Processing tools with hyperspectral cameras for metrology-based analysis

Abstract

Examples are disclosed that relate to a processing tool that includes a hyperspectral camera configured to acquire hyperspectral images of a processing chamber of the processing tool and / or a substrate within the processing tool, and metrology data derived from the hyperspectral images is used to control operation of the processing tool.

Description

[Background technology]

[0001] Semiconductor device manufacturing involves many steps of material deposition, patterning, and removal to form devices on a substrate. Metrology-based analyses can be performed on substrates throughout production for quality control checks. Exemplary metrology analyses that can be performed on substrates include film thickness, non-uniformity, refractive index (RI), stress, particles, and Fourier transform infrared (FTIR) spectroscopy. Summary of the Invention

[0002] This Summary is provided to introduce in a simplified form some concepts that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Moreover, the claimed subject matter is not limited to implementations that solve any or all of the disadvantages noted in any part of this disclosure.

[0003] Examples are disclosed relating to a processing tool including a hyperspectral camera configured to acquire hyperspectral images of a processing chamber of the processing tool and / or a substrate within the processing tool, and metrology data derived from the hyperspectral images is used to control operation of the processing tool.

[0004] In one example, a processing tool includes a processing chamber with an optical interface and a hyperspectral camera positioned to capture hyperspectral images of the interior of the processing chamber through the optical interface of the processing chamber.

[0005] In some such examples, the processing chamber alternatively or additionally comprises a pedestal, and the hyperspectral camera is positioned to capture hyperspectral images of a substrate positioned on the pedestal through the optical interface.

[0006] In some such examples, the processing chamber alternatively or additionally includes a showerhead located opposite the pedestal, and the optical interface is disposed on the showerhead.

[0007] In some such examples, the optical interface is alternatively or additionally disposed on a pedestal.

[0008] In some such examples, the optical interface is alternatively or additionally located on a sidewall of the processing chamber.

[0009] In some such examples, the processing tool alternatively or additionally further comprises one or more optical elements disposed between the optical interface and the hyperspectral camera, the one or more optical elements configured to direct electromagnetic radiation passing through the optical interface to the hyperspectral camera.

[0010] In some such examples, the processing chamber is alternatively or additionally a plasma reactor chamber.

[0011] In some such examples, the processing tool alternatively or additionally further comprises a computing system configured to execute a trained machine learning model configured to receive one or more hyperspectral images from the hyperspectral camera and to output metrology data for the processing chamber based on at least the one or more hyperspectral images.

[0012] In some such examples, the computing system is alternatively or additionally configured to adjust control parameters of a cleaning process for cleaning the processing chamber based on at least the metrology data for the processing chamber.

[0013] In some such examples, the trained machine learning model is alternatively or additionally configured to receive a series of hyperspectral images of a substrate in a processing chamber during a substrate processing cycle, and to output time-based metrology data for the substrate based on at least the series of hyperspectral images of the substrate, and the computing system is configured to adjust one or more control parameters of a process of the substrate processing cycle during the substrate processing cycle based on at least the time-based metrology data for the substrate.

[0014] In some such examples, the trained machine learning model is alternatively or additionally configured to receive one or more hyperspectral images of a first substrate in the processing chamber during or after a first substrate processing cycle, and to output metrology data for the first substrate based on the one or more hyperspectral images of at least the first substrate, and the computing system is configured, for a second substrate processing cycle for a second substrate, to adjust one or more control parameters of the process for the second substrate processing cycle based on the metrology data for at least the first substrate.

[0015] In another example, a computer-implemented method for controlling a processing tool includes receiving one or more hyperspectral images of a processing chamber of the processing tool from a hyperspectral camera, sending the one or more hyperspectral images to a trained machine learning model configured to output metrology data for the processing chamber based on at least the one or more hyperspectral images, and adjusting one or more control parameters of a process performed by the processing tool based on the metrology data for at least the processing chamber.

[0016] In some such examples, the process alternatively or additionally includes a cleaning process for cleaning the processing chamber, and the one or more control parameters include control parameters for the cleaning process.

[0017] In some such examples, the one or more hyperspectral images alternatively or additionally include a series of hyperspectral images of a substrate in a processing chamber, the series of hyperspectral images of the substrate being received from the hyperspectral camera during a substrate processing cycle for the substrate, the trained machine learning model being configured to output time-based metrology data for the substrate, and one or more control parameters being adjusted during the substrate processing cycle for the substrate based on at least the time-based metrology data for the substrate.

[0018] In some such examples, the one or more hyperspectral images alternatively or additionally include one or more hyperspectral images of a first substrate in the processing chamber during or after a first substrate processing cycle, and one or more control parameters are adjusted for a second substrate processing cycle for a second substrate based on the metrology data for at least the first substrate.

[0019] In some such examples, the processing chamber is alternatively or additionally a plasma reactor chamber, and one or more hyperspectral images of the plasma reactor chamber are taken by a hyperspectral camera while plasma is present in the plasma reactor chamber, and the plasma in the plasma reactor chamber is an illumination source for the hyperspectral camera.

[0020] In another example, a processing tool includes a hyperspectral camera positioned to capture hyperspectral images of a substrate within the processing tool, and a computing system configured to execute a trained machine learning model, the trained machine learning model configured to receive one or more hyperspectral images from the hyperspectral camera and output metrology data for the substrate based on at least the one or more hyperspectral images.

[0021] In some such examples, the metrology data includes the thickness of one or more layers of the substrate.

[0022] In some such examples, alternatively or additionally, the metrology data includes the condition of the gap in the substrate.

[0023] In some such examples, alternatively or additionally, the hyperspectral camera has a dynamically adjustable position.

[0024] In some such examples, alternatively or additionally, the hyperspectral camera has a dynamically adjustable angle.

[0025] In some such examples, alternatively or additionally, the metrology data includes a determined amount of stress and / or warpage in the substrate.

[0026] In some such examples, alternatively or additionally, the metrology data includes a determined amount of haze in the substrate. [Brief explanation of the drawings]

[0027] [Figure 1] FIG. 1 is a block diagram of an exemplary processing tool.

[0028] [Figure 2] FIG. 2 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber. [Figure 3] FIG. 3 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber. [Figure 4] FIG. 4 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber. [Figure 5] FIG. 5 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber. [Figure 6]FIG. 6 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber. [Figure 7] FIG. 7 is a schematic diagram illustrating different exemplary placements of a hyperspectral camera in a processing chamber.

[0029] [Figure 8] FIG. 8 is a diagram illustrating an exemplary hyperspectral image captured by a hyperspectral camera.

[0030] [Figure 9] FIG. 9 is a flow diagram illustrating an exemplary method for training and running a machine learning model to perform metrology-based analysis on a processing chamber and / or a substrate within the processing chamber.

[0031] [Figure 10] FIG. 10 is a flow diagram illustrating an exemplary method for implementing hyperspectral image-based metrology-based analysis to control a cleaning process of a processing chamber.

[0032] [Figure 11] FIG. 11 is a flow diagram illustrating an exemplary method for performing hyperspectral image-based metrology-based analysis for in-situ control of a processing tool during a substrate processing cycle.

[0033] [Figure 12] FIG. 12 is a flow diagram illustrating an exemplary method for performing hyperspectral image-based metrology-based analysis for ex-situ in-line control of a processing tool during a substrate processing cycle.

[0034] [Figure 13] FIG. 13 is a block diagram of an exemplary computing system.

[0035] [Figure 14]FIG. 14 is a diagram illustrating different exemplary states of a substrate during the process of filling gaps in the substrate.

[0036] [Figure 15] FIG. 15 is a diagram illustrating schematically different exemplary states of a substrate as a result of a filling process in which gaps in the substrate are filled.

[0037] [Figure 16] FIG. 16 is a diagram illustrating a schematic of an exemplary scenario in which a hyperspectral camera is configured to be dynamically adjustable to adjust the distance of the hyperspectral camera relative to the substrate being imaged.

[0038] [Figure 17] FIG. 17 is a diagram illustrating a schematic of an exemplary scenario in which a hyperspectral camera is configured to be dynamically adjustable to adjust the angle of the hyperspectral camera relative to the substrate being imaged.

[0039] [Figure 18] FIG. 18 is an exemplary graph of multiple plots of the spectral reflectance of a material on a substrate as a function of wavelength and angle of incidence of light on the material.

[0040] [Figure 19] FIG. 19 is an exemplary graph of two plots of the spectral reflectance of two different substrates as a function of wavelength and number of layers for two different substrates.

[0041] [Figure 20] FIG. 20 is a diagram illustrating a schematic of an exemplary configuration in which a hyperspectral camera can be used to measure stress and / or warpage in a substrate. [Figure 21] FIG. 21 is a diagram illustrating a schematic of an exemplary configuration in which a hyperspectral camera can be used to measure stress and / or warpage in a substrate. [Figure 22]FIG. 22 is a diagram illustrating a schematic of an exemplary configuration in which a hyperspectral camera can be used to measure stress and / or warpage in a substrate.

[0042] [Figure 23] FIG. 23 is an exemplary graph of spectral reflectance measured at different wavelengths at different points on a substrate to measure stress and / or warpage at different points on the substrate. [Figure 24] FIG. 24 is an exemplary graph of spectral reflectance measured at different wavelengths at different points on a substrate to measure stress and / or warpage at different points on the substrate.

[0043] [Figure 25] FIG. 25 illustrates an exemplary configuration in which a hyperspectral camera is configured to measure haze on a substrate.

[0044] [Figure 26] FIG. 26 is a diagram illustrating a schematic of an exemplary arrangement of spatially separated light emitters in the light source of a hyperspectral camera used to measure the haze of a substrate. [Figure 27] FIG. 27 is a diagram illustrating a schematic of an exemplary arrangement of spatially separated light emitters in the light source of a hyperspectral camera used to measure the haze of a substrate.

[0045] [Figure 28] FIG. 28 is an exemplary graph of the electromagnetic spectrum including different ranges of wavelengths that can be used for different operations, including measuring substrate haze.

[0046] [Figure 29] FIG. 29 is an exemplary graph containing haze measurements represented by the number of pixels with reflected light and their corresponding intensity.

[0047] [Figure 30]FIG. 30 is a schematic diagram of an exemplary processing tool including multiple hyperspectral cameras.

[0048] [Figure 31] FIG. 31 is a diagram illustrating an example of a processing module including multiple hyperspectral cameras.

[0049] [Figure 32] FIG. 32 is a flow diagram illustrating an exemplary method for dynamically controlling the position of a hyperspectral camera within a processing tool to vary the distance between the hyperspectral camera and the substrate for hyperspectral imaging and analysis.

[0050] [Figure 33] FIG. 33 is a flow diagram illustrating an exemplary method for dynamically controlling the position of a hyperspectral camera to capture hyperspectral images of different substrates from different angles.

[0051] [Figure 34] FIG. 34 is a flow diagram illustrating an exemplary method for performing metrology-based analysis using hyperspectral imagery for control of a processing tool. DETAILED DESCRIPTION OF THE INVENTION

[0052] The term "atomic layer deposition" (ALD) generally refers to a process in which a film is formed as one or more individual layers on a substrate by conformally adsorbing precursors to the substrate sequentially and reacting the adsorbed precursors to form the film layer. Examples of ALD processes include plasma-enhanced ALD (PEALD) and thermal ALD (TALD). PEALD and TALD utilize reactive gases and thermal plasma, respectively, to facilitate the chemical conversion of precursors adsorbed on the substrate into a film on the substrate.

[0053] The term "chemical vapor deposition" (CVD) generally refers to a process in which a solid film is formed on a substrate by directing a flow of one or more precursor gases over the substrate surface under conditions configured to chemically convert the precursor gases into a solid film. The term "plasma-enhanced chemical vapor deposition" (PECVD) generally refers to a CVD process in which a plasma is used to facilitate the chemical conversion of one or more precursor gases into a solid film on a substrate.

[0054] The term "cleaning process" generally refers to a process for cleaning deposited materials from the interior surfaces of a processing chamber. The deposited materials may include materials deposited on a substrate in a deposition process, by-products of a deposition process, residues from an etching process, and / or coatings of one or more materials applied to the processing chamber prior to performing a deposition or etching process.

[0055] The term "control parameter" generally refers to a controllable variable in a process performed in a process chamber. Exemplary control parameters include the temperature of the heater, the pressure in the chamber, the flow rate of each of one or more process gases, and the frequency and power level of the radio frequency power used to form the plasma in the process chamber.

[0056] The term "disposed on" generally describes a structural relationship in which one part is supported by another part. The term "disposed on" itself does not describe a specific relative location of one part to another. For example, an optical interface disposed on a part of a processing chamber or a component of a processing chamber may be flush with the surface of the part or component, may be disposed inward from the surface of the part or component, or may extend beyond the surface of the part or component.

[0057] The term "etching" and its variations generally refer to the removal of material from a structure. A substrate can be etched by a plasma in a plasma processing tool.

[0058] The term "hyperspectral camera" generally refers to an optical device configured to acquire hyperspectral images.

[0059] The term "hyperspectral image" generally refers to a data structure having multiple sub-images, each corresponding to a different wavelength or wavelength band of electromagnetic radiation. Each sub-image is a two-dimensional array of pixels. Each pixel in each sub-image stores an intensity value, which is the intensity of electromagnetic radiation at the corresponding wavelength or wavelength band for that sub-image received from the corresponding spatial location within the process chamber. In some examples, a hyperspectral image can include 100 or more sub-images corresponding to different wavelengths or wavelength bands. In other examples, a hyperspectral image can take the form of a multispectral image, including multiple sub-images each corresponding to a selected wavelength band associated with a different descriptive channel name. Examples of wavelength bands and descriptive channel names include BLUE in band 2 (0.45–0.51 micrometers (μm)), GREEN in band 3 (0.53–0.59 μm), RED in band 4 (0.64–0.67 μm), NEAR INFRARED (NIR) in band 5 (0.85–0.88 μm), SHORT-WAVE INFRARED (SWIR1) in band 6 (1.57–1.65 μm), SHORT-WAVE INFRARED (SWIR2) in band 7 (2.11–2.29 μm), PANCHROMATIC in band 8 (0.50–0.68 μm), CIRRUS in band 9 (1.36–1.38 μm), and THERMAL in band 10 (10.60–11.19 μm). Multispectral cameras can capture images of limited wavelength bands of interest, such as infrared (TIRS1) and thermal infrared (TIRS2) in band 11 (11.50-12.51 μm).

[0060] The term "illumination source" generally refers to a source that provides illumination light for a hyperspectral camera to capture an image.

[0061] The term "inhibitor" generally refers to a compound that can be introduced into a processing chamber and non-conformally deposited on a substrate surface, inhibiting the ALD growth of oxide films.

[0062] The term "metrology data" generally refers to data obtained by measuring one or more observable properties. For example, a hyperspectral camera can be used to obtain metrology data including electromagnetic energy intensity emanating from different spatial locations within a processing chamber. Exemplary observable properties include film thickness, non-uniformity, refractive index (RI), stress, particle detection, and Fourier transform infrared (FTIR) spectroscopy. One or more of such observable properties can be used to calibrate and validate a hyperspectral metrology model.

[0063] The term "optical interface" generally refers to an optically transparent structure positioned between the interior and exterior of a processing chamber for performing hyperspectral imaging of the processing chamber through the optical interface. The term "interior of the processing chamber" refers to the volume of space in which the substrate is located during processing. The optical interface can be located on a wall of the processing chamber or on a structure within the processing chamber, such as a pedestal or showerhead. The optical interface allows electromagnetic radiation for hyperspectral imaging to pass to the hyperspectral camera while preventing the passage of gases.

[0064] The term "optical element" generally refers to a structure configured to guide and / or modify electromagnetic radiation along an optical path. Exemplary optical elements include optical fibers and other waveguides, diffractive and refractive lenses and mirrors, and polarizers or other filters.

[0065] The term "optically transparent" with respect to a material generally refers to the material being suitably transparent to the electromagnetic energy bands imaged by a hyperspectral camera to obtain useful hyperspectral data.

[0066] The term "pedestal" generally refers to the structure that supports the substrate within a processing chamber.

[0067] The term "plasma" generally refers to an ionized gas containing gas-phase positive ions and free electrons.

[0068] The term "plasma reactor chamber" generally refers to a processing chamber in which a plasma can be generated to perform a chemical process on a substrate.

[0069] The term "processing chamber" generally refers to an enclosure in which chemical and / or physical processes are performed on a substrate. The pressure, temperature, gas flow rates, and atmospheric composition within the processing chamber may be controllable to perform the chemical and / or physical processes. Controllable aspects of the atmospheric composition include one or more of the gas mixture or plasma conditions.

[0070] The term "processing tool" generally refers to a machine that includes processing chambers and other hardware configured to perform substrate processing cycles.

[0071] The term "showerhead" generally refers to a structure for distributing gases across a substrate surface within a processing chamber.

[0072] The term "substrate" generally refers to any object that can be positioned on a pedestal in a processing tool for processing.

[0073] The term "substrate processing cycle" generally refers to a set of one or more processes used to cause physical and / or chemical changes on a substrate. For example, a substrate processing cycle can include a deposition cycle in which a thin film is formed on a substrate. The deposition cycle can be performed, for example, by a chemical vapor deposition (CVD) process or an atomic layer deposition (ALD) process. The substrate processing cycle can also include an etch cycle in which material is removed from the substrate. The etch cycle can be performed, for example, by plasma etching.

[0074] The term "time-based metrology data" generally refers to data corresponding to measurements of different properties of an object measured over a period of time.

[0075] The term "trained machine learning model" generally refers to a computer program that has been trained on a dataset to find specific patterns or outputs based on specific inputs. Training can involve, for example, adjusting weights between nodes in a neural network using an algorithm such as backpropagation.

[0076] The term "viewport" generally refers to an optically translucent or transparent window through which the interior of a processing chamber can be viewed.

[0077] As mentioned above, the fabrication of semiconductor devices involves many individual steps of material deposition, patterning, and removal. Metrology data can be collected and analyzed between process steps to monitor the process, both during process development and while performing control checks in production. Such metrology data is often obtained using offline techniques. Examples include scanning electron microscope (SEM) imaging of a substrate cross-section, ellipsometry, and Fourier transform infrared spectroscopy (FTIR).

[0078] The process of obtaining metrology data for a substrate can sometimes take at least two to three hours per substrate. During this time, production may be stopped to ensure the production process is operating within specification requirements. This production stoppage reduces productivity throughout the production process. In addition, some measurement processes are destructive, reducing overall production yield. Furthermore, when an out-of-spec substrate is discovered, it can require significant effort to determine the root cause. Also, multiple substrates may be processed before the problem is discovered. This can result in the need to scrap the substrate.

[0079] In contrast to offline ("ex-situ") metrology, in-situ metrology can be used to efficiently measure many more substrates, and potentially all substrates being processed. In-situ metrology refers to measurements performed on a substrate while it is within a processing tool.

[0080] When metrology is performed for quality control processes, the challenge is to perform sufficient measurements to ensure that processing tools are operating within specification limits. For example, it can be difficult to monitor multiple substrate parameters using data available with current in-situ metrology methods. Performing metrology on tools that utilize plasma for processing can be particularly challenging because the energy of the plasma can interfere with the measurements. Exemplary tools that utilize plasma are plasma deposition tools and plasma etch tools. Exemplary plasma deposition tools are plasma-enhanced atomic layer deposition (PEALD) tools and plasma-enhanced CVD (PECVD) tools.

[0081] Thus, examples are disclosed relating to performing in-situ metrology within a substrate processing tool using hyperspectral imaging of a processing chamber. Briefly, the processing tool can include a processing chamber with an optical interface and a hyperspectral camera positioned to capture hyperspectral images of the interior of the processing chamber through the optical interface. The hyperspectral image includes image data of the processing chamber at multiple different wavelengths of light. Each wavelength of light can potentially provide different information than other wavelengths of light. This can provide more data than other in-situ measurement methods. Furthermore, the hyperspectral images can be acquired in-situ during a substrate processing cycle. This allows a computing system to characterize the substrate in real time while the substrate is in the processing chamber, during processing, or immediately after processing. In some such examples, in-situ process control can be performed by adjusting one or more control parameters of one or more processes during the substrate processing cycle based at least on the acquired metrology data. Such in-situ process control allows for characterization of the substrate for quality in real time without destroying the substrate or stopping the substrate processing cycle. In this manner, in-situ process control provides the technical advantages of increasing substrate quality and substrate processing throughput while reducing costs.

[0082] In some examples, time-based metrology data can be generated from a series of hyperspectral images taken during a substrate processing cycle. The time-based metrology data includes hyperspectral imaging-based metrology measurements taken multiple times throughout the substrate processing cycle. The time-based metrology data can be used to build a time-based model of properties including film growth dynamics (e.g., nucleation delay, growth based on different process steps, etc.). Furthermore, process control can be performed by adjusting one or more control parameters of one or more processes during the substrate processing cycle based at least on the time-based metrology data and / or the time-based model. This can help improve substrate yield compared to not using time-based metrology.

[0083] Additionally, in some instances, metrology-based analysis can also be performed "ex-situ" inline between substrate processing cycles. Ex-situ inline metrology capabilities based on hyperspectral images can determine whether a process is operating within specification requirements. Changes can then be made for run-by-run process control. This can potentially help avoid tool downtime compared to obtaining metrology data using SEM or other ex-situ, destructive or non-destructive techniques.

[0084] In some examples, the computing system is configured to execute a trained machine learning model to analyze the hyperspectral image data. The trained machine learning model is configured to receive one or more hyperspectral images from the hyperspectral camera and output metrology data for the process chamber based on at least the one or more hyperspectral images. The computing system can be further configured to control operation of the process tool based at least on the metrology data.

[0085] Machine learning models can use temporal and spectral signatures to predict electrical and optical properties of films of interest. Furthermore, image data from different spectral bands can be specifically related to different properties of the film being measured. For example, infrared imaging can correlate temperature response with metrics such as film thickness, inhomogeneity, refractive index, resistivity, stress, and particle / defect concentration.

[0086] 1 shows a schematic diagram of an exemplary processing tool 100. The processing tool 100 includes a processing chamber 102 and a pedestal 104 within the processing chamber. The pedestal 104 is configured to support a substrate 106 disposed within the processing chamber 102. The pedestal 104 may include a substrate heater 108. In other examples, the heater may be omitted or located elsewhere within the processing chamber 102.

[0087] The processing tool 100 further includes a showerhead 110, a gas inlet 112, and flow control hardware 114. In other examples, the processing tool can include nozzles or other devices for supplying gases to the processing chamber 102, as opposed to or in addition to a showerhead. The flow control hardware 114 is connected to one or more process gas sources 116. If the processing tool 100 includes a deposition tool, the process gas source 116 can include, for example, one or more precursor sources and an inert gas source used as a diluent gas and / or a purge gas. If the processing tool 100 includes an etching tool or a CVD tool, the process gas source 116 can include, for example, one or more etchant gas sources and one or more inert gas sources.

[0088] Flow control hardware 114 can control gases to flow from a process gas source into the processing chamber 102 via the gas inlet 112. The flow control hardware 114 can include one or more flow controllers (e.g., mass flow controllers), valves, conduits, and other hardware for fluidly connecting one or more selected gas sources with the gas inlet 112. In other examples, the processing chamber can include one or more additional gas inlets.

[0089] The processing tool 100 further comprises an exhaust system 118. The exhaust system 118 is configured to receive gases exiting the processing chamber 102. In some examples, the exhaust system 118 is configured to actively remove gases from the processing chamber 102 and / or apply a partial vacuum. The exhaust system 118 can comprise any suitable hardware, including one or more pumps.

[0090] The processing tool 100 further includes an RF power source 120 electrically connected to the pedestal 104. The RF power source 120 is configured to form a plasma. The plasma can be used to form reactive species, such as radicals, in a film deposition or etching process. The showerhead 110 is configured as a grounded counter electrode in this example. In other examples, the RF power source 120 can supply RF power to the showerhead 110 or another suitable electrode structure. The processing tool 100 includes a matching network 122 for impedance matching of the RF power source 120. The RF power source 120 can be configured for any suitable frequency and power. Examples of suitable frequencies include frequencies in the range of 300 kHz to 90 MHz. More specific examples of suitable frequencies include 400 kHz, 13.56 MHz, 27 MHz, 60 MHz, 90 MHz, and 2.45 GHz. Examples of suitable power include power between 0 and 15 kilowatts. In some examples, the RF power source 120 is configured to operate at multiple different frequencies and / or powers. In other examples, the processing tool may alternatively or additionally include a remote plasma generator (not shown), which can be used to generate plasma at a location remote from the substrate being processed.

[0091] The processing chamber 102 further includes an optical interface 126 disposed on the sidewall 124 of the processing chamber. In other examples, the optical interface can be disposed on a different surface, such as the ceiling or floor of the processing chamber. The optical interface 126 is an interface that allows electromagnetic radiation of a desired wavelength band to pass from the interior of the processing chamber to a hyperspectral camera located outside the processing chamber while preventing the passage of gases. In the illustrated example, the optical interface includes an optically transparent window positioned in an opening formed in the sidewall of the processing chamber. In other examples, the optical interface can be configured as a window in the top or bottom wall of the processing chamber. The window in the wall of the processing chamber can be configured as a viewport. The term "viewport" generally refers to an optically transparent window in the wall of the processing chamber configured to allow an operator to observe the interior of the processing chamber during processing. As described below, in further examples, the optical interface can include an optically transparent surface located on a component within the processing chamber. Exemplary components include a pedestal and a showerhead.

[0092] The processing tool 100 also includes a hyperspectral camera 128 positioned to capture hyperspectral images of the interior of the processing chamber 102 through an optical interface 126 of the processing chamber 102. The processing tool 100 may include any suitable number of hyperspectral cameras for capturing hyperspectral images of the processing chamber 102 and / or the substrate 106. In some implementations, the processing tool may include multiple processing chambers / processing stations, and the processing tool may include one or more hyperspectral cameras positioned to capture hyperspectral images of some or all of the multiple processing chambers / processing stations.

[0093] The optical interface for hyperspectral imaging of the processing chamber can be located at any suitable location within the processing chamber. Figures 2-7 schematically illustrate different exemplary arrangements of the hyperspectral camera and optical interface for imaging the processing chamber.

[0094] First, FIG. 2 illustrates an exemplary process chamber 200 including a pedestal 202 on which a substrate 204 is positioned. The process chamber includes an optical interface 208 disposed on a top plate 206 of the process chamber 200. A hyperspectral camera 210 is positioned to capture hyperspectral images of the substrate 204 through the optical interface 208. The optical interface 208 can be formed from any material that is suitably transparent to the electromagnetic energy bands to be imaged and suitably opaque to the process gases. Exemplary electromagnetic energy bands include ultraviolet, visible, and infrared bands. Exemplary materials for the optical interface 208 include fused quartz, fused silica, sapphire, windows with coatings that reduce reflection, windows with coatings that prevent degradation, and windows with coatings that reduce the effect of materials deposited on the windows. In this example, the optical interface 208 is configured as a viewport through which the hyperspectral camera 210 can directly image the substrate 204. In the illustrated example, the hyperspectral camera 210 is positioned to image the substrate 204 through the optical interface. In other examples, the hyperspectral camera 210 can be positioned to image any other suitable structure within the processing chamber 200 through an optical interface. The hyperspectral camera 210 can be equipped with any suitable lens and / or other optical elements to enable imaging of a desired field of view (FOV).

[0095] FIG. 3 illustrates another exemplary processing chamber 300. The processing chamber 300 includes a pedestal 302 on which a substrate 304 is positioned. The pedestal 302 is configured for backside processing. In such processing, the side of the substrate 304 facing the pedestal is exposed to processing gases using one or more processing gas outlets (not shown) in the pedestal 302. Here, an optical interface 308 is disposed on a surface 306 of the pedestal 302. The optical interface 308 comprises an optically transparent structure that allows a hyperspectral camera 310 to capture hyperspectral images of the processing chamber 300 and / or the substrate 304. For example, images of the processing chamber 300 can be acquired during a processing chamber cleaning process in the absence of the substrate 304. Alternatively or additionally, images of the backside of the substrate 304 can be acquired during backside processing of the substrate. The hyperspectral camera 310 can include any suitable optical elements for imaging the substrate 304 and / or desired portions of the processing chamber 300. Additionally or alternatively, any suitable optical element may be positioned intermediate the hyperspectral camera 310 and the processing chamber 300 to image desired portions of the substrate 304 and / or processing chamber 300. In some examples, pedestals not configured for backside processing can also be equipped with optical interfaces for hyperspectral imaging. Such pedestal optical interfaces can be used, for example, to perform hyperspectral imaging measurements during a processing chamber cleaning process.

[0096] FIG. 4 illustrates another exemplary processing chamber 400. The processing chamber 400 includes a pedestal 402 on which a substrate 404 is positioned. The processing chamber 400 includes an optical interface 408 disposed on a sidewall 406. The optical interface 408 comprises an optically transparent structure on which a hyperspectral camera 410 is positioned to capture hyperspectral images of the processing chamber 400 and / or the substrate 404. The hyperspectral camera 410 is disposed outside the processing chamber 400, away from the optical interface 408. Additionally, an optical element 412 is disposed between the optical interface 408 and the hyperspectral camera 410. The optical element 412 is configured to direct electromagnetic radiation passing through the optical interface 408 to the hyperspectral camera 410. In this manner, the hyperspectral camera 410 captures hyperspectral images of the processing chamber 400 and / or the substrate 404 through the optical element 412. The optical element 412 is illustrated as an optical fiber or other waveguide. However, optical element 412 represents one or more optical elements that can collectively direct electromagnetic radiation passing through optical interface 408 to hyperspectral camera 410. Exemplary optical elements include an optical fiber, a bundle of optical fibers, another optical waveguide, one or more refractive / diffractive lenses, refractive / diffractive mirrors, a waveguide, and / or a filter such as a polarizer. In some examples, one or more optical elements can have adjustable optical power, which can be useful for focusing electromagnetic radiation of different wavelengths onto the image sensor of hyperspectral camera 410.

[0097] In some examples, the hyperspectral camera 410 can be calibrated to accommodate an oblique incidence angle of the optical interface 408 relative to the substrate 404. For example, a distortion correction transform can be applied to the hyperspectral image captured by the hyperspectral camera 410 to accommodate an oblique incidence angle based on the calibrated position of the hyperspectral camera 410.

[0098] 5 illustrates another exemplary processing chamber 500. The processing chamber 500 includes a pedestal 502 on which a substrate 504 is positioned. The processing chamber 500 further includes a showerhead 506 located opposite the pedestal 502. The showerhead 506 includes an optical interface 508 disposed on a surface 512 of the showerhead 506. The optical interface 508 includes an optically transparent structure that allows a hyperspectral camera 510 to capture hyperspectral images of the processing chamber 500 and / or the substrate 504.

[0099] FIG. 6 illustrates another exemplary processing chamber 600. The processing chamber 600 includes a pedestal 602 on which a substrate 604 is positioned. The processing chamber 600 further includes a showerhead 606 opposite the pedestal 602. An optical interface 608 is disposed on a surface 610 of the showerhead 606. The optical interface 608 includes an optically transparent structure that allows a hyperspectral camera 612 to capture hyperspectral images of the processing chamber 600 and / or the substrate 604. An optical element 616 is disposed between the optical interface 608 and the hyperspectral camera 612. The optical element 616 is configured to direct electromagnetic radiation from the optical interface 608 through the showerhead 606 to the hyperspectral camera 612. Thus, the hyperspectral camera 612 captures hyperspectral images of the processing chamber 600 and / or the substrate 604 through the optical element 616. In some examples, the optical element 616 includes an optical fiber or a bundle of optical fibers. Other optical elements may also be used. Examples include one or more refractive or diffractive lenses and / or mirrors.

[0100] FIG. 7 illustrates another exemplary processing chamber 700. The processing chamber 700 includes a pedestal 702 on which a substrate 704 is positioned. The processing chamber 700 further includes a showerhead 706 located opposite the pedestal 702. Multiple optical interfaces 708A, 708B, and 708C are disposed on a surface 710 of the showerhead 706. Each optical interface 708A, 708B, and 708C includes an optically transparent structure that allows a hyperspectral camera 712 to capture hyperspectral images of the processing chamber 700 and / or the substrate 704. The hyperspectral camera 712 is disposed at a location separate from the optical interfaces 708A, 708B, and 708C. Here, the hyperspectral camera 712 is located on a top plate 714 of the processing chamber 700. In other examples, the hyperspectral camera can be positioned at any other suitable location.

[0101] A plurality of optical elements 716A, 716B, 716C are disposed between corresponding optical interfaces 708A, 708B, 708C and the hyperspectral camera 712. The optical elements 716A, 716B, 716C are configured to direct electromagnetic radiation passing through the plurality of optical interfaces 708A, 708B, 708C through the showerhead 706 to the hyperspectral camera 712. In the illustrated example, the optical elements 716A, 716B, 716C each comprise an optical fiber, a bundle of optical fibers, or other optical waveguides or waveguide systems. Other optical elements, such as one or more refractive or diffractive lenses and / or mirrors, may alternatively or additionally be used. The plurality of optical interfaces 708A, 708B, 708C may be disposed on the surface 710 of the showerhead 706 in any suitable arrangement to collectively capture hyperspectral images of the processing chamber 700 and / or the substrate 704. Three optical interfaces 708A, 708B, 708C are shown in Figure 7. In other examples, any other suitable number of optical interfaces and associated optical elements may be used.

[0102] In some examples, the hyperspectral camera 712 is configured to capture images from the optical elements 716A, 716B, and 716C in spatially separated regions on the image sensor of the hyperspectral camera 712. In other examples, the hyperspectral camera 712 is configured to stitch together images collected from multiple optical elements 716A, 716B, and 716C to reconstruct a spatially contiguous hyperspectral image of the process chamber 700 and / or substrate 704. In other examples, the images from the optical elements 716A, 716B, and 716C are directed onto the hyperspectral camera's image sensor so that they partially or completely overlap. In such examples, the overlapping images from the optical elements 716A, 716B, and 716C can be analyzed using a trained machine learning function. Exemplary machine learning functions are described in more detail below.

[0103] The above-described arrangements are provided as non-limiting examples. The hyperspectral camera may be arranged in any suitable manner to capture images of the process chamber and / or substrate within the process chamber.

[0104] Returning to FIG. 1 , hyperspectral camera 128 is configured to capture a hyperspectral image including multiple sub-images, each corresponding to a different wavelength or wavelength band. In some examples, hyperspectral camera 128 is configured to capture a hyperspectral image including sub-images corresponding to multiple different wavelength bands in the 250-1000 nanometer range. In other examples, wavelengths outside this range may alternatively or additionally be captured. Furthermore, in some examples, hyperspectral camera 128 is configured to capture a hyperspectral image including 20 or more sub-images, each at a different wavelength or wavelength band. In other examples, the hyperspectral camera may be configured to capture fewer than 20 sub-images.

[0105] In some examples, the hyperspectral camera 128 includes wavelength-selective filters that isolate different wavelength bands for hyperspectral imaging. An example of such a filter is a diffraction grating. In some examples, the filters are adjustable to select different wavelength bands. In other examples, the high-resolution filter is configured to selectively filter multiple fixed wavelength bands.

[0106] In some examples, the hyperspectral camera 128 includes an illumination source 129. Such an illumination source 129 can include a broadband spectrum illumination source filtered by a high-resolution filter. In other examples, the illumination source 129 can be configured to emit light at specific wavelengths of interest. In some examples, where the processing chamber 102 is a plasma reactor chamber, the plasma present in the plasma reactor chamber can serve as the illumination source for the hyperspectral camera 128. In further examples, the hyperspectral camera 128 can capture hyperspectral images without an illumination source. In such examples, the hyperspectral camera 128 can instead utilize heat present in the processing chamber 102 to provide thermal-based hyperspectral data.

[0107] FIG. 8 schematically illustrates an exemplary hyperspectral image 800 captured by a hyperspectral camera, such as the hyperspectral camera 128 shown in FIG. 1 . The hyperspectral image 800 includes multiple sub-images 802 corresponding to different wavelength bands (λ) of the electromagnetic spectrum. Each sub-image includes multiple pixels 804. Each pixel of a sub-image has a location defined by an X-axis 806 and a Y-axis 808 and an intensity value at a wavelength (λ) associated with the wavelength band of the sub-image. Each pixel of the hyperspectral image 800 includes a spatially mapped set of hyperspectral data, which includes intensity data for each sub-image. If multiple hyperspectral images are captured over a period of time, an additional dimension (e.g., index / time step) can be added. Such a time dimension enables tracking the temporal response of a processing chamber and / or substrate to processing conditions. The hyperspectral data indicates the spectral signatures of different elements or materials imaged by the hyperspectral image 800. The hyperspectral data of the hyperspectral image 800 is processed to generate metrology data. Exemplary metrology data may include measurements of thickness, non-uniformity, stress, particles, FTIR spectroscopy, absorption, reflectance, and / or fluorescence spectral data for the substrate (or one or more layers of the substrate) or processing chamber at each pixel 804 of the hyperspectral image 800.

[0108] Returning to FIG. 1 , the controller 130 is operably coupled to the substrate heater 108, the flow control hardware 114, the exhaust system 118, the RF power supply 120, and the hyperspectral camera 128. The controller 130 may comprise any suitable computing system, examples of which are described below with reference to FIG. 13 . The controller 130 is configured to control various functions of the processing tool 100 to process substrates. As one example, the controller 130 is configured to operate the substrate heater 108 to heat the substrate 106 to a desired temperature. As another example, the controller 130 is also configured to operate the flow control hardware 114 to flow a selected gas or mixture of gases into the processing chamber 102 at a selected rate. As yet another example, the controller 130 is further configured to operate the exhaust system 118 to remove gases from the processing chamber 102. As yet another example, the controller 130 is further configured to operate the flow control hardware 114 and the exhaust system 118 to control the pressure within the processing chamber 102. As another example, the controller 130 is configured to operate the RF power source 120 to form the plasma.

[0109] The controller 130 is further configured to control the hyperspectral camera 128 to capture hyperspectral images of the process chamber 102 and / or the substrate 106. In some examples, the hyperspectral camera 128 can capture hyperspectral images using a point-to-point method, a line-scan method, or a snapshot method. In the point-to-point method, the hyperspectral camera 128 is configured to capture hyperspectral data for multiple wavelength bands pixel by pixel. In the line-scan method, the hyperspectral camera 128 is configured to capture hyperspectral data for multiple wavelength bands line by line (e.g., row by row). In the snapshot method, the hyperspectral camera 128 is configured to capture image sub-frames for each of the multiple wavelength bands, one by one. In some examples, the controller 130 controls the hyperspectral camera 128 to capture hyperspectral images during a substrate processing cycle as the substrate is being processed. In some examples, the controller 130 controls the hyperspectral camera 128 to capture a series of hyperspectral images throughout a substrate processing cycle to track the progress of the substrate during processing.

[0110] In some examples, the controller 130 can control the illumination source 129 of the hyperspectral camera 128 to output light that illuminates the substrate 106 or the process chamber 102 during image acquisition. In other examples, the controller 130 can control the hyperspectral camera 128 to acquire images while controlling the RF power source 120 to form a plasma. In such examples, the plasma can provide broadband spectrum light suitable for hyperspectral imaging. Furthermore, in some examples, the controller 130 controls the hyperspectral camera 128 to capture hyperspectral images upon completion of a substrate processing cycle. The controller 130 can control the hyperspectral camera 128 to capture any suitable number of hyperspectral images according to any suitable frame rate during and / or after a substrate processing cycle.

[0111] In some examples, the controller 130 is configured to execute a trained machine learning model 132. The trained machine learning model 132 is configured to receive one or more hyperspectral images from the hyperspectral camera 128 and output metrology data 134 about the process chamber 102 and / or the substrate 106 based on at least the one or more hyperspectral images. The metrology data 134 can characterize various properties of the process chamber 102 and / or the substrate 106. In some examples, the metrology data 134 includes absorption, reflectance, and / or fluorescence spectral data of the substrate 106 (and / or other materials within the process chamber 102). Alternatively or additionally, in some examples, the metrology data 134 includes measurements of stress exerted on the substrate 106. Alternatively or additionally, in some examples, the metrology data 134 includes measurements of the resistivity of the substrate 106. Alternatively or additionally, in some examples, the metrology data 134 includes measurements of the thickness of the substrate 106 and / or individual layers deposited on the substrate 106. Alternatively or additionally, in some examples, the metrology data 134 includes an assessment of non-uniformity of the substrate 106. Alternatively or additionally, in some examples, the metrology data 134 includes an indication of particle detection in the process chamber 102 and / or measurements of the size of particles detected in the process chamber 102. The metrology data 134 generated based on at least the hyperspectral images, in some examples, can measure properties of the substrate 106 and / or the process chamber 102 with higher resolution than other ex-situ metrology analysis methods that are not based on hyperspectral images.

[0112] In some implementations, the trained machine learning model 132 is configured to receive a series of hyperspectral images from the hyperspectral camera 128 over a period of time and output time-based metrology data 134 for the process chamber 102 and / or the substrate 106 based on at least the series of hyperspectral images. In some examples, the series of hyperspectral images is captured during a substrate processing cycle for in-situ analysis and control of the processing tool 100. In some such examples, the series of hyperspectral images is captured during a time period that begins before the start of a substrate processing cycle and ends after the completion of the substrate processing cycle. In other examples, the series of hyperspectral images is captured over a time segment that spans only a portion of a substrate processing cycle. In further examples, the series of hyperspectral images is captured over a longer period of time that encompasses multiple substrate processing cycles.

[0113] The trained machine learning model 132 may be a time-based model trained to analyze changes in metrology data and determine how the process chamber 102 and / or substrate 106 change over time. The time-based metrology data 134 may track changes in any suitable type of measurement over time. As one example, the time-based metrology data 134 may track the growth of a film deposited on the substrate 106 over time. As another example, the time-based metrology data 134 may measure nucleation delay at the start of a process. As a further example, the time-based metrology data 134 may measure the effectiveness of an inhibition process to control conformality. As another example, the time-based metrology data 134 may monitor the progress of an etch process. As a further example, the time-based metrology data 134 may monitor particle contamination of a substrate during processing. As yet another example, the time-based metrology data 134 may monitor the accumulation of material on the surfaces of the process chamber 102. As yet another example, the time-based metrology data 134 may monitor the non-uniformity of a film deposited on the substrate 106 over time. Traditionally, non-uniformity metrics, such as thickness, are performed offline using ellipsometry, XRF, or other methods on a small number of points (10–50). These points are used as locations for mapping thickness, refractive index, sheet resistance, or other properties to determine non-uniformity across a 300 mm wafer. However, using hyperspectral imaging, more detailed mapping can be achieved. For example, depending on the resolution of the hyperspectral camera, it is possible to obtain measurements with sub-mm resolution across a 300 mm wafer using a time-based model. In this way, not only is the evolution of thickness (or other properties) obtained point-by-point, but the evolution of non-uniformity can be obtained at a higher resolution than ex-situ measurements in the final state.

[0114] The trained machine learning model 132 may use any suitable method for processing the time-based measurement data 134. For example, the trained machine learning model 132 may use one or more convolutional neural networks (e.g., spatial and / or temporal convolutional neural networks for processing images and / or videos), recurrent neural networks (e.g., long short-term memory networks), support vector machines, associative memories (e.g., lookup tables, hash tables, Bloom filters, neural Turing machines, and / or neural random access memories), unsupervised spatial and / or clustering methods (e.g., nearest neighbor algorithms, topological data analysis, and / or k-means clustering), linear and / or Gaussian regression modeling, graphical models (e.g., Markov models, conditional random fields, and / or AI knowledge bases), and / or other methods for dimensionality reduction and modeling.

[0115] 9 shows a flow diagram illustrating an example method 900 for training and executing a machine learning model to perform metrology-based analysis on a process chamber and / or a substrate within the process chamber. For example, the method can be implemented to train and execute the trained machine learning model 132 shown in FIG. 1. In some examples, the controller 130 shown in FIG. 1 can implement the method. In other examples, a separate computing system can train the trained machine learning model 132, and the controller 130 can execute the trained machine learning model 132.

[0116] At 902, the method 900 includes receiving raw data for training a machine learning model. In some examples, the raw data includes hyperspectral images of a processing chamber under different conditions / states. For example, when training a machine learning model to monitor a processing chamber cleaning process, such conditions / states may include a clean processing chamber and a processing chamber with different levels of residue buildup after undergoing various processing cycles. In other examples, the raw data includes hyperspectral images of a substrate under different conditions / states for training a machine learning function to monitor substrate processing. For example, such conditions / states may include an unprocessed substrate, a substrate at different points in a process, and a substrate after undergoing different processes. The raw data may include any suitable type of training data for training a machine learning model to output metrology data based on one or more hyperspectral images. In some examples, the raw data includes metadata associated with properties of the hyperspectral camera. Exemplary hyperspectral camera properties include intrinsic and extrinsic properties of the hyperspectral camera. Exemplary intrinsic camera properties may include focal length, principal point, pixel dimensions, and pixel resolution, among other properties. Exemplary external camera properties may include, among other properties, the position and orientation of the camera in world space. In some examples, the raw data includes metadata related to the operation of the process chamber. Exemplary process chamber operating properties include information such as one or more process gases in the process chamber, the flow rate of each of the one or more process gases, the overall chamber pressure, the plasma power level, the plasma frequency, and the substrate temperature. By taking such metadata into account, the machine learning model can be updated / retrained based on hardware and / or process changes, making it more robust and accurate under different operating conditions compared to other machine learning models that have not been updated / retrained.

[0117] At 904, the method 900 includes preprocessing the raw data by filtering out data not required for training the machine learning model. In some examples, the filtered data includes overlapping hyperspectral images. In some examples, the filtered data includes hyperspectral data in wavelength bands not of interest. For example, if a machine learning model is trained to process a particular membrane that only responds to a specific wavelength band, hyperspectral data corresponding to other wavelength bands to which the membrane does not respond can be filtered out from processing. In other examples, the raw data is preprocessed using normalization and dimensionality reduction techniques, such as principal component analysis. The preprocessing step can optionally be performed to reduce the overall time to train the machine learning model.

[0118] At 906, method 900 includes training / developing a machine learning model. The machine learning model can be trained / developed according to any suitable training procedure. Non-limiting examples of training procedures for machine learning models include supervised training (e.g., using gradient descent or any other suitable optimization method), zero-shot learning, few-shot learning, and unsupervised learning (e.g., classification based on classes derived from unsupervised clustering methods), and reinforcement learning (e.g., feedback-based deep Q-learning). In some examples, training can be performed by backpropagation using an appropriate loss function. Exemplary loss functions that can be used for training include mean absolute error, mean squared error, cross-entropy, Huber loss, or other loss functions.

[0119] In some examples, the machine learning model may be trained through supervised training using labeled training data including a set of images with the same structure as the input image. In other words, the training data includes the same type of hyperspectral image captured by the hyperspectral camera that is provided as input to the trained machine learning model. For example, raw data or pre-processing data of substrates and / or processing chambers under different processing conditions.

[0120] At 908, the method 900 includes executing the machine learning model to perform metrology-based analysis on the processing chamber of the processing tool and / or the substrate within the processing chamber. In particular, the machine learning model receives as input one or more hyperspectral images of the processing chamber and / or the substrate and outputs metrology data based on the one or more hyperspectral images.

[0121] In some examples, metrology data representing one or more observable properties of the processing chamber and / or substrate can be used to calibrate and validate the hyperspectral metrology machine learning model. As an example, a film thickness on a substrate can be observed to determine whether the deposition process is operating within specifications based on the control proposed by the trained machine learning model. If the film thickness is within specifications, the controller can verify that the trained machine learning model is operating properly. Alternatively, if the film thickness is out of specifications, the trained machine learning model can be adjusted / recalibrated and the controls of the deposition process can be adjusted so that the film thickness is within specifications. The metrology data can be used to validate and / or calibrate the trained machine learning model in any suitable manner.

[0122] Returning to FIG. 1 , the controller 130 is configured to adjust the control of the processing tool 100 based on at least the metrology data 134 output by the trained machine learning model 132. In some examples, the trained machine learning model 132 is configured to output recommended control adjustments based on the metrology data 134. In other examples, separate trained machine learning models may be configured to recommend particular control adjustments based on at least the metrology data 134. In yet other examples, the controller 130 may include separate logic configured to adjust the control of the processing tool 100 based on at least the metrology data 134. In yet other examples, the controller 130 is configured to visually present the metrology data 134 to a human operator via a display, and the controller 130 is configured to adjust the operation of the processing tool 100 based on at least user input received from the human operator.

[0123] In some examples, the controller 130 is configured to adjust operation of the processing tool 100 based on metrology data 134 about the processing chamber 102 itself. The controller 130 can be configured to adjust any suitable control parameter of any suitable process performed by the processing tool 100 based on metrology data 134 about the processing substrate.

[0124] As described above, in some examples, the controller 130 can be configured to adjust control parameters of a cleaning process for cleaning the processing chamber 102 based on at least the metrology data 134 for the processing chamber 102. In one example, the metrology data 134 for the processing chamber 102 can indicate an amount of material accumulated inside the processing chamber 102, and the controller 130 can be configured to determine whether the amount of material accumulated inside the processing chamber 102 exceeds a threshold amount. If the amount of material exceeds the threshold amount, the controller 130 initiates the cleaning process. Additionally or alternatively, the controller 130 can monitor the progress and detect an endpoint of the cleaning process based on the amount of material accumulated inside the processing chamber 102. By intelligently controlling the cleaning process based on the metrology data 134 for the processing chamber 102, cleaning of the processing chamber can be performed more efficiently and as needed. This can reduce tool maintenance time compared to cleaning processes that are performed at a fixed frequency or for a fixed length / extent.

[0125] Alternatively or additionally, in some examples, the controller 130 may be configured to perform in-situ analysis of metrology data collected during a substrate processing cycle and adjust the control of the processing tool in real time during the substrate processing cycle.

[0126] In some such examples, the controller 130 can be configured to monitor particle contamination on the substrate surface or within the processing chamber during processing based at least on the metrology data 134. Such in-situ analysis enables intelligent scheduling of other inspection operations. This can help limit the number of substrates scanned with a light scattering tool for particle detection. This can also help limit the areas of the substrate on which an analysis, such as an energy dispersive X-ray (EDX) analysis, is performed to determine particle composition for troubleshooting. As another example, the controller 130 can be configured to perform in-situ analysis of time-based metrology data for the substrate collected during a substrate processing cycle. The controller can be further configured to adjust control of the processing tool in real time during the substrate processing cycle. The controller 130 can be configured to adjust any suitable control parameters of any suitable substrate process in real time based on the in-situ analysis of the time-based metrology data.

[0127] Alternatively or additionally, in some such examples, the controller 130 can be configured to track film thickness based at least on time-based metrology data during the film deposition process. The controller 130 can be further configured to regulate the deposition process and control the deposition rate. As a more specific example, the controller 130 can be configured to allow a high deposition growth rate until a first threshold thickness is detected. The controller can be further configured to adjust process conditions to decrease the deposition rate until a desired final thickness is achieved.

[0128] Alternatively or additionally, in some such examples, in processes utilizing inhibitors, the controller 130 can be configured to track the effectiveness of the inhibition process during the inhibition process based at least on time-based metrology data and dynamically adjust the inhibition time and / or number of inhibition cycles based on the effectiveness derived from the time-based metrology data. This can help ensure that film growth is adequately inhibited according to the desired process. As an example, an inhibition ALD process can be performed by first depositing the inhibitor on the feature such that a high concentration of the inhibitor is deposited on the substrate surface and a low concentration of the inhibitor is deposited in the substrate recesses. ALD can then be used to deposit a film such that the final film is thick in the substrate recesses and thin or completely inhibited on the substrate surface. In such examples, hyperspectral imaging can be performed to monitor inhibitor adsorption on the substrate surface. This can allow inhibitor deposition to continue until inhibitor adsorption reaches a desired level. The hyperspectral camera can also be used to monitor film growth on the substrate surface. This can be used to determine whether the inhibitor is effectively inhibiting film growth or whether additional inhibitor deposition cycles are required.

[0129] Alternatively or additionally, in some examples, the controller 130 may be configured to perform ex-situ in-line analysis of metrology data for the substrates and adjust control of the processing tool 100 between substrate processing cycles. The controller 130 may be configured to adjust any suitable control parameters of any suitable substrate process on a run-by-run basis based on the ex-situ in-line analysis of metrology data.

[0130] Ex-situ inline measurements can be performed in a variety of different ways. In some examples, a processing tool can include a separate module for hyperspectral imaging of substrates. As used herein, a "separate module" generally refers to a space within a processing tool that is separate from one or more processing chambers of the processing tool and into which substrates can be moved for hyperspectral imaging by a substrate handling system. In other examples, ex-situ inline measurements can be performed while substrates are being transferred into or out of processing chambers of a processing tool. For example, the hyperspectral camera 128 can be positioned outside a slit valve through which substrates are transferred between processing stations, and the substrate can be imaged by the hyperspectral camera 128 as the substrate passes through or exits the slit valve. In other examples, substrates can be imaged for ex-situ inline hyperspectral image-based metrology analysis when they are in a transfer module, a load lock module, a front-opening unified pod (FOUP), or an equipment front-end module (EFEM). In yet another example, a hyperspectral camera with an illumination source (e.g., tungsten quartz, xenon, LED set 400 nm to 1000 nm) can be mounted on a vacuum transfer arm that moves across the substrate, and the hyperspectral camera can thus function as a line scan camera that images the substrate as the vacuum transfer arm moves relative to the substrate.

[0131] In some such examples, the controller 130 can be configured to compare parameter values ​​of interest from the metrology data 134 for the substrate with expected / ideal parameter values. For example, the thickness of the deposited film after a deposition process cycle is completed can be measured via hyperspectral imaging and compared to a target thickness. If the measured thickness deviates from the target thickness by more than a threshold amount, the controller 130 can be configured to adjust the control parameters (e.g., power, pressure, gas flow parameters) for a subsequent substrate processing cycle for a different substrate such that the accuracy of the subsequent substrate processing cycle for a different substrate is improved compared to the previous substrate processing cycle.

[0132] Alternatively or additionally, in some such examples, if the controller 130 determines, based at least on the analysis of the metrology data 134, that the thickness of the film deposited on the substrate is outside of a threshold measurement of uniformity, the controller 130 can be configured to implement an automatic correction in the process control (e.g., modifying the process gap, modifying the spin index / index operation). In yet another example, the controller 130 is configured to trigger an alert for manual correction by a human engineer based on determining that the thickness of the film deposited on the substrate is highly non-uniform. For example, the controller 130 can trigger implementation of a showerhead-pedestal leveling process.

[0133] In some examples, the trained machine learning model 132 is configured to generate recommendations for control adjustments to be made by a human operator based on the measurement data 134.

[0134] 10 is a flow diagram illustrating an example method 1000 for using hyperspectral images to perform metrology-based analysis and control a cleaning process for a processing chamber. For example, method 1000 can be performed by controller 130 of FIG.

[0135] At 1002, method 1000 includes receiving one or more hyperspectral images of a processing chamber of a processing tool from a hyperspectral camera. At 1004, method 1000 includes sending the one or more hyperspectral images to a trained machine learning model configured to output metrology data for the processing chamber based on at least the one or more hyperspectral images. At 1006, method 1000 includes adjusting one or more control parameters of a process performed by the processing tool based on at least the metrology data for the processing chamber. In some embodiments, at 1008, method 1000 optionally includes adjusting one or more control parameters of a cleaning process during cleaning of the processing chamber. In some examples, the frequency at which the cleaning process is performed and / or the length / extent of the cleaning process are adjusted based on the amount of material accumulation on the processing chamber indicated by the metrology data for the processing chamber. Additionally or alternatively, in some examples, cleaning pressure, cleaning gas flow rate, and / or cleaning gas timing may be adjusted based on an analysis of the metrology data.

[0136] Alternatively or additionally, in some embodiments, the method 1000 may optionally include adjusting one or more control parameters of the inspection process and inspecting the process chamber at 1010. In one example, the frequency at which the inspection process is adjusted based on the detection of particles in the processing chamber as indicated by metrology data for the processing chamber.

[0137] Method 1000 can be implemented to intelligently control operation of a processing tool based on feedback provided by metrology data about the processing chamber. Such intelligent operation can include performing cleaning and / or inspection operations only when necessary as determined by the feedback. Such intelligent operation can increase the efficiency and throughput of the processing tool compared to processing tools that perform such operations without feedback. Method 1000 can be repeatedly implemented for any suitable number of process and / or processing cycles.

[0138] 11 shows a flow diagram illustrating an exemplary method 1100 for performing metrology-based analysis using hyperspectral images for in-situ control of a processing tool during a substrate processing cycle. For example, method 1100 can be performed by controller 130 shown in FIG.

[0139] At 1102, method 1100 includes receiving a series of hyperspectral images of a substrate in a processing chamber of a processing tool from a hyperspectral camera during a substrate processing cycle. At 1104, method 1100 includes sending the series of hyperspectral images during the substrate processing cycle to a trained machine learning model configured to output time-based metrology data for the substrate based at least on the series of hyperspectral images. At 1106, method 1100 includes adjusting one or more control parameters of a process of the substrate processing cycle based on the time-based metrology data for at least the substrate during the substrate processing cycle. Examples of control parameters that can be adjusted include one or more of process time, substrate temperature, showerhead temperature (if the showerhead has a heater), showerhead-to-pedestal spacing, total process pressure, partial pressure of each of one or more process gases, and radio frequency power. Method 1100 can thereby provide in-situ metrology-based analysis using hyperspectral images, enabling real-time adjustment and control of the processing tool. In some implementations, in-situ metrology-based analytical / metrology data can optionally be tracked across different processing cycles for multiple substrates to adjust the control of a particular process. For example, at a given process step (in a given iteration), a particular statistic / characteristic can be tracked for each of multiple substrates to determine if a drift / shift has occurred that can be corrected by adjusting the process. Method 1100 can be performed repeatedly across any suitable number of process and / or processing cycles.

[0140] FIG. 12 shows a flow diagram illustrating an exemplary method 1200 for performing metrology-based analysis using hyperspectral images for ex-situ inline control of a processing tool during substrate processing cycles. For example, method 1200 can be performed by controller 130 shown in FIG. 1. At 1202, method 1200 includes receiving one or more hyperspectral images of a first substrate at a processing tool during or after a first substrate processing cycle from a hyperspectral camera. Note that the substrate may be imaged by the hyperspectral camera within any suitable processing module of the processing tool or while being transferred between different processing modules of the processing tool for ex-situ inline metrology-based analysis. At 1204, method 1200 includes sending the one or more hyperspectral images to a trained machine learning model configured to output metrology data for the first substrate based on at least the one or more hyperspectral images. At 1206, method 1200 includes, during a second substrate processing cycle for a second substrate, adjusting one or more control parameters of the process for the second substrate processing cycle based on the metrology data for at least the first substrate. Method 1200 can be implemented to provide ex-situ in-line metrology-based analysis using hyperspectral images, thereby enabling run-by-run adjustment and control of a processing tool between substrate processing cycles. Furthermore, method 1200 can be implemented to provide ex-situ in-line metrology-based analysis performed over multiple processing cycles of multiple different substrates, such as to correct for operational drift / shifts over time.

[0141] In some embodiments, the methods and processes described herein may be coupled to the computing systems of one or more computing devices. In particular, such methods and processes may be implemented as computer application programs or services, application programming interfaces (APIs), libraries, and / or other computer program products.

[0142] 13 illustrates generally a non-limiting embodiment of a computing system 1300 capable of performing one or more of the methods and processes described above. The computing system 1300 is shown in simplified form. The computing system 1300 can take the form of one or more personal computers, workstations, computers integrated with wafer processing tools, and / or network-accessible server computers.

[0143] Computing system 1300 includes a logic machine 1302 and a storage machine 1304. Computing system 1300 may optionally include a display subsystem 1306, an input subsystem 1308, a communication subsystem 1310, and / or other components not shown in Figure 13. Controller 130 is an example of a computing system 1300.

[0144] The logical machine 1302 includes one or more physical devices configured to execute instructions. For example, the logical machine may be configured to execute instructions that are part of one or more applications, services, programs, routines, libraries, objects, components, data structures, or other logical constructs. Such instructions may be implemented to perform a task, implement a data type, transform the state of one or more components, achieve a technical effect, or otherwise arrive at a desired result.

[0145] A logical machine may include one or more processors configured to execute software instructions. Additionally or alternatively, a logical machine may include one or more hardware or firmware logical machines configured to execute hardware or firmware instructions. The processors of a logical machine may be single-core or multi-core, and the instructions executed thereon may be configured for sequential, parallel, and / or distributed processing. Individual components of a logical machine may optionally be distributed across two or more separate devices that may be remotely located and / or configured for cooperative processing. Aspects of a logical machine may be virtualized and executed by remotely accessible networked computing devices configured in a cloud computing configuration.

[0146] The storage machine 1304 includes one or more physical devices configured to hold instructions 1312 executable by a logical machine to implement the methods and processes described herein. When such methods and processes are implemented, the state of the storage machine 1304 can be transformed, for example, to hold different data.

[0147] The storage machine 1304 may include removable and / or internal devices. The storage machine 1304 may include, among other things, optical memory (e.g., CDs, DVDs, HD-DVDs, Blu-ray discs, etc.), semiconductor memory (e.g., RAM, EPROM, EEPROM, etc.), and / or magnetic memory (e.g., hard disk drives, floppy disk drives, tape drives, MRAM, etc.). The storage machine 1304 may include volatile, non-volatile, dynamic, static, read / write, read-only, random-access, sequential-access, position-addressable, file-addressable, and / or content-addressable devices.

[0148] It will be understood that the storage machine 1304 includes one or more physical devices, however, aspects of the instructions described herein may alternatively be propagated by a communication medium (e.g., electromagnetic signals, optical signals, etc.) that is not retained for a finite period by a physical device.

[0149] Aspects of the logic machine 1302 and the storage machine 1304 may be integrated into one or more hardware logic components, which may include, for example, field programmable gate arrays (FPGAs), program and application specific integrated circuits (PASICs / ASICs), program and application specific standard products (PSSPs / ASSPs), systems on a chip (SOCs), and complex programmable logic devices (CPLDs).

[0150] If included, the display subsystem 1306 can be used to present a visual representation of the data maintained by the storage machine 1304. This visual representation can take the form of a graphical user interface (GUI). As the methods and processes described herein modify the data maintained by the storage machine, thereby transforming the state of the storage machine, the state of the display subsystem 1306 can be similarly transformed to visually represent the changes in the underlying data. The display subsystem 1306 can include one or more display devices utilizing virtually any type of technology. Such display devices can be combined with the logic machine 1302 and / or the storage machine 1304 within a common enclosure, or such display devices can be peripheral display devices.

[0151] If included, the input subsystem 1308 may comprise or interface with one or more user input devices, such as a keyboard, mouse, or touchscreen. In some embodiments, the input subsystem may comprise or interface with selected natural user input (NUI) components. Such components may be integrated or peripheral, and translation and / or processing of input actions may be handled on-board or off-board. Exemplary NUI components may include microphones for speech and / or voice recognition, and infrared, color, stereo, and / or depth cameras for machine vision and / or gesture recognition.

[0152] If included, the communications subsystem 1310 can be configured to communicatively couple the computing system 1300 with one or more other computing devices. The communications subsystem 1310 can include wired and / or wireless communications devices compatible with one or more different communications protocols. As a non-limiting example, the communications subsystem can be configured to communicate over a wireless telephone network, or a wired or wireless local or wide area network. In some embodiments, the communications subsystem can enable the computing system 1300 to send and receive messages to and from other devices over a network, such as the Internet.

[0153] 1 , the controller 130 is configured to execute the trained machine learning model 132. The trained machine learning model 132 is configured to receive one or more hyperspectral images from the hyperspectral camera 128 and to output metrology data 134 about the process chamber 102 and / or the substrate 106 based on at least the one or more hyperspectral images. The metrology data 134 may characterize various properties of the process chamber 102 and / or the substrate 106.

[0154] In some implementations, the machine learning model 132 is trained to predict / identify various properties of the substrate 106 and / or other materials (e.g., gases) within the process chamber 102 based on at least one of the spectral signatures generated in the hyperspectral image output from the hyperspectral camera 128. Different spectral signatures are generated from different gases and substrates because different gases and substrates transmit and reflect different wavelengths and intensities of light. Furthermore, variations in the spectral signatures captured in the hyperspectral image can be caused by variations in the gas / plasma density, gas / plasma composition (e.g., different types of gases transmit and reflect different wavelengths and intensities, thereby generating different spectral signatures captured in the hyperspectral image), the gas / plasma flow path during processing, the composition of the substrate 106, the thickness of the substrate 106, and / or the density of the substrate 106. To identify these properties, the hyperspectral camera 128 takes multiple hyperspectral images "in-situ" while the substrate is within the processing tool, and the machine learning model 132 is trained to predict / identify these properties based on an analysis of at least the multiple hyperspectral images. The machine learning model 132 can be trained to predict / identify any or all of these properties based on an analysis of the multiple hyperspectral images of at least the process chamber 102 and / or the substrate 106. The machine learning model 132 can be trained to identify any suitable properties of the process chamber 102, the substrate 106, and / or other materials within the process chamber 102 based on an analysis of the spectral signatures corresponding to these different elements captured in at least the multiple hyperspectral images.

[0155] In some implementations, the controller 130 is configured to execute multiple machine learning models, each trained to predict / identify different properties of the process chamber 102, the substrate 106, and / or other materials within the process chamber 102, based on analysis of at least the multiple hyperspectral images. The controller 130 can execute multiple machine learning models simultaneously to analyze the multiple hyperspectral images and predict / identify different properties of the process chamber 102, the substrate 106, and / or other materials within the process chamber 102.

[0156] In some implementations, the hyperspectral camera 128 is configured to capture a series of hyperspectral images of the substrate 106 in situ during a process being performed on the substrate 106 to determine whether the process is performing properly according to specifications or within specified tolerance levels. In one example, the machine learning model 132 is trained to analyze differences in reflectance spectra across the substrate 106 during the process and determine whether the substrate 106 is within specifications or within specified tolerance levels.

[0157] FIG. 14 schematically illustrates different exemplary states of a substrate 1400 during a process in which a gap 1402 of a feature of the substrate 1400 is being filled by atomic layer deposition (ALD). ALD allows for conformal growth of the film, allowing the film to have a substantially uniform thickness throughout. At 1404, a first state of the gap 1402 is shown, where the gap 1402 is empty. For example, the gap 1402 can be in the first state at the start of the process. At 1406, a second state of the gap 1402 is shown, where the gap 1402 is partially filled. At 1408, a third state of the gap 1402 is shown, where the gap 1402 is more partially filled than in the second state. At 1410, a fourth state of the gap 1402 is shown, where the gap 1402 is completely filled. For example, the gap 1402 can be in the fourth state at the end of the process.

[0158] In some embodiments, the hyperspectral camera 128 is configured to capture hyperspectral images of the substrate 1400 at each of different states 1404-1410 during the process of filling the gap 1402. The reflectance spectrum of the region in the hyperspectral image corresponding to the substrate 1400, and more particularly the gap 1402, is different in each of the different states 1404-1410. The differences in the reflectance spectra in the hyperspectral images allow for the identification of different states of the gap 1402 / substrate 1400 through analysis of the hyperspectral images. In some embodiments, the machine learning model 132 is trained using hyperspectral images including reflectance spectra corresponding to different substrates at different states during the process (e.g., including different fill levels of the gaps on the substrate), such that the trained machine learning model 132 can identify the state of the substrate at any given time during the process based at least on analysis of the hyperspectral images of the substrate captured by the hyperspectral camera 128 during the process.

[0159] In some embodiments, the controller 130 is configured to generate a thickness map of the substrate from multiple different hyperspectral images of the substrate taken at different times during the process. The thickness map provides a visual representation of the film growth (or various other conditions of the substrate) over the course of the process. By training the machine learning model 132 in this manner, the trained machine learning model 132 can determine whether the substrate is experiencing problems during the process and, if so, identify the type of problem.

[0160] FIG. 15 schematically illustrates different exemplary states of a substrate 1500 during a fill process, in which voids form in gaps 1502 of a feature of the substrate 1500. At 1504, the film 1505 being deposited has tapered sidewalls near the bottom of the gap 1502. This may be due, for example, to the ALD process not saturating the substrate surface in the gap 1502. Continuing at 1506, the tapered sidewalls remain even as the film thickens. At 1510, it can be seen that voids 1511 remain after the gap fill process is complete. The gap 1502 with voids 1511 can have a different reflectance spectrum in a hyperspectral image compared to a gap filled with a void-free film.

[0161] In each of the above examples, a problem (or lack thereof) with the substrate 1500 manifests as a change in the reflectance spectrum in a hyperspectral image of the substrate 1500. In these examples, the machine learning model 132 can identify the problem with the substrate 1500 based on an analysis of hyperspectral images of the substrate 1500 taken at least before, during, and / or after the fill process, and determine the change in the reflectance spectrum of the substrate 1500 before, during, and / or after the fill process. The machine learning model 132 can analyze these hyperspectral images to determine whether the process is performing properly, and the controller 130 can dynamically adjust the process to compensate for any problems identified by the machine learning model 132. Depending on the implementation, such control can be performed in situ during the process or between different batches of the process.

[0162] In some implementations, the location of the hyperspectral camera 128 is configured to be dynamically adjustable to adjust the distance of the hyperspectral camera 128 relative to the scene / object being imaged. Varying the distance of the hyperspectral camera 128 relative to the scene / object being imaged changes the field of view of the hyperspectral camera 128 and, therefore, the physical size of the pixels in the hyperspectral images generated at different distances. This allows for more control over hyperspectral measurements of areas of interest, particularly relatively small areas of interest (such as areas used to determine the degree of haze in a layer of a substrate). Furthermore, different hyperspectral images of a scene / object taken at different distances relative to the imaged scene / object can be compared to each other to distinguish the associated measurement data from noise.

[0163] 16 schematically illustrates an example scenario in which a hyperspectral camera 1600 is configured to be dynamically adjustable to adjust the distance of the hyperspectral camera 1600 relative to a substrate 1602 being imaged. The hyperspectral camera 1600 is located in a transfer module 1604 of a processing tool, such as the processing tool 100 shown in FIG. 1. In other examples, the hyperspectral camera 1600 can be located in a different part of the processing tool, such as a processing chamber module, a load lock module, or a front-end equipment module (EFEM).

[0164] The hyperspectral camera 1600 is located above a slit valve 1606 within a transfer module 1604. A robotic arm 1608 holds and moves the substrate 1602 within the transfer module 1604. The robotic arm 1608 passes the substrate 1602 through the slit valve 1606 as the substrate 1602 is transferred from the transfer module 1604 to a processing module. Additionally, the robotic arm 1608 receives the substrate 1602 from the slit valve 1606 as the substrate 1602 is transferred from a processing module to the transfer module 1604. In the illustrated example, the height of the hyperspectral camera 1600 is adjustable within the transfer module, thereby adjusting the distance between the hyperspectral camera 1600 and the substrate 1602. The hyperspectral camera 1600 can capture one or more hyperspectral images of the substrate 1602 from different distances as the substrate enters or exits the slit valve 1606.

[0165] In one example, at a first time (T1), the hyperspectral camera 1600 is positioned at a first height (H1) within the transfer module 1604, such that the hyperspectral camera 1600 is a first distance (D1) from the substrate 1602. At the first distance (D1), the substrate 1602 is positioned entirely within the field of view 1610 of the hyperspectral camera 1600. The hyperspectral camera 1600 captures one or more hyperspectral images of the substrate 1602 from this first position. In some examples, the hyperspectral camera 1600 is a line scan camera that scans the substrate 1602 as it passes under the hyperspectral camera 1600 and enters the slit valve 1606. In other examples, the hyperspectral camera 1600 is configured to capture a snapshot of the entire substrate 1602 at a specific index location before the substrate enters the slit valve 1606.

[0166] At a second time (T2), the hyperspectral camera 1600 is dynamically adjusted relative to the substrate 1602. In particular, the hyperspectral camera 1600 is lowered in the transfer module 1604 to a second height (H2), such that the hyperspectral camera 1600 is at a second distance (D1) from the substrate 1602 that is closer than the first distance (D1). At the second distance (D2), only a portion of the substrate 1600 is positioned within the field of view 1610 of the hyperspectral camera 1600. The hyperspectral camera 1600 captures one or more hyperspectral images of the substrate 1602 from this second position. Because the hyperspectral camera 1600 is positioned closer to the substrate 1602 at the second position than at the first position, the pixels of the hyperspectral image captured by the hyperspectral camera 1600 at the second position correspond to a smaller or more granular region of the substrate 1602 compared to the pixels of the hyperspectral image captured when the hyperspectral camera 1600 was at the first position.

[0167] Alternatively or additionally, in some implementations, the hyperspectral camera 1600 may include one or more optical components (e.g., a zoom lens) configured to optically adjust the distance between the image sensor of the hyperspectral camera 1600 and the scene / object being imaged (e.g., the substrate 1602). The one or more optical components may be dynamically adjusted to adjust the distance of the hyperspectral camera 128 relative to the scene / object being imaged.

[0168] Alternatively or additionally, in some embodiments, the substrate 1602 can be moved by a robotic arm 1708 relative to the position of the hyperspectral camera 1600 to dynamically adjust the distance between the hyperspectral camera 1600 and the substrate 1602.

[0169] In some implementations, the machine learning model 132 is trained based on hyperspectral images of a scene (e.g., a process chamber) and / or an object (e.g., a substrate) taken at different distances relative to the hyperspectral camera that captured the hyperspectral images. The trained machine learning model 132 can be configured to receive one or more hyperspectral images of a substrate taken at a first distance relative to the substrate and output metrology data 134 for the substrate based on the one or more hyperspectral images taken at the first distance. Furthermore, the trained machine learning model 132 can be configured to receive one or more hyperspectral images of a substrate taken at a second distance relative to the substrate that is different from the first distance and output metrology data 134 for the substrate based on the one or more hyperspectral images taken at the second distance. For example, the second distance may be smaller than the first distance. The change in distance can be implemented dynamically by physically moving the hyperspectral camera or optically by adjusting optical components of the hyperspectral camera, depending on the implementation.

[0170] In some examples, the metrology data 134 for the substrate may vary depending on different distances of the hyperspectral camera relative to the substrate. In some examples, the hyperspectral camera 1600 is brought closer to the substrate to obtain metrology data 134 for a particular feature or region of interest, such as inspecting one or more gaps or other features on the substrate to be filled. In other examples, the hyperspectral camera is dynamically adjusted to capture a larger portion of the substrate (e.g., the entire substrate) within the hyperspectral camera's field of view, and the machine learning model can output metrology data 134 for the substrate based on the hyperspectral image captured at that distance. In some implementations, the machine learning model 132 is configured to receive hyperspectral images of the substrate captured at different distances, compare the reflectance spectra of the different hyperspectral images to distinguish actual spectral information from noise, and output denoised metrology data 134 for the substrate.

[0171] In some implementations, the hyperspectral camera 128 is configured to be dynamically adjustable to adjust the angle of the hyperspectral camera 128 relative to the scene / object being imaged. The angle of incidence of light emitted from the hyperspectral camera 128 relative to the scene / object being imaged can change how the light interacts with the surface of the scene / object and affect the reflectance spectrum. Depending on the material being imaged by the hyperspectral camera 128, some angles of incidence may be more optimal than others for the accuracy of the predicted output. In some examples, a selected set of angles may be optimized for a particular material.

[0172] 17 schematically illustrates an exemplary scenario in which a hyperspectral camera 1700 is configured to be dynamically adjustable to adjust the angle of incidence of light emitted from the hyperspectral camera 1700 relative to different substrates 1702, 1702′ being imaged. The hyperspectral camera 1700 is located within a transfer module 1704 of a processing tool, such as the processing tool 100 shown in FIG. 1. In other examples, the hyperspectral camera 1700 can be located in a different part of the processing tool, such as a processing chamber module, a load lock module, or a front-end equipment module (EFEM).

[0173] The hyperspectral camera 1700 is positioned above a slit valve 1706 within a transfer module 1704. A robotic arm 1708 holds and moves the substrates 1702, 1702′ within the transfer module 1704. The robotic arm 1708 passes the substrates 1702, 1702′ through the slit valve 1706 as they are transferred from the transfer module 1704 to a processing module. Additionally, the robotic arm 1708 receives the substrates 1702, 1702′ from the slit valve 1706 as they are transferred from a processing module to the transfer module 1704. In the illustrated example, the angle of the hyperspectral camera 1700 is adjustable within the transfer module, thereby adjusting the angle of incidence of light emitted from the hyperspectral camera 1700 on the substrate being imaged. The hyperspectral camera 1700 can take one or more hyperspectral images of the substrate 1702 , 1702 ′ from different angles of incidence as the substrate 1702 , 1702 ′ enters and exits the slit valve 1706 .

[0174] In one example, at a first time (T1), the hyperspectral camera 1700 is positioned at a first angle (θ1) relative to a first substrate 1702 having a surface film including a first material. For example, the first angle (θ1) may be selected based on at least how the first material of the surface film is optimized to respond to different wavelengths of light at a selected angle of incidence. The hyperspectral camera 1700 captures one or more hyperspectral images of the substrate 1702 from this first angle (θ1). In some examples, the hyperspectral camera 1700 is a line scan camera that scans the substrate 1702 as it passes under the hyperspectral camera 1700 and enters the slit valve 1706. In other examples, the hyperspectral camera 1700 is configured to capture a snapshot of the entire substrate 1702 at a specific index location before the substrate 1702 enters the slit valve 1706.

[0175] At a second time (T2), the hyperspectral camera 1700 is dynamically adjusted to a second angle (θ2) relative to a second substrate 1702 having a surface film including a second material different from the first material of the surface film of the first substrate 1702. For example, the second angle (θ2) may be selected based on optimizing at least how the second material of the surface film responds to different wavelengths of light at a selected angle of incidence. The hyperspectral camera 1700 captures one or more hyperspectral images of the second substrate 1702′ from this second angle (θ2).

[0176] Alternatively or additionally, in some embodiments, the substrate 1702, 1702' can be moved by a robotic arm 1708 relative to the position of the hyperspectral camera 1700 to dynamically adjust the angle between the hyperspectral camera 1700 and the substrate 1702, 1702'.

[0177] FIG. 18 shows an exemplary graph 1800 of multiple plots of the spectral reflectance of a material on a substrate as a function of wavelength. Each of the multiple plots corresponds to a different angle of incidence of light reflected from the substrate and collected by a hyperspectral camera. The multiple plots can be generated from a hyperspectral image of the substrate. A first plot 1802 corresponds to light having a first angle of incidence (θ1) with respect to the material on the substrate. A second plot 1804 corresponds to light having a second angle of incidence (θ2) with respect to the material on the surface of the substrate. In this example, the second angle of incidence (θ2) is greater than the first angle of incidence (θ1). A third plot 1806 corresponds to light having a third angle of incidence (θ3) with respect to the material on the surface of the substrate. In this example, the third angle of incidence (θ3) is greater than the second angle of incidence (θ2). Note that the angle of incidence changes how light interacts with the material on the surface of the substrate, thus affecting the reflectance spectrum. In other words, the spectral reflectance of a material is different for different angles of incidence at different wavelengths. Furthermore, note that the variation in spectral reflectance between different angles of incidence varies non-uniformly at different wavelengths. The periodic nature of plots 1802, 1804, and 1806 is caused by interference patterns produced by light passing through the material on the substrate.

[0178] In some implementations, the machine learning model 132 is trained based on hyperspectral images of a scene (e.g., a process chamber) and / or an object (e.g., a substrate) captured at least at different angles relative to the hyperspectral camera that captured the hyperspectral images. In some examples, the angle of the hyperspectral camera for the training hyperspectral images is selected based at least on the material of the film on the substrate being imaged. By labeling the training hyperspectral images with the angle of incidence of the hyperspectral camera 128, the machine learning model 132 can be trained to make accurate predictions regarding the spectral reflectance of the material on the substrate or other metrology data for the imaged object. The trained machine learning model 132 can be configured to receive one or more hyperspectral images of a substrate having a film including a first material captured at a first angle selected based at least on the first material of the film, and to output metrology data 134 for the first substrate based at least on the one or more hyperspectral images captured at the selected first angle. Additionally, the trained machine learning model 132 can be configured to receive one or more hyperspectral images of a second substrate having a film including a second material different from the first material taken at a second angle selected based on at least the second material of the film, and to output metrology data 134 for the second substrate based on the one or more hyperspectral images taken at at least the second angle. In some examples, the metrology data 132 can include the spectral reflectance of the material as a function of wavelength at different angles of incidence of the hyperspectral camera 128.

[0179] In some implementations, the machine learning model 132 can be trained to distinguish between differences in the number of layers of material on a substrate and / or differences in the thickness of one or more layers of material on a substrate based on an analysis of at least a hyperspectral image of the substrate. FIG. 19 shows an example graph 1900 of two plots of the spectral reflectance of two different substrates as a function of wavelength. The plots can be generated from a hyperspectral image of the substrate. The two substrates have the same overall thickness but different numbers of layers. A first plot 1902 represents the spectral reflectance of a first substrate having a first number of layers (N1). A second plot 1902 represents the spectral reflectance of a second substrate having a second number of layers (N2). In this example, the second number of layers (N2) is greater than the first number of layers (N1). Note that the first plot 1902, corresponding to the first substate, generally has a greater spectral reflectance than the second plot 1904, corresponding to the second substrate. The difference in spectral reflectance may be due to the layers of the first substrate being thicker than the layers of the second substrate. This information may be applied to training the machine learning model 132. In particular, the machine learning model 132 may be trained using hyperspectral images of different substrates, each having a different number of layers and layer thicknesses. The trained machine learning model 132 may be configured to identify the number of layers on a substrate and / or the thickness of the layers on a substrate based on an analysis of at least the hyperspectral images of the substrates.

[0180] Stress and warpage are metrics that can be used for process control to assess whether a process is performing as expected on a substrate and to monitor the tool health of the processing tool 100. In some embodiments, the processing tool 100 can be configured to scan and measure the warpage and / or stress of the substrate using the hyperspectral camera 128. Indications of stress and warpage of the substrate can appear at least as a change in curvature, where some pixels in a hyperspectral image of the substrate receive more light than other pixels. Furthermore, indicators of stress and warpage can appear at least as a shift in the reflectance spectrum, which may appear as a simple right or left shift in a spectrum without fringes, or as a change in compressional waves in a spectrum with multiple fringes. Stress and / or warpage can be measured locally at multiple different points across the substrate via hyperspectral imaging, and an overall measurement of the stress and / or warpage of the substrate can be determined based on at least the multiple local measurements of stress and / or warpage.

[0181] 20-22 schematically illustrate exemplary configurations in which a hyperspectral camera may be used to scan and measure stress and / or warpage of a substrate in a processing tool, such as the processing tool 100 shown in FIG. 1. In some implementations, the hyperspectral camera may be configured to perform in-situ scanning and measurement of stress and / or warpage of a substrate during processing in a processing chamber of the processing tool. In other examples, the hyperspectral camera may be configured to perform ex-situ or in-line scanning and measurement of stress and / or warpage of a substrate while the substrate is in a transfer module, a load lock module, or an equipment front-end module (EFEM).

[0182] Figures 20-21 show a schematic diagram of an exemplary configuration in which a hyperspectral camera can be used to simulate a spectroscopic ellipsometer to scan and measure the stress and / or warpage of a substrate. The measurement is based on analyzing changes in the polarization state of reflected light from the substrate. Ellipsometer measurements provide information about the complex refractive index and thickness of a film on the substrate. Changes in surface curvature can alter the complex refractive index, and by changing the polarization of the received light, changes in stress / warpage can be captured as changes in ellipsometer parameters, such as psi (Ψ), delta (Δ), angle of incidence (Θi), wavelength (λ), and refractive indices (n and k). Psi (Ψ) is the amplitude ratio between the p-polarized and s-polarized components of the reflected light. This is related to the phase shift between the two polarization states. Delta (Δ) is the phase difference between the p-polarized and s-polarized components of the reflected light. This is related to the shift in the polarization ellipse. The angle of incidence (Θi) is the angle at which the light strikes the sample surface. The angle of incidence can affect the sensitivity of thin film properties in ellipsometry measurements. Wavelength (λ) is the wavelength of the incident light. Ellipsometers often operate over a specific wavelength range, and measurements at different wavelengths can be used to extract more information about the sample. Refractive Index (n and k) is the complex refractive index (n+ik) of the thin film on the substrate. The real part (n) and imaginary part (k) relate to the amplitude and absorption of the light, respectively.

[0183] In FIG. 20, hyperspectral camera 2000 includes a light source 2002 and an image sensor 2004. Light source 2002 emits light of different selected wavelengths across the electromagnetic spectrum (e.g., UV, visible, near-IR, IR) directly onto substrate 2006. Image sensor 2004 captures hyperspectral images of the different wavelengths of light reflected from substrate 2006 to image sensor 2004. This configuration notes pixel-to-pixel wavelength shifts (compressive shifts) and intensity variations resulting from curvature- and / or stress-induced anisotropic behavior to determine measurements of warpage and / or stress. As an example, a flat surface provides an equal amount of reflected light to each pixel. However, a curved surface creates additional interference between the incident and reflected light within the dielectric film. This results in reflected light with different phases / k at specific wavelengths, varying in amplitude.

[0184] In FIG. 21 , hyperspectral camera 2100 includes light source 2102, image sensor 2104, and polarizer 2106. Light source 2102 emits light of different wavelengths across the electromagnetic spectrum (e.g., UV, visible, near-IR, IR) directly onto substrate 2108. Light reflected from substrate 2108 passes through polarizer 2106 to image sensor 2104. Polarizer 2106 varies the polarization of the light reflected from substrate 2108, allowing variations in birefringence of substrate 2108 to be observed by image sensor 2104 and determining stress and / or warpage of substrate 2108. In particular, image sensor 2104 can observe changes in the reflectivity of S and P light passing through polarizer 2106 due to stress and / or warpage of the substrate. Additionally, the polarizer 2106 can remove components of the reflected light that may interfere with the stress and / or warpage measurements by the hyperspectral camera 2100, amplifying birefringence changes to noise in the signal, thereby enabling more accurate measurements of stress and / or warpage. The hyperspectral camera 2100 can use any suitable type of polarizer to vary the polarization of the reflected light and remove components of the reflected light that are unwanted for the image sensor 2104 of the hyperspectral camera 2100. Examples include rotating polarizers, linear polarizers, elliptical polarizers, and other types of polarizers.

[0185] FIG. 22 schematically illustrates an exemplary configuration in which a hyperspectral camera 2200 can be used to obtain local and global stress and / or warpage measurements via coherent gradient sensing. Coherent gradient sensing involves measuring the surface tilt and slope of a substrate with high precision and extracting information about the surface tilt by analyzing the interference pattern of coherent light by comparing the phase shift between multiple points as the light's optical path changes. In the illustrated configuration, a coherent light source 2202 emits coherent light into a beam splitter 2204. In some embodiments, the coherent light source 2202 includes one or more lasers that generate laser light of different wavelengths. In some embodiments, the coherent light source 2202 includes a broadband light source. The coherent light emitted from the coherent light source 2202 is used to generate a distinct interference pattern. The beam splitter 2204 is configured to split the coherent light emitted from the coherent light source 2202 into two beams. One of the split beams serves as a reference beam, while the other serves as a test beam that interacts with the substrate 2206. The test beam illuminates the surface of the substrate 2206, and the reflected light interacts with surface features of the substrate 2206. The reflected test beam interferes with the reference beam, forming an interference pattern. The interference pattern is sensitive to phase changes induced by surface gradients on the substrate 2206 and is indicative of stress and / or warpage. The hyperspectral camera 2200 can optionally include a pattern filter 2208 configured to filter out unwanted light and tailor the characteristics of the light incident on the image sensor of the hyperspectral camera 2200 to the requirements of the measurement system. The pattern filter 2208 improves the quality and accuracy of the resulting data by removing light that could otherwise increase noise. The choice of filter depends on factors such as the nature of the material being measured, the wavelength range of interest, and the specific requirements of the measurement device. Exemplary types of optical filters that can be used in the hyperspectral camera 2200 may include wavelength-selective filters, spatial filters, polarization filters, and / or frequency filters. The image sensor of the hyperspectral camera 2200 captures the interference pattern.The hyperspectral camera 2200 analyzes the changes in the phase of the interference pattern to extract information about the stress and / or warpage of the substrate 2206. The hyperspectral camera 2200 generates hyperspectral images of the interference pattern at different wavelengths because the phase difference can vary based at least on the different wavelengths. In some examples, the hyperspectral images of the substrate 2206 captured by the hyperspectral camera 2200 can be used to determine a stress coefficient of the substrate. In some examples, the hyperspectral images of the substrate 2206 captured by the hyperspectral camera 2200 can be used to reconstruct a three-dimensional surface profile of the substrate 2206 that is indicative of the stress and / or warpage on the substrate.

[0186] The hyperspectral camera configuration shown in Figures 20-22 and described above can measure the reflectivity of a substrate on a pixel-by-pixel basis and across the entire substrate via hyperspectral images to determine local stress and / or warpage at different pixels as well as global stress and / or warpage across the surface of the substrate. In some embodiments, the hyperspectral camera can perform in-line measurements of reflectivity, which can quickly provide insight into the stress and / or warpage of the wafer, allowing for in-process determination of recipe quality and tool health. Furthermore, in some embodiments, the processing tool 100 can dynamically adjust to correct for issues related to substrate stress and / or warpage. For example, the processing tool 100 can transfer the substrate to a different processing chamber to deposit a film on the backside of the substrate, thereby balancing the stress caused by surface processing.

[0187] In some implementations, the machine learning model 132 is trained based on hyperspectral images of different substrates that are subject to at least different levels of stress and / or warpage. The trained machine learning model 132 can be configured to receive one or more hyperspectral images of the substrate and output metrology data 134 that includes a determination of the amount of stress and / or warpage of the substrate based on at least the one or more hyperspectral images. In some examples, the determination of stress and / or warpage can be localized to different points on the substrate. In other examples, the determination of stress and / or warpage applies across the entire substrate.

[0188] 23-24 show exemplary graphs of spectral reflectance measured at different wavelengths at different points on a substrate. FIG. 23 shows graph 2300 of spectral reflectance measurements taken at a pixel corresponding to the center point of the substrate. Graph 2300 includes a first plot 2302 showing the spectral reflectance at the lowest warpage point (warpage A) on the substrate within the pixel at different wavelengths, and a second plot 2304 showing the spectral reflectance at the highest warpage point (warpage B) on the substrate within the pixel at different wavelengths. Plots 2302 and 2304 collectively show the stress and / or warpage of the substrate measured at the center point of the substrate.

[0189] FIG. 24 shows a graph 2400 of spectral reflectance measurements taken at a pixel corresponding to a region near the edge of a substrate. Graph 2400 includes a first plot 2402 showing the spectral reflectance at the lowest warpage point (warpage A) on the substrate within the pixel at different wavelengths, and a second plot 2404 showing the spectral reflectance at the highest warpage point (warpage B) on the substrate within the pixel at different wavelengths. Plots 2402 and 2404 collectively show the stress and / or warpage of the substrate measured at the edge of the substrate. Comparing graph 2300 of FIG. 23 with graph 2400 of FIG. 24, plot 2404 has a vertical compressive shift and amplitude difference compared to plot 2304, indicating a greater amount of stress and / or warpage at the edge of the substrate compared to the center point of the substrate. When a substrate is curved or warped, a phase shift occurs, changing the k-vector (the angular wave vector of the light reflected from the substrate). Not only does there appear to be a shift with position on the same substrate, but for different warpage values, the shift appears more pronounced, as in the case of plot 2404 compared to plot 2304.

[0190] Haze is a metric that provides insight into the diffuseness of a surface and provides an indication of surface roughness. Furthermore, surface roughness measurements can be used to improve the accuracy of substrate thickness determinations and provide that information by providing a roughness correction factor that is taken into account in the thickness determination. In some embodiments, the processing tool 100 can be configured to measure the haze of a substrate using the hyperspectral camera 128. FIG. 25 shows an exemplary configuration in which a hyperspectral camera 2500 is configured to measure haze on a substrate 2502. In some embodiments, the hyperspectral camera 2500 can be configured to perform in-situ measurement of the haze of a substrate during processing in a processing chamber of the processing tool 100. In other examples, the hyperspectral camera 2500 can be configured to perform ex-situ or in-line measurement of the haze of a substrate when the substrate is in a transfer module, a load lock module, a front-opening unified pod (FOUP), or a front-end of equipment module (EFEM).

[0191] The hyperspectral camera 2500 includes a light source 2504 and an image sensor 2506. The light source 2504 includes multiple spatially separated light emitters. In some examples, the light emitters include broadband light sources. In some examples, the light emitters comprise LEDs configured to emit light at specific wavelengths. The spatially separated light emitters are configured to emit light to different spatially separated regions of the substrate 2502. Each region corresponds to multiple pixels that are sufficiently separated that light from one light emitter only emits light to a single region of the substrate 2502 and not to other regions. In the illustrated example, light emitted from the light emitters of the light source 2504 illuminates a pixel 2508 on the substrate 2502. If the surface were more ideally specular (smooth), the pixels surrounding the illuminated pixel 2508 would be dark. If the surface were not specular, the surrounding pixels would exhibit diffuse reflectance 2510. Diffuse reflectance 2510 can be caused by a variety of factors, including, but not limited to, the presence of particles, crystalline structure, defects, crystalline orientation (resulting in anisotropic dispersion), surface roughness (topographical variations), or any combination thereof. In one example, haze is measured by looking at the ratio of incident light to scattered light. The greater the angular spread from the point of incidence, the more diffuse the substrate's surface. While the surface condition of the substrate affects the scattering of incident light, using hyperspectral imaging and a broadband light source, the instrument can also indicate which wavelengths of light are more or less affected. This can act as a potential correction factor for thickness / stress, as well as indicate the presence of larger defects / contaminants / particles.

[0192] Note that the illustrated example shows a single pixel 2508 on the substrate 2502 being illuminated to measure haze. In other examples, various other pixels spaced across the substrate 2502 can be simultaneously illuminated with different illuminants to measure haze in different regions of the substrate.

[0193] The multiple spatially separated light emitters can be arranged in different arrangements in the light source 2504 in different embodiments. FIGS. 26-27 schematically illustrate exemplary arrangements of spatially separated light emitters in a light source. In FIG. 26, the light source 2600 includes multiple spectral light emitters 2602, multiple near-IR light emitters 2604, and multiple collimated red light emitters 2606. In some examples, the multiple spectral light emitters 2602 include a set of red, green, and blue light emitters (e.g., LEDs). The spectral light emitters 2602 are evenly spaced apart from one another across the light source 2600 in this example. The near-IR light emitters 2604 are evenly spaced apart from one another across the light source 2600 in this example. The collimated red light emitters 2606 are designed for measuring haze and are spaced apart from one another more than the spectral light emitters 2602 and near-IR light emitters 2604. It should be noted that the different types of light emitters can have any suitable spacing on the light source 2600. By placing the collimated red light emitter 2606 farther away than the other light emitters in light source 2600, the light emitted from these emitters can be directed to different areas of the substrate to measure haze without light from the other collimated red light emitters contaminating that area with unintended light. In one example, in this arrangement, the collimated red light emitter can be used to measure haze, and the spectral light emitter 2602 and near-IR light emitter 2604 can be used to perform thickness / stress / warpage measurements, among other operations.

[0194] 27, light source 2700 includes multiple light emitter sets 2702. Each light emitter set 2702 includes one or more spectral light emitters 2704, one or more near-IR light emitters 2706, and one or more collimated red light emitters 2708. Light emitter sets 2702 are spaced apart from other light emitter sets so that light emitter sets 2702 illuminate different areas of the substrate for haze measurements without adding light pollution to other areas. In the illustrated example, all light emitters in the different sets can be used to measure haze in addition to other metrics / behaviors (e.g., thickness, stress, warpage measurements).

[0195] In some embodiments, light of different wavelengths can be selected to illuminate a substrate to measure the substrate's haze. FIG. 28 shows an example graph 2800 of the electromagnetic spectrum, including different ranges of wavelengths that can be used for different operations. In particular, light in a spectral wavelength range (e.g., red, green, blue) 2802 and a near-IR wavelength range 2804 can be designated for use in metrics such as measuring thickness, stress, and warpage. Additionally, wavelength ranges outside the spectral range and near-IR range, such as wavelength range 2806, can be designated for measuring haze. By using wavelength range 2806 to measure haze, the spectral wavelength range and near-IR wavelength range do not pose light pollution because these wavelength ranges are not considered when quantifying the substrate's haze.

[0196] 29 shows an example graph 2900 including a haze measurement 2902 represented by the number of pixels with reflected light and their corresponding intensity. The height of the haze measurement 2902 corresponds to the smoothness of the surface, and the spread of the haze measurement 2902 corresponds to the roughness of the surface. The haze measurement 2902 corresponds to a single light emitter.

[0197] In some embodiments, a processing tool includes multiple hyperspectral cameras positioned at different locations within the processing tool to provide input for control of the processing tool. Figure 30 schematically illustrates an exemplary processing tool 3000 including multiple modules 3002, 3004, and 3006 connected to a vacuum transfer chamber 3008. The vacuum transfer chamber 3008 includes multiple hyperspectral cameras 3010, 3012, and 3014 corresponding to the multiple modules 3002, 3004, and 3006. Substrates can be transferred between the different modules 3002, 3004, and 3006 to perform different processes on the substrate. As the substrate is transferred from one module to another, it passes through the vacuum transfer chamber 3008, and the corresponding hyperspectral camera can capture a hyperspectral image of the substrate. For example, each time a substrate enters or leaves a module, the corresponding hyperspectral camera can capture a hyperspectral image of the substrate to determine how the substrate has been changed by the processes performed by the modules. Additionally, the processing tool 3000 may be configured to control how the substrate is processed based on a hyperspectral image of the substrate and a corresponding analysis performed on the hyperspectral image (e.g., by the machine learning model 132 shown in FIG. 1).

[0198] In one example, a film deposition process is performed for a substate in module 3004. Hyperspectral camera 3012 captures hyperspectral images of the substrate before and after the process is performed. The hyperspectral images are analyzed by machine learning model 132, which determines that the process caused the substrate to warp based on at least the analysis of the hyperspectral images. Processing tool 3000 transfers the substrate to module 3006 and performs backside film deposition on the substrate based at least on the output of machine learning model 132 to compensate for the warpage on the other side of the substrate. Processing tool 3000 may be configured to dynamically adjust controls of processing tool 3000 to perform any suitable process on the substrate based at least on the analysis of the hyperspectral images of the substrate performed by machine learning model 132.

[0199] In some embodiments, a module includes multiple hyperspectral cameras positioned at different locations within the module to provide input for control of a process performed by the module. FIG. 31 schematically illustrates an exemplary module 3100 including multiple hyperspectral cameras. For example, module 3100 may correspond to either the processing tool 3000 shown in FIG. 30 or the processing tool 100 shown in FIG. 1. Module 3100 is configured to perform processes on four different substrates 3102, 3104, 3106, and 3108 at a time. Substrates can be indexed and transferred between four different locations within module 3100 to perform different processes on the different substrates. Module 3100 includes four hyperspectral cameras 3110, 3112, 3114, and 3116 corresponding to the four substrates 3102, 3104, 3106, and 3108. The four hyperspectral cameras 3110, 3112, 3114, 3116 are configured to capture hyperspectral images of the four substrates 3102, 3104, 3106, 3108 before, during, and / or after performing processes on the four substrates 3102, 3104, 3106, 3108. The machine learning model 132 analyzes the hyperspectral images and outputs metrology data for the four substrates 3102, 3104, 3106, 3108. The processing tool dynamically controls the processes performed on the substrates by the module 3100 based on at least the metrology data output from the machine learning model 132.

[0200] In one example, a film deposition process is performed on a substrate 3102 in a first position of the module 3100. The hyperspectral camera 3110 takes a series of hyperspectral images of the substrate 3102 during the process. The machine learning model 132 analyzes the series of hyperspectral images and determines that the amount of film growth on the substrate is less than expected. The processing tool dynamically adjusts the process to increase the film growth rate based at least on the output of the machine learning model 132 to compensate for the determined shortfall in film growth and achieve the expected amount of film growth on the substrate.

[0201] In another example, a film deposition process is performed on a substrate 3102 at a first position in the module 3100. The hyperspectral camera 3110 takes hyperspectral images of the substrate 3102 before and after performing the process on the substrate 3102. The machine learning model 132 analyzes the hyperspectral images taken by the hyperspectral camera 3110 and outputs metrology data indicating that the process performed on the substrate 3102 was improper. The processing tool dynamically adjusts a next process to be performed on the substrate at a second position in the module 3100 based on at least the output of the machine learning model 132. The substrate 3102 is moved to the second position in the module 3100, and the hyperspectral camera 3112 takes hyperspectral images of the substrate 3102 before and after performing the next dynamically adjusted process on the substrate 3102 at the second position. The machine learning model 132 analyzes the hyperspectral images taken by the hyperspectral camera 3112 and outputs metrology data indicating that the process performed on the substrate 3102 was as expected. Thus, the substrate 3102 continues to be processed in the remaining locations within the module 3100.

[0202] The module 3100 may be configured to dynamically adjust a process performed on a substrate based on at least an analysis of the hyperspectral image of the substrate performed by the machine learning model 132. Additionally, the module 3100 may be configured to dynamically adjust future processes to be performed on the substrate based on at least an analysis of the hyperspectral image of the substrate performed by the machine learning model 132.

[0203] FIG. 32 shows a flow diagram illustrating an example method 3200 for dynamically controlling the position of a hyperspectral camera within a processing tool and varying the distance between the hyperspectral camera and a substrate for hyperspectral imaging and analysis. For example, method 3200 can be performed by controller 130 of processing tool 100 of FIG. 1 . At 3202, method 3200 includes receiving one or more hyperspectral images of a substrate within the processing tool from a hyperspectral camera positioned at a first location. At 3204, method 3200 includes sending the one or more hyperspectral images taken at the first distance to a trained machine learning model configured to output metrology data for the substrate based on at least the one or more hyperspectral images and the first distance between the hyperspectral camera and the substrate. At 3206, method 3200 includes dynamically adjusting the position of the hyperspectral camera to a second position that is a second distance from the substrate. Alternatively or additionally, in some implementations, optical components (e.g., a zoom lens) of the hyperspectral camera can be dynamically adjusted to adjust the optical distance between the hyperspectral camera and the substrate. Alternatively or additionally, in some implementations, the substrate can be moved relative to the position of the hyperspectral camera to dynamically adjust the distance between the hyperspectral camera and the substrate. At 3208, method 3200 includes receiving one or more hyperspectral images of the substrate from a hyperspectral camera positioned at a second distance from the substrate. At 3210, method 3200 includes sending the one or more hyperspectral images taken at the second distance to a trained machine learning model configured to output metrology data for the substrate based on at least the one or more hyperspectral images and the second distance between the hyperspectral camera and the substrate.Method 3200 can be implemented to capture hyperspectral images of a substrate at different distances and analyze these images differently with a trained machine learning model to obtain different types of metrology data for the substrate based at least on the distance between the hyperspectral camera and the substrate when the hyperspectral images were captured. For example, the distance between the hyperspectral camera and the substrate can be set to capture hyperspectral images of the entire substrate, providing global metrology data for the entire substrate. Additionally, the distance between the hyperspectral camera and the substrate can be dynamically reduced, causing the hyperspectral camera to capture images of specific features or regions of interest on the substrate, providing local metrology data for the specific features or regions of interest on the substrate.

[0204] FIG. 33 shows a flow diagram illustrating an example method 3300 for dynamically controlling the position of a hyperspectral camera to capture hyperspectral images of different substrates from different angles. For example, method 3300 can be performed by controller 130 of processing tool 100 of FIG. 1 . At 3302, method 3300 includes receiving one or more hyperspectral images of a first substrate in the processing tool from a hyperspectral camera positioned at a first angle relative to the first substrate. At 3304, method 3300 includes sending the one or more hyperspectral images of the first substrate captured at the first angle to a trained machine learning model configured to output metrology data for the first substrate based on at least the one or more hyperspectral images and the first angle between the hyperspectral camera and the first substrate. At 3306, method 3300 includes dynamically adjusting the position of the hyperspectral camera such that the hyperspectral camera is positioned at a second angle relative to a second substrate. Alternatively or additionally, in some implementations, the substrate can be moved relative to the position of the hyperspectral camera to dynamically adjust the angle between the hyperspectral camera and the second substrate. At 3308, the method 3300 includes receiving one or more hyperspectral images of the second substrate from a hyperspectral camera positioned at a second angle relative to the second substrate. At 3310, the method 3300 includes sending the one or more hyperspectral images of the second substrate taken by the hyperspectral camera at the second angle to a trained machine learning model configured to output metrology data for the second substrate based on at least the one or more hyperspectral images and the second angle between the hyperspectral camera and the second substrate.Method 3300 can be implemented to capture hyperspectral images of different substrates at different angles and analyze these images in different ways with a trained machine learning model to obtain different types of metrology data for different substrates based at least on the angle between the hyperspectral camera and the different substrates when the hyperspectral images were captured. For example, different substrates may include films including different materials that reflect incident light from different angles in different ways. In some examples, a particular angle of incident light for a particular material may yield more accurate metrology data compared to other angles. Thus, the angle between the hyperspectral camera and the substrate may be set / dynamically adjusted to capture hyperspectral images based at least on the material of the substrate.

[0205] FIG. 34 shows a flow diagram illustrating an example method 3300 of performing metrology-based analysis using hyperspectral images for control of a processing tool. For example, method 3400 can be performed by controller 130 of processing tool 100 of FIG. 1 . At 3402, method 3400 includes receiving one or more hyperspectral images of a substrate in the processing tool from a hyperspectral camera. At 3404, method 3400 includes sending the one or more hyperspectral images to a trained machine learning model configured to output metrology data for the substrate based on at least the one or more hyperspectral images. In some embodiments, at 3406, the metrology data can include a gas / plasma composition and / or a gas / plasma flow path in a processing chamber containing the substrate. In some embodiments, at 3408, the metrology data can include a thickness and / or density of the substrate. In some embodiments, at 3410, the metrology data can include, and the metrology data can include, the number of layers in the substrate. In some embodiments, at 3412, the metrology data can identify voids in the substrate that were not properly filled during a process performed on the substrate. In some embodiments, at 3414, the metrology data may include an amount of stress and / or warpage in the substrate. In some embodiments, at 3416, the metrology data may include an amount of haze in the substrate. At 3418, the method 3400 includes adjusting one or more control parameters of a process performed by the processing tool based on the metrology data for at least the substrate. The method 3400 may use hyperspectral images to provide metrology-based analysis and enable in-situ or in-line adjustment and control of the processing tool.

[0206] It will be understood that the configurations and / or approaches described herein are exemplary in nature, and that these specific embodiments or examples are not to be considered limiting, as numerous variations are possible. The specific routines or methods described herein may represent one or more of any number of processing strategies. As such, the various operations shown and / or described may be performed in the order shown and / or described, in other orders, in parallel, or omitted. Similarly, the order of the processes described above may be changed.

[0207] The subject matter of the present disclosure includes all novel and non-obvious combinations and subcombinations of the various processes, systems and configurations, and other features, functions, operations and / or properties disclosed herein, and any and all equivalents thereof.

Claims

1. 1. A processing tool comprising: a processing chamber having an optical interface; a hyperspectral camera positioned to capture hyperspectral images of the interior of the processing chamber through the optical interface of the processing chamber; A processing tool comprising:

2. 10. The processing tool of claim 1, A processing tool, wherein the processing chamber comprises a pedestal, and the hyperspectral camera is positioned to capture a hyperspectral image of a substrate positioned on the pedestal through the optical interface.

3. 3. A processing tool according to claim 2, The processing tool, wherein the processing chamber includes a showerhead located opposite the pedestal, and the optical interface is disposed on the showerhead.

4. 3. A processing tool according to claim 2, The processing tool, wherein the optical interface is disposed on the pedestal.

5. 10. The processing tool of claim 1, The processing tool, wherein the optical interface is disposed on a sidewall of the processing chamber.

6. 10. The processing tool of claim 1, 10. The processing tool, further comprising: one or more optical elements disposed between the optical interface and the hyperspectral camera, the one or more optical elements configured to direct electromagnetic radiation passing through the optical interface to the hyperspectral camera.

7. 10. The processing tool of claim 1, The processing tool, wherein the processing chamber is a plasma reactor chamber.

8. 10. The processing tool of claim 1, 10. The processing tool, further comprising: a computing system configured to execute a trained machine learning model, the trained machine learning model configured to receive one or more hyperspectral images from the hyperspectral camera and output metrology data for the processing chamber based on at least the one or more hyperspectral images.

9. 9. A processing tool according to claim 8, The processing tool, wherein the computing system is configured to adjust control parameters of a cleaning process for cleaning the processing chamber based on the metrology data for at least the processing chamber.

10. 9. A processing tool according to claim 8, 10. The processing tool of claim 9, wherein the trained machine learning model is configured to receive a series of hyperspectral images of a substrate in the processing chamber during a substrate processing cycle and to output time-based metrology data for the substrate based on at least the series of hyperspectral images of the substrate, and wherein the computing system is configured to adjust one or more control parameters of a process of the substrate processing cycle based on the time-based metrology data for at least the substrate during the substrate processing cycle.

11. 9. A processing tool according to claim 8, the trained machine learning model is configured to receive one or more hyperspectral images of a first substrate in the processing chamber during or after a first substrate processing cycle and output metrology data for the first substrate based on the one or more hyperspectral images of at least the first substrate; and the computing system is configured to, for a second substrate processing cycle for a second substrate, adjust one or more control parameters of a process for the second substrate processing cycle based on the metrology data for at least the first substrate.

12. 1. A computer-implemented method for controlling a processing tool, comprising: receiving one or more hyperspectral images of a process chamber of the process tool from a hyperspectral camera; sending the one or more hyperspectral images to a trained machine learning model configured to output metrology data for the process chamber based on at least the one or more hyperspectral images; 10. A method for controlling a processing tool, comprising adjusting one or more control parameters of a process performed by the processing tool based on the metrology data for at least the processing chamber.

13. 13. A method for controlling a processing tool according to claim 12, comprising:

10. The method for controlling a processing tool, wherein the process is a cleaning process for cleaning the processing chamber, and the one or more control parameters comprise control parameters of the cleaning process.

14. 13. A method for controlling a processing tool according to claim 12, comprising:

10. A method for controlling a processing tool, wherein the one or more hyperspectral images comprise a series of hyperspectral images of a substrate in the processing chamber, the series of hyperspectral images of the substrate being received from the hyperspectral camera during a substrate processing cycle for the substrate, the trained machine learning model being configured to output time-based metrology data for the substrate, and the one or more control parameters being adjusted during the substrate processing cycle for the substrate based on the time-based metrology data for at least the substrate.

15. 13. A method for controlling a processing tool according to claim 12, comprising:

10. A method for controlling a processing tool, wherein the one or more hyperspectral images include one or more hyperspectral images of a first substrate in the processing chamber during or after a first substrate processing cycle, and wherein the one or more control parameters are adjusted for a second substrate processing cycle for a second substrate based on the metrology data for at least the first substrate.

16. 13. A method for controlling a processing tool according to claim 12, comprising:

1. A method for controlling a processing tool, wherein the processing chamber is a plasma reactor chamber, the one or more hyperspectral images of the plasma reactor chamber are taken by the hyperspectral camera while a plasma is present in the plasma reactor chamber, and the plasma in the plasma reactor chamber is an illumination source for the hyperspectral camera.

17. 1. A processing tool comprising: a hyperspectral camera positioned to capture hyperspectral images of substrates within the processing tool; and a computing system configured to execute a trained machine learning model, the trained machine learning model configured to receive one or more hyperspectral images from the hyperspectral camera and to output metrology data for the substrate based on at least the one or more hyperspectral images.

18. 18. A processing tool according to claim 17, The processing tool, wherein the metrology data includes thicknesses of one or more layers of the substrate.

19. 18. A processing tool according to claim 17, The processing tool, wherein the metrology data includes a gap condition in a feature of the substrate.

20. 18. A processing tool according to claim 17, The hyperspectral camera has a dynamically adjustable position.

21. 18. A processing tool according to claim 17, The hyperspectral camera has a dynamically adjustable angle.

22. 18. A processing tool according to claim 17, The processing tool, wherein the metrology data includes a determined amount of stress and / or warpage in the substrate.

23. 18. A processing tool according to claim 17, The metrology data includes a determined amount of haze on the substrate.

Images