an atomic deposition operation, etc. In some embodiments, the processing operation can be a chamber clean operation. In some embodiments, generating a performance indication of the processing operation can include providing the data (e.g., the spectral data, the CWI data, etc.) to a trained machine learning model. The machine learning model can be trained using, e.g., the historical light sensor data as training inputs. The machine learning model can be trained using supervised learning (e.g., provided with labeled training data, e.g., training data indicative of quality classification, anomaly classification, historical anomaly label data, etc.). The machine learning model can be trained using unsupervised learning (e.g., configured to perform clustering, anomaly detection, etc.).

[0105] At block 526, processing logic causes corrective action to be taken in view of the performance indication of the processing chamber. In some embodiments, the corrective action may include one or more of providing an alert to a user (e.g., on a graphical user interface), updating a processing operation recipe (e.g., updating one or more steps or operations of an etch recipe, a clean recipe, or a deposition recipe), scheduling preventive maintenance, scheduling corrective maintenance, etc. In some embodiments, updating the recipe may include updating an end point of a processing operation, e.g., extending the operation if time is determined to be insufficient, shortening the operation to increase chamber throughput, etc.

[0106] FIG. 5C is a flow diagram of a method 500C of maintaining the integrity of a processing chamber, according to some embodiments. At block 530, processing logic receives first data from an optical sensor of the processing chamber. The first data is associated with a first operation of the processing chamber. In some embodiments, the first operation is an operation that changes a condition of a coating on an inner surface of the chamber, e.g., a deposition or etching operation. In some embodiments, the operation is a cleaning operation, e.g., an in situ chamber cleaning operation. In some embodiments, the processing logic receives a plurality of historical data, e.g., the first data is one of many optical data associated with many operations. In some embodiments, the first data includes spectral data. In some embodiments, the first data includes an index, e.g., a CWI. In some embodiments, the first data indicates the evolution of the spectral data throughout the processing operation, e.g., a time trace of CWI values, a time trace of spectral values, etc.

[0107] At block 532, the processing logic receives second data from the optical sensor of the processing chamber. The second data is associated with a second operation of the processing chamber. In some embodiments, the second data is associated with a processing operation corresponding to the processing operation of the first data, such as the same operation of the same recipe performed in a separate run on a separate wafer, etc. The second data can include similar characteristics to the first data, such as spectral data, the evolution of the data over time, etc.

[0108] At block 534, the processing logic generates first and second performance indications of the processing chamber based on the first and second data. The first performance indication can be associated with performance of the first processing operation. The second performance indication can be associated with performance of the second processing operation. In some embodiments, the performance indications can be generated by a machine learning model, e.g., the second data and / or the first data can be provided to a trained machine learning model that generates the performance indications. In some embodiments, the machine learning model can be configured to detect outliers, anomalies, etc. based on the optical sensor data.

[0109] At block 536, the processing logic determines one or more performance deltas between the first and second operations. In some embodiments, the performance deltas can indicate a variation in performance over time. In some embodiments, multiple historical performance data are considered. In some embodiments, the evolution of performance over time can indicate a trend in performance, for example, over many historical examples. In some embodiments, multivariate analysis can be performed. The optical sensor data can be used along with other sensor data, manufacturing parameters, hardware data, metrology data, etc. to determine the performance of the processing operation, the performance of the chamber, the performance variation over time, the performance variation in chamber activity time, etc. At block 538, the processing logic causes a corrective action to be taken based on the one or more deltas in the performance data. The corrective action can include updating the processing recipe, such as updating an endpoint condition (e.g., endpoint time, target endpoint sensor reading, etc.) of the processing recipe. The corrective action can include scheduling maintenance, e.g., scheduling replacement of a component, scheduling preventive maintenance, scheduling further investigation of the subsystem, etc. The corrective action can include providing an alert to a user, for example, via a user interface.

[0110] 5D is a flow diagram of a method 500D for detecting and correcting anomalies and / or inconsistencies in a group of processing chambers, according to some embodiments. At block 540, processing logic receives first data from an optical sensor of a first processing chamber. The first data can include spectral data. The first data can include processed data, e.g., an index, such as a CWI index. The first data can indicate a characteristic of a coating on an inner surface of the processing chamber, e.g., a thickness of the coating, a composition of the coating, etc. At block 542, processing logic receives second data from an optical sensor of a second processing chamber. The second data can share characteristics with the first data.

[0111] At block 544, the processing logic generates a first performance indication for the first chamber based on the first data and a second performance indication for the second chamber based on the second data. These performance indications may include an index indicative of the state of the chamber, e.g., a dimensionless index such as the CWI. The performance indications may incorporate additional data, e.g., the generation may include multivariate analysis. In some embodiments, data indicative of the optical condition of the chamber (e.g., OWPS sensor data, data indicative of the thickness and / or composition of the deposited film in the processing chamber, etc.) may be provided to the trained machine learning model to generate the performance indication. In some embodiments, other sensor data, metrology data, process parameter data, etc. may also be provided to the machine learning model. The trained machine learning model may be configured to determine a performance differential, e.g., detect outliers, anomalies, etc. In some embodiments, the performance of the processing chamber may be compared to the performance of multiple chambers, e.g., identify abnormal performance in a group of chambers. The performance of the processing chamber may be compared to historical data of multiple processing chambers.

[0112] At block 546, processing logic determines one or more performance deltas (e.g., quality of performance) between the first and second processing chambers. In some embodiments, the operations of block 546 may be accomplished by providing data to a trained machine learning model. In some embodiments, the operations of block 546 may be performed by comparing data from one processing chamber to an average, median, range, etc., of multiple other chambers (e.g., the multiple including the first processing chamber).

[0113] At block 548, the processing logic causes a corrective action to be performed based on one or more differences between the performance of the first processing chamber and the performance of the second processing chamber. In some embodiments, the corrective action may be selected to modify the performance of the second processing chamber. In some embodiments, the corrective action may modify the performance of the second processing chamber such that the modified performance of the second processing chamber is more similar to the performance of the first processing chamber than the unmodified performance of the second processing chamber. In some embodiments, the performance of the second chamber may be matched to the first chamber. The corrective action may include updating a process recipe. In some embodiments, a process recipe associated with the optical sensor data may be updated. For example, a process time, a process temperature or pressure, a gas flow, etc. may be adjusted to more closely match the performance of the second chamber to the performance of the first chamber. In some embodiments, a process other than the process associated with the optical sensor data may be updated. For example, a cleaning treatment recipe may be updated, and updating the cleaning treatment may affect the performance of the processing recipe (e.g., substrate processing treatment) of the second chamber.

[0114] FIG. 5E is a flow diagram of a method 500E of utilizing optical sensor data collected over time, according to some embodiments. At block 550, processing logic receives first sensor data. The first sensor data includes optical data, e.g., data from an OWPS. The first data is indicative of a condition of a wall of a first processing chamber. At block 552, processing logic receives second sensor data. The second sensor data can have similar characteristics as the first sensor data. The operations of blocks 550 and 552 can share characteristics with the operations of block 502 of FIG. 5A.

[0115] At block 554, the processing logic compares the first sensor data to the second sensor data. In some embodiments, the first sensor data can include historical sensor data. In some embodiments, the historical data can be considered to determine corrective actions. Historical optical sensor data (e.g., spectral data generated by the OWPS) of many process runs can be recorded and stored (e.g., as the first sensor data). In some embodiments, historical data collected from the same process chamber, process runs performing the same recipe, process runs sharing a target end product, etc. can be utilized. The second sensor data (e.g., a current dimensionless value derived from the OWPS sensor data) can be compared to the historical optical sensor data. In some embodiments, statistical analysis can be utilized for anomaly detection, for example, to determine whether a dimensionless value associated with the OWPS sensor data is outside a threshold range, such as three standard deviations from the mean value of the historical data. In some embodiments, the historical optical data (e.g., spectral data, parameters generated based on the spectral data, etc.) can be used to train a machine learning model. The machine learning model can be configured to accept as input current data (e.g., spectral data indicative of the condition of the lining deposited on the chamber walls) and generate as output an indication of whether a process run associated with the input data is likely to be anomalous (e.g., resulting in a product that does not meet a performance threshold). In some embodiments, the machine learning model can be trained via unsupervised learning (e.g., generated to determine whether the input data is similar to historical data without explicit labeling). In some embodiments, the machine learning model can be trained using supervised learning (e.g., trained with data labeled with classification labels, such as labeled data associated with acceptable products, labeled data with unique anomaly causes for anomaly cause classification, etc.).

[0116] In some embodiments, data associated with the optical sensor can be monitored over time (e.g., over many processed substrates, long chamber operation times, long radio frequency times, etc.). In some embodiments, the optical data can be utilized to maintain run-to-run consistency, e.g., to adjust process parameters to compensate for process equipment component variations, aging, etc. Run-to-run consistency operations can share features with anomaly detection operations, e.g., as described above in connection with block 506 of FIG. 5A. Run-to-run consistency operations can include statistical calculations, providing current sensor data to machine learning models, etc. In some embodiments, chamber conditions can be tracked over time, and process operations (e.g., process recipes) can be updated to maintain chamber consistency. In some embodiments, tracking changes in process conditions, changes in recipes (e.g., changes made to maintain consistent process conditions), etc. can enable understanding of chamber performance over time, upcoming maintenance to be performed, etc. In some embodiments, monitoring data from the chamber over time can include monitoring spectral data and / or monitoring index data, e.g., values ​​designed to represent quality of a coating, chamber, etc., based on the spectral data. In some embodiments, evaluating performance over time can include evaluating data generated at the end of the process, e.g., wall index values ​​(CWI values) related to the spectral data at the end of the process. In some embodiments, evaluating performance over time can include evaluating data generated throughout the process, e.g., the evolution of wall index values ​​throughout the process. In some embodiments, the OWPS data can be utilized to determine whether an update to the process recipe had the intended effect, e.g., whether the update corrected a processing disturbance. In some embodiments, one or more other values ​​can be tracked over time, e.g., derivatives or gradients of the spectral data or index values, higher order derivatives, averages over time (e.g., in a processing operation), etc.

[0117] In some embodiments, data from several processing operations (e.g., processing runs performed over a period of time) can be analyzed. In some embodiments, data can be tracked over time (e.g., an index such as CWI can be tracked across multiple processing runs). Process performance can be correlated to the tracked optical data, e.g., etch rate can be correlated to CWI values. Based on the tracked optical data, variations can be detected within a chamber, within a process, within a population, etc. Corrective action can be taken in light of the variations detected over time. For example, OWPS data can be generated during a cleaning procedure. OWPS data collected at the end of a cleaning procedure can be tracked over time (e.g., across multiple cleaning procedure runs). Variations in the OWPS data can be detected (e.g., by calculating the slope of an index associated with the depicted OWPS data over time, calculating the difference between index values ​​over time, etc.). Variation detection can lead to process adjustments, e.g., cleaning procedure endpoints can be adjusted. In some embodiments, a procedure can be adjusted to provide improved results, e.g., cleaning procedures can be extended for more thorough cleaning. In some embodiments, a procedure can be adjusted to provide improved efficiency, e.g., cleaning procedures can be shortened for increased throughput, reduced costs, etc.

[0118] 6A-6D are flow diagrams of processes 600A-D associated with determining process operation updates based on chamber wall values, according to some embodiments. Processes 600A-D may be performed by processing logic, which may include hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (e.g., instructions executed on a processing device, general-purpose computer system, or dedicated machine), firmware, microcode, or combinations thereof. In some embodiments, processes 600A-D may be performed in part by modeling system 470 of FIG. 4 (e.g., server machine 472, server machine 480, and server machine 492 of FIG. 4). Processes 600A-D may be performed by server machine 480 (e.g., training engine 482, etc.). Processes 600A-D may be performed by server machine 492 (e.g., chamber value component 494). In some embodiments, processes 600A-D may be performed by server 420 (e.g., one or more of chamber matching tool 422, sensor matching tool 424, recipe development tool 426, and / or process control tool 428). In some embodiments, a non-transitory storage medium stores instructions that, when executed by a processing device (e.g., modeling system 470, server machine 480, server machine 492, server 420, etc.), cause the processing device to perform processes 600A-D.

[0119] Referring to FIG. 6A, a process 600A for updating a chamber cleaning recipe is shown. The process 600A may be performed by the recipe development tool 426 of FIG. 4. In some embodiments, at block 602, a processing device (e.g., of the recipe development tool 426) may receive OWPS data. The OWPS data may be indicative of one or more optical thin film properties of a process film layer deposited on an inner surface of a process chamber wall (e.g., on a process chamber liner). The OWPS data may be indicative of a thickness of a process film layer deposited on an inner wall (e.g., a liner) of the process chamber. At block 604, the processing device may determine a first characteristic of the film deposited on the wall. In some embodiments, the processing device may determine a thickness of the film. In some embodiments, the processing device may determine a chemical composition of the film. In some embodiments, the processing device may determine a chemical arrangement or packing geometry of the film (e.g., a crystal structure, a molecular or atomic packing structure, etc.). In some embodiments, the processing device may determine another characteristic of the film. The processing device may determine a characteristic of the process film layer prior to a process operation. The processing device can determine a first dimensionless value indicative of a process film layer characteristic before a process operation, hi some embodiments, the first dimensionless value can be determined before a chamber clean operation.

[0120] The processing device may begin a processing operation at block 608. In some embodiments, the processing operation that is begun may include a cleaning process, an etching process, a deposition process, or the like.

[0121] After the processing operation, in block 610, the processing device can determine a second characteristic of the film on the processing chamber wall, such as a characteristic of the film at a second time. The processing device can determine a second dimensionless value indicative of the process film layer characteristic after the processing operation. The processing device can determine the second value based on sensor data received from the OWPS after the processing operation.

[0122] At block 612, the processing device may determine whether a threshold process layer characteristic has been met. In some embodiments, the threshold process layer characteristic includes a critical wall thickness. In some embodiments, the threshold process layer characteristic includes a target chemical composition. The processing device may make the determination based on a comparison of the first dimensionless value to the second dimensionless value. As an example, the processing device may determine whether the processing chamber has been cleaned such that the thickness of the process film layer on the processing chamber wall is below a predetermined value (e.g., a critical wall thickness). If the threshold thickness has not been met, the processing device may cause an additional chamber clean (block 608). If the threshold thickness has been met, the processing device may update the cleaning recipe (e.g., based on any additional cleaning time, etc.) at block 614. After the cleaning process has been updated, the processing device may release the processing chamber to a production mode. Although the operation of process 600A has been described with reference to a chamber cleaning operation, one skilled in the art should understand that process 600A may be applied to any chamber processing operation, such as a deposition process, an etch process, etc.

[0123] Referring to FIG. 6B, a process 600B is shown for calibrating a sensor in a processing chamber. The process 600B may be performed by the sensor matching tool 424 of FIG. 4. The process 600B may be performed to compensate for sensor variations over time. In some embodiments, at block 602, a processing device (e.g., of the sensor matching tool 424 of FIG. 4) may receive OWPS data. The OWPS data may be indicative of one or more optical thin film properties of a process film layer deposited on an inner surface of a processing chamber wall (e.g., on a processing chamber liner). The OWPS data may be indicative of a thickness of the process film layer, a chemical composition of the process film layer, and the like. At block 624, the processing device may determine a characteristic of the film deposited on the processing chamber wall. The characteristic may correspond to an amount of material deposited on the surface, a thickness of the film on the surface, a chemical composition of the film, or another characteristic of the film. In some embodiments, the thickness of the film may be determined from the data indicative of the amount of material deposited proximate to the sensor. The processing device may determine a dimensionless value indicative of the process film layer characteristic at block 626. The processing device can use this dimensionless value to calibrate one or more sensors associated with the processing chamber at block 628. The one or more sensors can include the OWPS of the processing chamber, but can also include other sensors of the processing chamber, sensors associated with one or more other process chambers, etc. Calibration of the sensors can include baseline subtraction, percent response, or another functional adjustment, parameter-dependent adjustment (e.g., wavelength-dependent adjustment), etc. Calibration of the sensors by the processing device can be in response to sensor and / or process variations caused by the buildup of a process film layer on one or more interior surfaces of the processing chamber.

[0124] Referring to FIG. 6C, a process 600C for updating a chamber process recipe is shown. The process 600C may be performed by the chamber matching tool 422 of FIG. 4. In some embodiments, at block 630, a processing device (e.g., of the chamber matching tool 422 of FIG. 4) may receive first OWPS data corresponding to a first processing chamber. At block 632, the processing device may determine a first dimensionless value based on the first OWPS data. The first dimensionless value may be indicative of a process film layer characteristic of a film deposited on an inner wall (or other inner surface) of the first processing chamber. At block 634, the processing device may receive second OWPS data corresponding to a second processing chamber. The processing device may determine a second dimensionless value based on the second OWPS data at block 636. The second dimensionless value may be indicative of a process film layer characteristic of a film deposited on an inner wall of the second processing chamber. In some embodiments, the processing device may normalize the first dimensionless value and the second dimensionless value by a data normalization operation. The first and second dimensionless values ​​can be normalized based on process variations caused by an accumulation of process material (e.g., an increase in thickness of a process film layer) in the first and / or second processing chambers.

[0125] At block 638, the processing device can update a process recipe associated with the first processing chamber (e.g., process action 406 of FIG. 4) based on the first dimensionless value and the second dimensionless value. In some embodiments, the processing device can update the process recipe based on the normalized first dimensionless value and / or the normalized second dimensionless value. Updating the process recipe associated with the first processing chamber based on the first and second dimensionless values ​​can match or align the first processing chamber to the second processing chamber. Updating the process recipe associated with the first processing chamber based on the first and second dimensionless values ​​can be included in a chamber matching action for matching the first and second chambers. In some embodiments, the process recipe set points, e.g., temperature, pressure, gas flow, etc., can be adjusted. In some embodiments, the process recipe operation duration can be adjusted, e.g., an end point or transition point can be adjusted.

[0126] Referring to Figure 6D, a process 600D for updating a chamber process recipe endpoint is shown. The process 600D may be performed by the process control tool 428 of Figure 4. In some embodiments, at block 602, a processing device (e.g., of the process control tool 428 of Figure 4) may receive OWPS data indicative of one or more optical thin film properties of a process film layer deposited on an inner surface of a processing chamber wall (e.g., on a processing chamber liner). The OWPS data may indicate a thickness of the process film layer, a chemical composition of the process film layer, a structure or packing arrangement of the process film layer, etc.

[0127] At block 624, the processing device may determine a characteristic of the film deposited on the processing chamber wall. This determination may be based on the OWPS data. In some embodiments, the processing device may determine a dimensionless value indicative of a thickness of the process film layer at block 626. The dimensionless value may be a ratio, a percentage, etc. of a target value of one or more characteristics of the thin film. In some embodiments, the dimensionless value may indicate a percentage thickness of an acceptable process film layer thickness (e.g., a critical wall thickness). The processing device may update a process operation recipe taking into account the dimensionless value. At block 648, the processing device may update a process recipe endpoint (e.g., a cleaning recipe endpoint, a deposition recipe endpoint, an etch recipe endpoint, etc.), one or more recipe processing parameters (e.g., temperature, pressure, gas flow, etc.), one or more hardware parameters (e.g., internal component placement) based on the dimensionless value. The process update of the process recipe may be correlated to a target process recipe endpoint condition. For example, the process recipe update can correlate to a process recipe that terminates when a process film layer thickness reaches (e.g., above or below) a predetermined value (e.g., a critical wall thickness). In some embodiments, the process recipe update terminates the process recipe before the process film layer characteristic reaches a critical value (e.g., before the film layer reaches a critical wall thickness). In some embodiments, the process recipe update terminates the process recipe after the process film layer characteristic falls below a critical value.

[0128] FIG. 7 is a flow diagram of a method 700 for improving processing in a processing chamber using one or more optical film properties of a process film layer on a wall of a process chamber (e.g., on a transparent thin film (e.g., transparent thin film 307 of FIG. 3)) located above a transparent window (e.g., transparent window 120 of FIG. 1) in accordance with various aspects of the present disclosure. The method 700 can be performed by processing logic, which can include hardware (circuitry, dedicated logic, etc.), software (such as running on a general purpose computer system or dedicated machine), firmware, or some combination thereof. For example, the method 700 can be performed by the controller 109 (e.g., processing device 130), as these components are referred to herein. In some embodiments, various baselines or thresholds (such as thickness or variations in n and k) can be stored and indexed for a particular configuration in the memory 134 of the controller 109. These values ​​can be used by comparison by various methods 700 discussed herein. Although shown in a particular sequence or order, unless otherwise specified, the order of the processes can be modified. Therefore, the illustrated embodiments should be understood as examples only, and the illustrated processes may be performed in a different order, and some processes may be performed in parallel. In addition, in various embodiments, one or more processes may be omitted. Thus, not all processes are required in all embodiments. Other process flows are possible.

[0129] 7, the method 700 may begin at block 710 where processing logic processes the reflected measurement signal to fit the reflected measurement signal to a thin film optical model to determine one or more film properties or characteristics (e.g., thickness, n and k values, material composition, etc.) of the process film layer 309 deposited on the transparent thin film 307 (see FIG. 3). When a measurement is made during processing of a substrate in a processing chamber, the method 700 may proceed to block 715 where processing logic determines whether the property or characteristic, such as thickness, matches a baseline measurement for the process film layer, e.g., taking into account the composition of the process film layer 309. The processing logic may determine a dimensionless value indicative of a characteristic (e.g., thickness) of the process film layer and match the characteristic to a baseline, standard, or golden run measurement (e.g., a measurement associated with a processing run deemed to be of sufficiently high quality, performance, etc.). If a match exists, e.g., one or more property or characteristic values ​​of the process film layer meet a threshold, then the method 700 may proceed to block 705, where the substrate is processed in the processing chamber. If a match does not exist, then the method 700 may proceed to block 720. At block 720, processing logic determines a variance in the rate of change of the value of one or more properties of the process film layer. This rate of change may not meet a threshold performance metric (e.g., the layer may be deposited too fast or too slow).

[0130] In various embodiments, the threshold variation (e.g., variation compared to a baseline measurement, a standard measurement, a golden run measurement, an average or median of several measurements, a two sigma variation from the average, etc.) may be sufficient to merit a change in a processing operation, e.g., a deposition process, an etch process, a cleaning process, etc. Any such threshold variation may be application specific and can be empirically determined based on how much variation is sufficient to induce a process change. The process being monitored may be an etch process, a deposition process, a cleaning process, etc.

[0131] Method 700 may proceed to block 725, where processing logic triggers a process in the processing chamber to correct the deposition or removal rate of the process film layer. In this manner, this portion of method 700 may use differential measurements over time compared to a baseline measurement (e.g., a golden run measurement, an average of several measurements, etc.) to determine if the process in the processing chamber has changed (or fluctuated) over time. In some embodiments, machine learning models may be utilized to identify anomalous fluctuations in the rate of change of the film. In some embodiments, unsupervised machine learning models may be configured to detect optical measurements (or optical measurement derivatives, dimensionless parameters, etc.) indicative of process faults, chamber fluctuations or aging, component failures, etc.

[0132] The method 700 may also proceed to block 730, where the processing logic determines whether the material composition of the process film layer 309 matches a baseline composition expected to be deposited on the transparent thin film. If the measured composition matches the baseline composition (within an acceptable error threshold), the method 700 may proceed to block 705, where processing of the substrate in the processing chamber continues. If there is no match, the method 700 may proceed to block 735, where the processing logic sends a feedback signal to the controller 109 (or other computing device with an interface to a processing system operator) to report errors due to process changes detected by the processing chamber. Process changes may be detected when comparing the detected composition to an expected composition, for example, determining that the composition of the process film has changed. This portion of the method 700 may be performed when the processing system is inactive, in some embodiments.

[0133] In some embodiments, the method 700 can proceed to block 740, where the processing device determines whether the process film layer reaches an accumulation limit (e.g., an accumulation thickness limit within a specification range) during or after processing the substrate in the processing chamber. If not, the method 700 can proceed to block 705, where the processing device processes the substrate in the processing chamber. If the process film layer reaches an accumulation limit, the method 700 can proceed to block 745, where the processing logic triggers the processing chamber to start a cleaning process in the processing chamber. The processing device can determine whether the process film layer reaches an accumulation limit based on a dimensionless value indicating the process film layer thickness. The cleaning process can be intended to clean the surfaces of the processing equipment and the accumulated film to improve future processing results and / or to return the processing equipment to a certain specification. The cleaning process can also use a plasma process, and thus the reflectance measurement signal of the process film can be determined by subtracting the spectrum determined by the OES with the light source on.

[0134] In various embodiments, the method 700 may proceed to block 750, where it is determined whether the process film layer has been removed to a predetermined threshold thickness during such cleaning process triggered by operation 745. This determination may be made based on a dimensionless value indicative of the process film layer thickness. Such a determination may be to ascertain whether the cleaning process has sufficiently reduced the process film layer. After the process film layer has been removed to or beyond the predetermined threshold thickness, the method 700 may proceed to block 755, where the processing logic triggers the processing chamber to terminate the cleaning process being performed in the processing chamber. Once complete, the method 700 may proceed to block 705, where the substrate is processed in the processing chamber.

[0135] Additional or similar methods to method 700 of Figure 7 are contemplated. For example, during processing of a substrate in a processing chamber, processing logic may detect when a process film layer reaches a threshold level of thickness on transparent thin film 307. Processing logic may further trigger an end of a deposition process that deposits a deposition film layer in the processing chamber. Processing logic may make other similar decisions to update the process or process state of the processing chamber to improve substrate throughput, quality, and / or reduce process variations.

[0136] According to some embodiments, the processing chamber can perform a cleaning process after all substrates (or a group of substrates) have been processed, after a predefined number of operating hours, after a predefined range of times with the plasma in the chamber active, after a volume of processing material has been consumed, or at some other period. The processing logic can determine exactly the moment when the endpoint of this cleaning is reached on the distal surface 120B or 420B of the transparent membrane 307. The processing logic can then send a trigger to the processing chamber to stop cleaning and proceed to the next step, thereby improving throughput. If the distal surface 120B or 420B does not clean at the same rate as the rest of the chamber, this difference can be pre-characterized and compensated for via a look-up table in the controller 109.

[0137] FIG. 8 illustrates model training and model application workflows for prediction and simulation of chamber operation updates according to aspects of the disclosure. In an embodiment, the model training workflow 805 can be executed on a server, which may or may not include a chamber operation update prediction application, to which the trained model is provided, and which can execute the model application workflow 817. The model training workflow 805 and the model application workflow 817 can be executed by processing logic executed by a processor of a computing device (e.g., modeling system 470 of FIG. 4). One or more of these workflows 805, 817 can be executed by one or more machine learning modules executed on the processing device and / or other software and / or firmware executed on the processing device. The workflows can be executed by one or more physics-based models, such as optical models. The workflows can be executed by one or more digital twin models. As used herein, a digital twin is a digital replica of a physical asset, such as a manufactured part. The digital twin includes features of the physical asset at each stage of the manufacturing process, including, but not limited to, coordinate dimensions, weighting features, material features (e.g., density, surface roughness), electrical features (e.g., conductivity), optical features (e.g., reflectivity), etc. In some embodiments, the physics-based and digital twin models can be trained, for example, to compensate for unknown information (e.g., variations in manufacturing tolerances of components), correct for insufficient assumptions (e.g., non-ideal fluid flow), etc.

[0138] A model training workflow 805 is for training one or more machine learning models (e.g., deep learning models), physics-based models, etc., to perform one or more of determining, predicting, correcting, updating, etc. tasks associated with a chamber operation update predictor (e.g., determining updates to process operation associated with a processing chamber) and / or associated with a chamber matching tool, a sensor matching tool, a recipe development tool, and / or a process control tool, as shown in Figure 4. A model application workflow 817 is for applying one or more trained models to perform determining and / or updating, etc. tasks on chamber data (e.g., raw sensor data indicative of a condition of a processing chamber, synthetic data). One or more of the machines can receive process result data (e.g., chamber operation updates).

[0139] Various model outputs are described herein. A particular number and arrangement of models are described and illustrated. However, it should be understood that the number and type of models used, as well as the arrangement of such models, can be modified to achieve the same or similar end result. Thus, the arrangement of models (e.g., machine learning models, physics-based models, etc.) described and illustrated are merely examples and should not be construed as limiting.

[0140] In some embodiments, one or more models are trained to perform one or more of the aforementioned tasks. Each task may be performed by a separate model. Alternatively, a single model may perform each of the tasks or a portion of the tasks. Additionally or alternatively, different models may be trained to perform different combinations of the tasks. In one example, one or more machine learning models may be trained, where the trained machine learning (ML) model is a single shared neural network with multiple shared layers and multiple higher level individual output layers, each of the output layers outputting a different prediction, classification, discrimination, etc. Tasks that the one or more trained models may be trained to perform may include operations such as chamber operation update predictor 864, chamber quality index generator 870, performance matching model 874, corrective action predictor 878, etc. As previously discussed, various input data, such as sensor data, pre-processing data, synthetic data, etc., indicative of the state of the processing chamber before and / or after a chamber operation (e.g., a cleaning operation, a deposition operation, an etch operation, etc.), may be received and processed by various machine learning models, rule-based models, physics-based models, etc., associated with the model application workflow 817.

[0141] The chamber operation update predictor 864 can output various values ​​(e.g., chamber update values, process recipe value adjustments, etc.) corresponding to various chamber operation updates based on various characteristic values ​​of a process film layer on the inner surface of the processing chamber wall (e.g., on the processing chamber liner). The various values ​​can correspond to an index indicating a state of the processing chamber. In some embodiments, the various values ​​correspond to a ratio and / or percentage of an acceptable process film layer thickness (e.g., critical wall thickness). The various values ​​can be used by a processing device (e.g., of the server 420 of FIG. 4 ) to determine updates to the chamber operation (e.g., operational parameters, control knobs, durations, etc.). The chamber operation update predictor 864 generates chamber update data 866 that can be utilized to modify the chamber, chamber operation, etc., e.g., to improve the performance of the chamber.

[0142] The chamber quality index generator 870 can generate quality index data 872. In some embodiments, the quality index data 872 can include a dimensionless index indicative of the quality of the chamber. In some embodiments, the quality index data 872 can include one or more CWI indices. In some embodiments, the chamber quality index values ​​(e.g., CWI values) can be utilized as inputs for additional analysis (e.g., input data 862 to the chamber operation update predictor 864, performance matching model 874, corrective action predictor 878, etc. can include one or more CWI values).

[0143] The performance matching model 874 can accept input data 862 and generate as output performance matching data 876. The performance matching model 874 can include one or more algorithms, physics-based models, machine learning models, etc. configured to accept input data (e.g., spectral data, OWPS data, CWI data) and generate as output performance matching data 876. The performance matching data 876 can include data associated with corrective actions to be taken associated with one or more forms of matching, e.g., chamber-to-chamber matching, matching of current chamber performance to historical chamber performance, etc.

[0144] Corrective action predictor 878 can accept input data 862 and generate corrective action data 880 as an output. The input data can include sensor data, process parameter data, recipe data, product data, optical data, chamber quality index data, etc. The output can include recommended component replacements, recommended subsystem investigations (e.g., pressure subsystem, plasma generation subsystem, etc.), recommended maintenance, etc. In some embodiments, one or more blocks shown in FIG. 8 may not be executed, for example, the analysis data flow may not include the performance matching operations associated with blocks 874 and 876.

[0145] 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 with a classifier or regression layer that maps features to a desired output space. A convolutional neural network (CNN), for example, hosts multiple layers of convolutional filters. Pooling can be performed to address nonlinearities at the bottom layer, and a multi-layer perceptron is typically added on top to map the top layer features extracted by the convolutional layer to a decision (e.g., classification output). Deep learning is a class of machine learning algorithms that uses 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 learned supervised (e.g., classification) and / or unsupervised (e.g., pattern analysis). Deep neural networks include a hierarchy of layers, with different layers learning different levels of representations corresponding to different levels of abstraction. In deep learning, each level learns to transform its input data into a slightly more abstract and complex representation. In plasma process tuning, for example, the raw inputs can be chamber data (e.g., raw sensor chamber measurement parameters such as temperature, pressure, material thickness, and vacuum conditions in the chamber), a second layer can consist of feature data associated with parameters associated with the substrate processing procedure or process recipe, and a third layer can include historical thresholds and chamber conditions under similar conditions. In particular, the deep learning process can teach itself which features to optimally place at which level. The "deep" in "deep learning" refers to the number of layers through which data is transformed. More precisely, deep learning systems have substantial contribution allocation path (CAP) depth. A CAP is a chain of transformations from input to output. A CAP describes the potentially causal connections between inputs and outputs. For forward neural networks, the CAP depth can be the depth of the network, which can be the number of hidden layers plus one.For recurrent neural networks, where a signal can propagate through a layer more than once, CAP depth is potentially unlimited.

[0146] In one embodiment, the one or more machine learning models are recurrent neural networks (RNNs). RNNs are a type of neural network that includes memory to enable the neural network to capture time dependencies. RNNs are capable of learning input-output mappings that depend on both current and past inputs. RNNs address past and future sensor and / or process result measurements and make predictions based on this continuous metrology information. RNNs can be trained using a training data set to generate a fixed number of outputs (e.g., to determine a set of substrate processing rates, determine chamber conditions, and production mode requirements and condition thresholds). One type of RNN that can be used is a long short-term memory (LSTM) neural network.

[0147] Training a neural network can be accomplished with supervised learning, which involves feeding the network a training dataset of labeled inputs, observing its output, defining an error (by measuring the difference between the output and the label value), and adjusting the network weights across all layers and nodes to minimize the error using techniques such as deep gradient descent and backpropagation. In many applications, repeating this process across many labeled inputs in the training dataset results in a network that can produce correct outputs when presented with inputs different from those present in the training dataset.

[0148] For model training workflow 805, a training data set including hundreds, thousands, tens of thousands, hundreds of thousands, or more of chamber data 810 (e.g., sensor data indicative of associated processing chamber conditions, OWPS data, synthetic data) must be used to form the training data set. In some embodiments, the training data set may also include associated process result data 812 (e.g., measured chamber variables (e.g., critical dimensions, uniformity requirements, film thickness results, etc.)) to form the training data set, where each data point may include various labels or classifications of one or more types of useful information. In either case, it may include, for example, data indicative of the condition of one or more processing chambers undergoing a process operation and / or associated process results of a substrate evaluated during and / or after the process operation. This data may be processed to generate one or more training data sets 836 for training one or more machine learning models. The machine learning models may be trained to predict, for example, chamber process operation parameter updates.

[0149] To perform the training, the processing logic inputs the training dataset 836 to one or more untrained machine learning models. Prior to inputting the first input to the machine learning models, the machine learning models may be initialized. The processing logic trains the untrained machine learning models based on the training dataset to generate one or more trained machine learning models that perform the various operations described above.

[0150] Training can be performed by inputting one or more of the chamber data 810 and the process result data 812 into the machine learning model one at a time.

[0151] The machine learning model processes this input to generate an output. An artificial neural network includes an input layer consisting of the values ​​of the data points. The next layer is called the hidden layer, where the nodes of the hidden layer each receive one or more of the input values. Each node includes parameters (e.g., weights) to apply to the input values. Each node thus essentially inputs these input values ​​into a multivariate function (e.g., a nonlinear mathematical transformation) to create an output value. The next layer can be another hidden layer or an output layer. In either case, the nodes of the next layer receive the output values ​​from the nodes of the previous layer, and each node applies weights to those values ​​and then generates its own output value. This can be done at each layer. The last layer is the output layer, where there is one node for each class, prediction, and / or output that the machine learning model can make.

[0152] Thus, the output may include one or more predictions or inferences. For example, the output predictions or inferences may include one or more corrections to the plasma process data (e.g., corrections to one or more plasma exposure durations), updates to the cleaning recipe (e.g., updates to the duration of the cleaning recipe). The processing logic may cause the processing chamber to perform the updated process operation. The processing logic may compare the results of the process operation after the update to the results of the process operation before the update. The processing logic may determine an error (i.e., chamber update error) based on the difference between the identified process operation results. The processing logic adjusts the weights of one or more nodes in the machine learning model based on the error. An error term or delta may be determined for each node in the artificial neural network. Based on the error, the artificial neural network adjusts one or more of its parameters (weights for one or more inputs of a node) for one or more of its nodes. The parameters may be updated in a backpropagation manner, such that the nodes in the highest layer are updated first, followed by the nodes in the next layer, and so on. An artificial neural network includes multiple layers of "neurons," each of which receives as inputs values ​​from neurons in a previous layer. The parameters for each neuron include weights associated with the values ​​received from each of the neurons in the previous layer. Adjusting the parameters can thus include adjusting weights assigned to each of the inputs to one or more neurons in one or more layers within the artificial neural network.

[0153] After the model parameters are optimized, model validation can be performed to determine whether the model has been improved and to determine the current accuracy of the deep learning model. After one or more training rounds, the processing logic can determine whether a stopping criterion has been met. The stopping criterion can be a target level of accuracy, a target number of processed data from the training dataset, a target amount of parameter change relative to one or more previous data points, combinations thereof, and / or other criteria. In one embodiment, the stopping criterion is met when at least a minimum number of data points have been processed and at least a threshold accuracy has been achieved. The threshold accuracy can be, for example, 70%, 80%, or 90% accuracy. In one embodiment, the stopping criterion is met when the accuracy of the machine learning model stops improving. If the stopping criterion is not met, further training is performed. If the stopping criterion is met, training may be completed. After the machine learning model is trained, the model can be tested using a reserved portion of the training dataset.

[0154] After the one or more trained machine learning models 838 are generated, these trained machine learning models 838 can be stored in model storage 845 and can be added to a manufacturing system (e.g., manufacturing system 402 in FIG. 4), a modeling system (e.g., modeling system 470 in FIG. 4), etc. The modeling system, manufacturing system, etc. can then use the one or more trained ML models 838 as well as additional processing logic to perform an automatic mode in which user manual input of information is minimized, or in some cases may even be eliminated.

[0155] For the apply model workflow 817, according to one embodiment, the input data 862 can be input to a chamber operation update predictor 864, which can include a trained machine learning model. Based on the input data 862, the chamber operation update predictor 864 outputs information indicative of updates associated with chamber process operations corresponding to conditions of the processing chamber.

[0156] In some embodiments, the input data 862 can be provided to a chamber quality index generator 870, which can output an indication of the chamber quality as quality index data 872. The chamber quality index generator 870 can include one or more trained machine learning models. In some embodiments, the chamber quality index values ​​can be utilized as inputs to further analytical models, for example, machine learning models.

[0157] In some embodiments, the input data 862 can be provided to a performance matching model 874. The performance matching model 874 can include one or more machine learning models. The performance matching model 874 can be configured to generate data associated with performance matching, such as performance matching data 876. Performance matching can include matching the chamber performance to the performance of another chamber (e.g., changing the performance of the chamber to be more similar to the second chamber), matching the performance of the chamber to historical performance (e.g., changing the performance of the chamber to be more similar to a previously performed process run that resulted in a product that met a performance threshold), etc.

[0158] In some embodiments, the input data 862 can be provided to a corrective action predictor 878. The corrective action predictor 878 can include one or more trained machine learning models. The corrective action predictor 878 can include a supervised machine learning model, e.g., a model trained using data labeled with associated corrective actions. The corrective action predictor 878 can generate corrective action data 880 as an output, where the corrective action data 880 can include proposed corrective actions associated with the input data 862.

[0159] 9 shows a block diagram of an exemplary computing device that operates in accordance with one or more aspects of the present disclosure. In various illustrative examples, various components of computing device 900 may represent various components of client device 450, measurement system 410, server 420, data store 140, and modeling system 470 shown in FIG.

[0160] The exemplary computing device 900 can be connected to other computer devices in a LAN, an intranet, an extranet, and / or the Internet. The computing device 900 can operate in the capacity of a server in a client-server network environment. The computing device 900 can be a personal computer (PC), a set-top box (STB), a server, a network router, switch or bridge, or any device capable of executing (sequentially or otherwise) a set of instructions that specify actions to be taken by the device. Furthermore, although only a single exemplary computing device is shown, the term "computer" shall also be taken to include any collection of computers that individually or together execute a set (or sets) of instructions to perform any one or more of the methodologies discussed herein.

[0161] The exemplary computing device 900 may include a processing device 902 (also referred to as a processor or CPU), a main memory 904 (e.g., read only memory (ROM), flash memory, dynamic random access memory (DRAM), such as synchronous DRAM (SDRAM), etc.), a static memory 906 (e.g., flash memory, static random access memory (SRAM), etc.), and a secondary memory (e.g., a data storage device 918), which may communicate with each other via a bus 930.

[0162] The processing device 902 represents one or more general-purpose processing devices, such as a microprocessor, a central processing unit, or the like. More specifically, the processing device 902 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. The processing device 902 may also be one or more special-purpose processing devices, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, or the like. According to one or more aspects of the present disclosure, the processing device 902 may be configured to execute instructions implementing the methods 500A-D illustrated in FIGS. 5A-5D and / or the processes 600A-D illustrated in FIGS. 6A-6D.

[0163] The exemplary computing device 900 may further comprise a network interface device 908, which may be communicatively coupled to a network 920. The exemplary computing device 900 may further comprise a video display 910 (e.g., a liquid crystal display (LCD), touch screen, or cathode ray tube (CRT)), an alphanumeric input device 912 (e.g., a keyboard), a cursor control device 914 (e.g., a mouse), and an audio signal generating device 916 (e.g., a speaker).

[0164] The data storage device 918 may include a computer-readable storage medium (or, more specifically, a non-transitory machine-readable storage medium) 928 having stored thereon one or more sets of executable instructions 922. According to one or more aspects of the present disclosure, the executable instructions 922 may include executable instructions associated with performing the methods 500A-D illustrated in Figures 5A-5D and / or the processes 600A-D illustrated in Figures 6A-6D.

[0165] The executable instructions 922 may also reside, completely or at least partially, within the main memory 904 and / or within the processing device 902 during execution thereof by the exemplary computing device 900, with the main memory 904 and the processing device 902 also constituting computer-readable storage media. The executable instructions 922 may further be transmitted or received over a network via the network interface device 908.

[0166] Although the computer-readable storage medium 928 is illustrated in FIG. 9 as a single medium, the term "computer-readable storage medium" should be interpreted to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store one or more sets of operating instructions. The term "computer-readable storage medium" should also be interpreted to include any medium capable of storing or encoding a set of instructions for execution by a machine that causes the machine to perform any one or more of the methodologies described herein. Thus, the term "computer-readable storage medium" should be interpreted to include, but is not limited to, solid-state memory, and optical and magnetic media.

[0167] Some portions of the above detailed descriptions are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here conceived to be a self-consistent sequence of steps leading to a desired result. These steps require physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, primarily for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.

[0168] It should be noted, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise, as will be apparent from the discussion that follows, discussions utilizing terms such as "identify," "determine," "store," "adjust," "cause," "return," "compare," "create," "stop," "load," "copy," "inject," "exchange," "execute," "receive," and the like throughout the description will be understood to refer to operations and processes of a computer system or similar electronic computing device that manipulates and transforms data, represented as physical (electronic) quantities in the computer system's registers and memory, into other data, similarly represented as physical quantities in the computer system's memory or registers or other such information storage, transmission, or display device.

[0169] An example of the present disclosure also relates to an apparatus for carrying out the methods described herein. This apparatus can be specially constructed for the required purpose, or can be a general-purpose computer system selectively programmed by a computer program stored in the computer system. Such a computer program can be stored in a computer-readable storage medium, such as any type of disk, including, but not limited to, optical disk, compact disk read-only memory (CD-ROM), and magneto-optical disk, read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic disk storage media, optical storage media, flash memory devices, other types of machine-accessible storage media, or any type of media suitable for storing electronic instructions, each of which is coupled to a computer system bus.

[0170] The above description has set forth numerous specific details, such as examples of specific systems, components, methods, etc., to provide a good understanding of some embodiments of the present disclosure. However, it will be apparent to those skilled in the art that at least some embodiments of the present disclosure can be practiced without these specific details. In other instances, well-known components or methods have not been described in detail or have been presented in simple block diagram form to avoid unnecessarily obscuring the present disclosure. Thus, the specific details described are merely illustrative. It is contemplated that certain embodiments may vary from these illustrative details and still be within the scope of the present disclosure.

[0171] Throughout this specification, a reference to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment. Thus, the phrases "in one embodiment" or "in an embodiment" appear in various places throughout this specification, but do not necessarily all refer to the same embodiment. In addition, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". When the term "about" or "approximately" is used herein, it is intended to mean that the nominal value presented is accurate to within ±10%.

[0172] Although the operations of the methods herein are illustrated and described in a particular order, the order of the operations of each method may be changed, such that certain operations may be performed in reverse order, certain operations may be performed at least in part concurrently with other operations, etc. In alternative embodiments, separate operation instructions or sub-operations may be performed intermittently and / or alternately.

[0173] It is to be understood that the above description is intended to be illustrative, and not limiting. Many other embodiments will be apparent to those skilled in the art upon reading and understanding the above description. The scope of the present disclosure should therefore be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

1. receiving, by a processing device, first sensor data indicative of a condition of a wall corresponding to a first process chamber, the first sensor data including optical spectrum data; determining, by the processing device, a first value based on the first sensor data, the first value corresponding to a first amount of product disposed along a surface of the wall at a first time; determining, by the processing device, a first update of a first process operation associated with the first processing chamber based on the first value; performing, by the processing device, one or more of: (i) preparing a notification indicative of the first update for presentation on a graphical user interface (GUI); or (ii) causing the first process operation to be performed in accordance with the first update. The method includes:

2. The first sensor data is received from an optical wall process sensor (OWPS), and the method further comprises: The method of claim 1 , further comprising determining, by the processing device, an update of at least one of a sensor gain value or a sensor offset value corresponding to the OWPS based on the first value.

3. the first value corresponds to the first amount of the product disposed along the surface at a first time, the method comprising: determining, by the processing device, a second value based on the first sensor data, the second value corresponding to a second amount of the product disposed along the surface of the wall at a second time; 2. The method of claim 1, further comprising: determining, by the processing device, a process speed associated with the first process operation using the first value and the second value, wherein the first update is determined further using the process speed.

4. The method of claim 3 , wherein the first update corresponds to at least one of a duration of the first process operation or an endpoint condition of the first process operation.

5. 4. The method of claim 3, wherein at least one of the process rate or the end point condition is related to at least one of a cleaning rate, an etch rate, or a deposition rate corresponding to the first processing chamber.

6. The method of claim 1 , wherein the first value indicates a thickness of the product disposed along the surface of the wall.

7. receiving, by the processing device, second sensor data indicative of a second condition of a second wall corresponding to a second process chamber, the second sensor data including second optical spectrum data; determining, by the processing device, a second value based on the second sensor data, the second value corresponding to a second amount of the product disposed along a second surface of the second wall at a second time; the first update is determined further using the second value. The method of claim 1.

8. performing, by the processing device, a calibration of one or more sensors associated with the first processing chamber based on a comparison between the first value and the second value. The method of claim 7 further comprising:

9. performing, by the processing device, data normalization using the first sensor data to determine the first value, the first value comprising a normalized dimensionless value. The method of claim 1.

10. using, by the processing device, the first sensor data as an input to a machine learning (ML) model; receiving, by the processing device, one or more outputs of the ML model, the one or more outputs indicating the first update; The method of claim 1 further comprising:

11. a processing chamber including a first wall, the first wall having an inner surface positioned within the processing chamber; an optical sensor coupled to the processing chamber, the optical sensor configured to acquire spectral data associated with the interior surface; Memory, a processing device coupled to the memory, the processing device comprising: receiving first sensor data indicative of a condition of a wall corresponding to a first process chamber, the first sensor data including optical spectrum data; determining a first value based on the first sensor data, the first value corresponding to a first amount of product disposed along a surface of the wall at a first time; determining a first update of a first process operation associated with the first processing chamber based on the first value; (i) preparing a notification indicative of the first update for presentation on a graphical user interface (GUI); or (ii) causing the first process operation to be performed in accordance with the first update. system.

12. The first sensor data is received from an optical wall process sensor (OWPS), and the processing device further comprises: The system of claim 11 , for determining an update of at least one of a sensor gain value or a sensor offset value corresponding to the OWPS based on the first value.

13. the first value corresponds to the first amount of the product disposed along the surface at a first time, and the processing device further comprises: determining a second value based on the first sensor data, the second value corresponding to a second amount of the product disposed along the surface of the wall at a second time; 12. The system of claim 11, for determining a process speed associated with the first process operation using the first value and the second value, the first update being determined further using the process speed.

14. The system of claim 11 , wherein the first value indicates a thickness of the product disposed along the surface of the wall.

15. the processing device further comprising: receiving second sensor data indicative of a second condition of a second wall corresponding to a second process chamber, the second sensor data including second optical spectrum data; determining a second value based on the second sensor data, the second value corresponding to a second amount of the product disposed along a second surface of the second wall; The system of claim 11 , wherein the first update is determined further using the second value.

16. A non-transitory machine-readable storage medium storing instructions that, when executed, cause a processing device to: receiving first sensor data indicative of a condition of a wall corresponding to a first process chamber, the first sensor data including optical spectrum data; determining a first value based on the first sensor data, the first value corresponding to a first amount of product disposed along a surface of the wall at a first time; determining a first update of a first process operation associated with the first processing chamber based on the first value; 23. A non-transitory machine-readable storage medium that causes one or more of: (i) preparing a notification indicating the first update for presentation on a graphical user interface (GUI); or (ii) causing the first process operation to be performed in accordance with the first update.

17. The first sensor data is received from an optical wall process sensor (OWPS), and the processing device further comprises:

17. The non-transitory machine-readable storage medium of claim 16, for determining an update of at least one of a sensor gain value or a sensor offset value corresponding to the OWPS based on the first value.

18. the first value corresponds to the first amount of the product disposed along the surface at a first time, and the processing device further comprises: determining a second value based on the first sensor data, the second value corresponding to a second amount of the product disposed along the surface of the wall at a second time; 20. The non-transitory machine-readable storage medium of claim 17, for determining a process speed associated with the first process operation using the first value and the second value, the first update being determined further using the process speed.

19. the processing device further comprising: receiving second sensor data indicative of a second condition of a second wall corresponding to a second process chamber, the second sensor data including second optical spectrum data; determining a second value based on the second sensor data, the second value corresponding to a second amount of the product disposed along a second surface of the second wall; 20. The non-transitory machine-readable storage medium of claim 16, wherein the first update is determined further using the second value.

20. the processing device further comprising: using the first sensor data as an input to an ML model; 20. The non-transitory machine-readable storage medium of claim 16, for receiving one or more outputs of the ML model, the one or more outputs indicating the first update.

Images