Machine learning and deep learning techniques for predicting eco-efficiency in substrate processing

Abstract

In some embodiments, a method includes receiving a process recipe including process recipe settings data. The method further includes inputting the process recipe to one or more trained machine learning models that output predicted environmental resource usage data indicative of environmental resource consumption associated with processing a substrate in a processing chamber according to the process recipe. The method further includes outputting a recommendation associated with the process recipe based at least in part on the predicted environmental resource usage data.

Description

[Technical Field]

[0001] This specification relates generally to the environmental impact of manufacturing equipment, such as semiconductor manufacturing equipment. More specifically, this specification relates to machine learning and deep learning techniques for predicting eco-efficiency in substrate processing. [Background technology]

[0002] The continuing demand for electronic devices is creating an ever-increasing demand for semiconductor wafers. The increased manufacturing required to produce these wafers is taking a significant toll on the environment in terms of resource utilization and the generation of environmentally harmful waste. Therefore, there is an increasing demand for more eco-friendly and environmentally responsible methods of wafer manufacturing, and manufacturing methods in general. Given the energy-intensive nature of wafer processing, there is value in decoupling the growth of the semiconductor industry from its environmental impact. The growing demand for chips and their increasing complexity are increasing resource consumption, which impacts the environment. Furthermore, the increasing complexity of chips increases the difficulty of determining eco-efficient substrate processing recipes. Summary of the Invention

[0003] The following is a simplified summary of the disclosure to provide a basic understanding of some aspects of the disclosure. This summary is not an extensive overview of the disclosure. It is not intended to delineate the scope of particular embodiments of the disclosure or the scope of the claims. Its sole purpose is to present some concepts of the disclosure in a simplified form as a prelude to the more detailed description that is presented later.

[0004] In some embodiments, a method includes receiving a process recipe including process recipe settings data. The method further includes inputting the process recipe to one or more trained machine learning models that output predicted environmental resource usage data indicative of environmental resource consumption associated with processing a substrate in a processing chamber according to the process recipe. The method further includes outputting a recommendation associated with the process recipe based at least in part on the predicted environmental resource usage data.

[0005] In some embodiments, the system includes one or more process chambers configured to process a substrate. The one or more chambers include a plurality of sensors. The system further includes a system controller for controlling the one or more process chambers. The system controller is operative to receive a process recipe including process recipe settings data. The system controller is further operative to input the process recipe to one or more trained machine learning models that output predicted environmental resource usage data indicative of environmental resource consumption associated with processing the substrate in the process chamber(s) according to the process recipe. The system controller is further operative to output a recommendation associated with the process recipe based at least in part on the predicted environmental resource usage data.

[0006] In some embodiments, a non-transitory machine-readable storage medium includes instructions that, when executed by a processing device, cause the processing device to train a first machine learning model to form a first trained machine learning model. The first trained machine learning model is trained to output predicted metrology data based on a process recipe input to the first trained machine learning model. The processing device further trains a second machine learning model using training data including the predicted metrology data output from the first trained machine learning model to form a second trained machine learning model. The second trained machine learning model is trained to output predicted environmental resource usage data indicative of environmental resource consumption associated with processing a substrate in a processing chamber according to the first process recipe input to the second trained machine learning model.

[0007] Aspects and embodiments of the present disclosure will be more fully understood from the following detailed description and accompanying drawings, which are intended to illustrate aspects and embodiments by way of example and not by way of limitation. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a schematic top view of an exemplary manufacturing system according to one embodiment. [Figure 2A] FIG. 1 is a block diagram illustrating a logical view of an exemplary eco-efficiency platform, according to one embodiment. [Figure 2B] FIG. 1 is a simplified block diagram illustrating a logical view of an exemplary eco-efficiency prediction platform, according to some embodiments of the present disclosure. [Figure 3] FIG. 1 is a block diagram illustrating an exemplary system architecture in which embodiments of the present disclosure may function. [Figure 4] 1 illustrates an exemplary digital replica according to some embodiments of the present disclosure. [Figure 5] 10A-10C are exemplary diagrams of process parameter value windows according to some embodiments of the present disclosure. [Figure 6]1 is a flowchart of a method for generating a training dataset for training a machine learning model, according to an aspect of the present disclosure. [Figure 7] FIG. 1 illustrates a flow diagram of a method for training a machine learning model to determine predicted cooling parameter values, according to an aspect of the present disclosure. [Figure 8A] 1 is a flow diagram of a method for obtaining recommendations for processing a substrate according to some embodiments of the present disclosure. [Figure 8B] FIG. 1 is a flow diagram of a method for obtaining predicted process recipe setting data according to some embodiments of the present disclosure. [Figure 8C] FIG. 1 is a flow diagram of a method for obtaining predicted environmental resource usage data according to some embodiments of the present disclosure. [Figure 9A] 1 illustrates a chart showing observed versus predicted environmental resource consumption data according to some embodiments of the present disclosure. [Figure 9B] 1 illustrates a chart showing predicted or actual time series environmental resource consumption data according to some embodiments of the present disclosure. [Figure 10] FIG. 1 illustrates a block diagram of an exemplary computing device that operates in accordance with one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0009] Eco-efficiency characterization is a complex technique used to determine various levels of inputs (e.g., resources, utilization, etc.) associated with a particular manufacturing tool during its use. Eco-efficiency characterization is used to determine how changes in inputs affect the eco-efficiency of the manufacturing tool. Eco-efficiency characterization and / or eco-efficiency prediction can be useful during manufacturing tool development to aid in the development of manufacturing tools that maximize per-unit (or per-hour) eco-efficiency and minimize adverse environmental impacts. Eco-efficiency characterization can also be useful after tool development to fine-tune the per-unit eco-efficiency characterization values ​​of the tool and / or process recipe in light of the specific parameters the tool will operate under during its operation.

[0010] Embodiments described herein provide a system for predicting and optimizing the eco-efficiency of substrate processing recipes through the design, development, and implementation of the process recipes. In some embodiments, the methods disclosed herein can assist engineers in the development, optimization, and / or operation of processes that meet both material engineering and eco-efficiency specifications. In some embodiments, sensor data and / or models are utilized to provide predictions of the eco-efficiency of numerous manufacturing systems, individual processing chambers of the manufacturing systems, and / or specific processing recipes executed within individual processing chambers. Furthermore, the methods described herein enable optimization of process recipes to increase eco-efficiency while maintaining targets for processed substrates. For example, the output of a predictive model can be used as an indicator of environmental resource usage to select and / or optimize process recipes to reduce consumption of environmental resources while meeting established substrate target results. According to some embodiments described herein, optimization and / or selection of process recipes to increase eco-efficiency can be achieved prior to the actual execution of the process recipe. For example, historical data can be used to develop and utilize models to determine the eco-efficiency of process recipes under development. In some embodiments, a comparison of the eco-efficiencies of several process recipes can be performed, and the process recipe with the greatest eco-efficiency that meets manufacturing goals can be selected. In this manner, the eco-efficiency and / or environmental impact of a substrate process recipe can be predicted and / or improved without physical testing or empirical results.

[0011] In some embodiments, an eco-efficiency prediction platform (e.g., software in a system controller) can receive process (e.g., process recipe setpoint data) and sensor data to form an eco-efficiency prediction. The process recipe and / or sensor data can be input to one or more models, such as one or more trained machine learning models, physics-based models (e.g., digital twins), and / or one or more additional models. In some embodiments, the process recipe is determined by a first model (e.g., a first predictive model, a trained machine learning model, etc.) based on targets for the processed substrates input to the model. For example, a user (e.g., an engineer, technician, etc.) can input target process results (e.g., for the processed substrates) to the model. The first model can be trained to output possible process recipes for processing the substrates, where each output process recipe meets the target process result. Because there can be multiple ways to achieve the target result (e.g., many recipes can produce substrates that meet the target), the first model can output multiple different process recipes, each of which meets the target result.

[0012] Each output recipe output by the model can be input to a second model (e.g., a second predictive model, a second trained machine learning model, etc.) configured to predict an eco-efficiency associated with the input process recipe. Predicted eco-efficiency data corresponding to each process recipe can be output by the second model. The predicted eco-efficiency values ​​can include predicted environmental resource usage data indicating environmental resource consumption (e.g., consumption of chemicals, gas, electricity, water, etc.). As used herein, environmental resource usage data can include data related to resource and / or chemical consumption, the environmental impact of the resources and / or chemicals used / consumed, energy consumption, and / or the environmental impact of the energy consumed. In some examples, the predicted data includes time series data indicating power consumption and / or gas flow associated with the process recipe. Each predicted eco-efficiency data can be analyzed and / or compared to determine the most eco-efficient process recipe (e.g., the least resource-intensive process recipe).

[0013] In some embodiments, recommendations for processing a substrate are output based on the eco-efficiency data corresponding to the process recipe. The recommendations may indicate that a particular process recipe should be executed to process the substrate to meet the process goals. In some embodiments, the recommendations may include modifying one or more process recipes and / or one or more additional goals and / or constraints to increase the process recipes' respective eco-efficiencies. The recommendations may be input into a first model (to predict a process recipe), which may output a further predicted or recommended process recipe. These further process recipes may be processed by a second model to determine resource consumption and / or eco-efficiency values ​​for the further process recipes. Analysis of the further recipes may be performed again, taking into account the eco-efficiency values ​​associated with these recipes, to provide further recommendations. This process may be iterative, such that the cooperating models converge on the most eco-efficient process recipe that meets the product goals.

[0014] In some embodiments, eco-efficiency is calculated on a unit basis. Typically, per-unit eco-efficiency is not considered during the development process of a manufacturing tool and / or process recipe. In addition, characterizing per-unit eco-efficiency while a manufacturing tool is in use (e.g., while the tool is being used to manufacture wafers) can be a cumbersome and complicated process to adjust the tool or process recipe settings. Furthermore, previous solutions required personnel and specialized eco-efficiency training of specialized engineers and analysts for eco-efficiency characterization. Embodiments of the present disclosure provide improved methods, systems, and software for unit-based eco-efficiency characterization. These methods, systems, and software can be used by individuals without specialized eco-efficiency training.

[0015] In some embodiments, characterization and / or prediction of eco-efficiency can be performed by software tools at all stages of the manufacturing equipment life cycle, including the design and operation stages of the manufacturing equipment. Eco-efficiency can include the amount of environmental resources (e.g., electrical energy, water, gas, chemicals, etc.) consumed per production run by the equipment (e.g., per wafer or device produced). Eco-efficiency can also be characterized as the amount of environmental impact (e.g., CO2 emissions, heavy metal waste, etc.) generated per production run by the equipment.

[0016] Any measurable quantity (e.g., substrate, die, area (cm)) manipulated by a manufacturing tool 2A per-unit analysis (e.g., by unit of production, period, device, etc.) allows for a more accurate characterization of eco-efficiency. "Per-unit" eco-efficiency allows for an accurate determination of resource use and environmental impact per unit produced and can be easily manipulated as a measure of value. For example, a particular manufacturing tool may be determined to have an electrical energy eco-efficiency rating of 1.0-2.0 kWh per substrate pass (in other embodiments, the eco-efficiency rating may be less than 0.5 kWh, up to 20 kWh, or greater than 20 kWh per substrate pass), indicating that each substrate processed by the manufacturing tool may use, for example, 1.0-2.0 kWh of electrical energy per processed substrate. In other embodiments, various other amounts of electrical energy may be used. Determining eco-efficiency on a substrate pass basis allows for easy comparison with other manufacturing tools that have different annual electrical energy consumption values ​​due to variations in annual substrate throughput. In one embodiment, eco-efficiency can also be determined on a per-device basis by dividing the per-substrate eco-efficiency characteristic value by the number of devices per wafer.

[0017] Eco-efficiency characterization or calculation can be performed on an operational manufacturing tool and / or process recipe. The manufacturing tool can access real-time variables, such as utilization and utility usage data, from a first sensor on the tool and a second sensor, which is an external sensor not part of the tool, and use the real-time variables in one or more eco-efficiency models. The manufacturing tool can fine-tune settings on the tool to maximize eco-efficiency given the current operating conditions of the tool. Similarly, sensor data (e.g., from the first sensor and / or the second sensor) can be input into a model (e.g., a trained machine learning model, a deep learning model, etc.) along with process recipe data (e.g., process recipe setpoint data) so that the model predicts eco-efficiency data corresponding to the process recipe. In some embodiments, the sensor data is input into the model to inform the model of physical constraints (e.g., physical constraints to form a physics-based model, etc.).

[0018] In some embodiments, modifications to a manufacturing process (e.g., a subset or multiple processes, process recipe operations, etc.) may be determined based on the environmental resource usage data or the characterization and / or prediction of eco-efficiency. For example, one or more modifications to process recipe parameters (e.g., setpoints) of a particular recipe may be determined based on predicted environmental resource usage data for multiple process recipes output from a model. The modifications to the process recipe may be associated with improved eco-efficiency (e.g., reduced environmental resource consumption and / or reduced environmental impact) of the selected manufacturing process.

[0019] In some embodiments, eco-efficiency is based on resource consumption, such as energy consumption, chemical consumption (e.g., gases such as hydrogen, nitrogen, chemicals used for etching or depositing thin films, and / or liquids that can be vaporized, atomized, or converted to a gaseous state via a bubbler, injector, or atomizer), clean dry air (CDA), and / or water consumption (e.g., process cooling water (PCW), de-ionized water (DIW), and ultrapure water (UPW)). However, in some embodiments, eco-efficiency is based on lifetime data of components associated with manufacturing equipment. For example, environmental resource consumption and / or environmental impact associated with an eco-efficiency characterization may be associated with replacement procedures or maintenance procedures for consumable parts of manufacturing equipment. However, it should be understood that such embodiments also apply to the consumption of chemicals having other states, such as chemicals in a liquid state. Embodiments described herein with respect to gas consumption equally apply to the consumption of other types of chemicals, such as liquids.

[0020] As described above, in some embodiments, the eco-efficiency prediction platform predicts environmental resource usage for a processing chamber that executes a manufacturing process according to a process recipe based on predicted process recipe data, sensor data, and / or substrate processing targets. By predicting eco-efficiency based on process targets and / or sensor data (e.g., sensor data used to inform a model), the predicted amount of resources used for a process in the processing chamber may be more accurately determined. Improving the accuracy of the eco-efficiency platform using such data may, in some embodiments, result in better process development and reduced overall resource consumption.

[0021] FIG. 1 is a schematic top view of an exemplary processing system 100 (also referred to herein as a manufacturing system) according to one embodiment. In some embodiments, processing system 100 can be an electronics processing system configured to perform one or more processes on a substrate 102. In some embodiments, processing system 100 can be an electronic device manufacturing system. Substrate 102 can be any suitable rigid, fixed-dimension, planar article (e.g., a silicon-containing disk or wafer, a patterned wafer, or a glass plate) suitable for fabricating electronic devices or circuit components thereon. In some embodiments, processing system 100 is a semiconductor processing system. Alternatively, processing system 100 can be configured to process other types of devices, such as display devices.

[0022] The processing system 100 includes a processing tool 104 (e.g., a mainframe) and a factory interface 106 coupled to the processing tool 104. The processing tool 104 includes a housing 108 having a transfer chamber 110 therein. The transfer chamber 110 includes one or more processing chambers (also referred to as process chambers) 114, 116, 118 disposed around and coupled to the transfer chamber 110. The processing chambers 114, 116, 118 may be coupled to the transfer chamber 110 through respective ports, such as slit valves. The processing chambers 114, 116, 118 may be configured to process substrates.

[0023] The processing chambers 114, 116, and 118 may be adapted to perform any number of processes on the substrate 102. The same or different substrate processes may be performed in each processing chamber 114, 116, and 118. Examples of substrate processes include atomic layer deposition (ALD), physical vapor deposition (PVD), or chemical vapor deposition (CVD), etching, annealing, hardening, pre-cleaning, metal or metal oxide removal, and the like. In one example, a PVD process is performed in one or both processing chambers 114, an etching process is performed in one or both processing chambers 116, and an annealing process is performed in one or both processing chambers 118. Other processes may be performed on substrates therein. The processing chambers 114, 116, and 118 may each include a substrate support assembly. The substrate support assembly may be configured to hold a substrate in a fixed position while a substrate process is performed.

[0024] The transfer chamber 110 also includes a transfer chamber robot 112. The transfer chamber robot 112 can include one or more arms, each including one or more end effectors at the end of the arm. The end effectors can be configured to handle specific objects, such as wafers. In some embodiments, the transfer chamber robot 112 is a selective compliance assembly robot arm (SCARA) robot, such as a two-axis SCARA robot, a three-axis SCARA robot, and a four-axis SCARA robot.

[0025] A load lock 120 may also be coupled to the housing 108 and the transfer chamber 110. The load lock 120 may be configured to connect and couple to the transfer chamber 110 on one side and couple to the factory interface 106 on the other side. The load lock 120 may have an environmentally controlled atmosphere that is changed from a reduced pressure environment (where substrates are transferred to and from the transfer chamber 110) to an inert gas environment at or near atmospheric pressure (where substrates are transferred to and from the fabrication interface 106). In some embodiments, the load lock 120 is a stacked load lock, having a pair of upper and lower internal chambers located at different vertical levels (e.g., one above the other). In some embodiments, the pair of upper internal chambers is configured to receive processed substrates from the transfer chamber 110 for removal from the processing tool 104, while the pair of lower internal chambers is configured to receive substrates from the factory interface 106 for processing within the processing tool 104. In some embodiments, the load lock 120 is configured to perform substrate processing (eg, etching or pre-cleaning) on ​​one or more received substrates 102 .

[0026] The factory interface 106 may be any suitable enclosure, such as an Equipment Front End Module (EFEM). The factory interface 106 accommodates substrate carriers 122 (e.g., Front Opening Unified Platforms (FOUPs)) docked to various load ports 124 of the factory interface 106. Pod, front-opening unified pod). A factory interface robot 126 (shown in dotted lines) may be configured to transfer substrates 102 between substrate carriers 122 (also referred to as containers) and the load lock 120. In other and / or similar embodiments, the factory interface 106 is configured to receive replacement parts from a replacement parts storage container 123. The factory interface robot 126 may include one or more robot arms and may be or include a SCARA robot. In some embodiments, the factory interface robot 126 has more axes and / or more degrees of freedom than the transfer chamber robot 112. The factory interface robot 126 may include an end effector at the end of each robot arm. The end effector may be configured to pick up and handle specific objects, such as wafers. Alternatively or additionally, the end effector may be configured to handle objects, such as rings of a process kit.

[0027] Any conventional robot type may be used for the factory interface robot 126. Transfers may be performed in any order or in any direction. In some embodiments, the factory interface 106 may be maintained in a non-reactive gas environment (e.g., using nitrogen as the non-reactive gas), for example, at a slight positive pressure.

[0028] Processing system 100 may also include a system controller 128. System controller 128 may be and / or include a computing device such as a personal computer, a server computer, a programmable logic controller (PLC), a microcontroller, etc. System controller 128 may include one or more processing devices, which may be a general-purpose processing device such as a microprocessor or a central processing unit. More specifically, the processing device may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, or a processor implementing other instruction sets or a combination of instruction sets. The processing device may also be one or more special-purpose processing devices, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, etc. System controller 128 may include data storage devices (e.g., one or more disk drives and / or solid state drives), main memory, static memory, network interfaces, and / or other components. System controller 128 is capable of executing instructions to perform any one or more of the methods and / or embodiments described herein. The instructions may be stored on a computer-readable storage medium, which may include main memory, static memory, secondary storage, and / or a processing unit (during which the instructions are executed).In an embodiment, execution of instructions by system controller 128 causes the system controller to perform the methods of Figures 8A and 8B. System controller 128 may also be configured to allow input and display of data, processing commands, and the like by a human operator.

[0029] In some embodiments, the system controller 128 includes an eco-efficiency module 129, which may be a local server, executing on the system controller 128 of the processing system 100. The eco-efficiency module 129 may be responsible for processing first sensor data generated by sensors in one or more of the process chambers 114, 116, 118 and second sensor data from additional sensors 140, 142, 144 external to the process chambers 114, 116, 118. The first sensor data may be generated by sensors integral with the process chambers 114, 116, 118. Such sensors may include, for example, temperature sensors, power sensors, current sensors, pressure sensors, concentration sensors, etc. The first sensor data output by the integrated sensors in the processing chambers 114, 116, 118 may include measurements of current, voltage, power, flow rate (e.g., of one or more gases, CDA, water, etc.), pressure, concentration (e.g., of one or more gases), velocity (e.g., of one or more moving parts, gases, etc.), acceleration (e.g., of one or more moving parts, gases, etc.), or temperature (e.g., of a substrate being processed, at various locations within the processing chamber, etc.). In one embodiment, each chamber includes between about 20 and about 100 sensors.

[0030] To obtain additional data not typically accessible by the integrated sensors of the process chambers 114, 116, 118, one or more external sensors 140, 142, 144, 152 may be attached to the process chambers 114, 116, 118 and / or to supply lines to and / or from the process chambers 114, 116, 118, and / or to subcomponents (e.g., pumps and / or abatement systems) operating on behalf of the process chambers 114, 116, 118. In one embodiment, each process chamber includes approximately three to six external sensors attached to the process chamber, subsystems associated with the process chamber, and / or inputs / outputs between the process chambers. Secondary sensor data output by the external sensors 140, 142, 144, 152 may include, for example, current, flow rate, temperature, eddy current, concentration, vibration, voltage, or power factor. Examples of external sensors 140, 142, 144, 152 that may be used include clamp sensors that measure AC or DC current (also called current clamps), clamp sensors that measure voltage, and clamp sensors that measure leakage current. Other examples of external sensors include vibration sensors, temperature sensors, ultrasonic sensors (e.g., ultrasonic flow sensors), accelerometers (e.g., acceleration sensors), etc.

[0031] In the illustrated example, the abatement system 130, gas supply system 134, water system 132, and / or CDA system 136 can provide environmental resources to the processing chambers 114, 116, 118 and / or other components of the processing system 100 (e.g., transfer chamber, factory interface, load lock, etc.). In embodiments, the abatement system 130 performs abatement on residual gases, reactants, and / or output associated with processes performed in the processing chambers 114, 116, 118. The abatement system 130 can, for example, combust the residual gases and / or reactants to ensure they do not pose an environmental risk. Additionally, in embodiments, one or more pumps can be attached to and / or operate on behalf of one or more processing chambers 114, 116, 118. External sensors 140, 142, 144, 152 are shown for simplicity with respect to one processing chamber 116. However, it should be understood that similar external sensors may be mounted on additional processing chambers and / or on lines between such additional chambers and / or in subsystems associated with such additional processing chambers.

[0032] External sensors 140, 142, 144, 152 may be IoT sensors in some embodiments. In some embodiments, external sensors include a power source, such as a battery. In some embodiments, external sensors are wired sensors that connect to a power source, such as an AC power outlet. In some embodiments, external sensors do not include a power source, but instead receive sufficient power to operate based on environmental conditions. For example, sensors that detect voltage, power, and / or current may be powered wirelessly by such power or current (e.g., by obtaining energy from the current flowing through wires securing the sensor).

[0033] In one embodiment, the external sensors 140, 142, 144, 152 are embedded systems. An embedded system is a class of computing device that is embedded in another device as a component of that device. The external sensors 140, 142, 144, 152 typically also include other hardware, electrical, and / or mechanical components that may be connected to the embedded system. An embedded system is typically configured to handle a specific task or set of tasks for which the embedded system may be optimized (e.g., generating and / or transmitting measurements). Therefore, an embedded system may have minimal cost and size compared to a typical computing device.

[0034] Each embedded system may include a communications module (not shown) that enables the embedded system (and thus the external sensors 140, 142, 144, 152) to connect to a LAN, a hub 150, and / or a wireless carrier network (e.g., implemented using various data processing devices, communication towers, etc.). The communications module may be configured to manage security, manage sessions, manage access control, manage communications with external devices, etc.

[0035] In one embodiment, the communication modules of the external sensors 140, 142, 144, 152 are configured to communicate using Wi-Fi®. Alternatively, the communication modules may be configured to communicate using Bluetooth®, Zigbee®, 6LowPAN (Internet Protocol version 6 over Low power Wireless Area Networks), power line communication (PLC), Ethernet (e.g., 10 Megabit (Mb), 100 Mb, and / or 1 Gigabit (Gb) Ethernet), or other communication protocols. If the communications module is configured to communicate with a wireless carrier network, the communications module may communicate using Global Systems for Mobile Communications (GSM), Code-Division Multiple Access (CDMA), Universal Mobile Telecommunications Systems (UMTS), 3GPP Long Term Evaluation (LTE), Worldwide Interoperability for Microwave Access (WiMAX), or any other second generation (2G), third generation (3G), fourth generation (4G), or other wireless technologies.

[0036] In one embodiment, the communication module is configured to communicate with the hub 150, which may be, for example, a Wi-Fi router or other type of router, switch, or hub. The hub 150 may be configured to communicate with the communication modules of each of the external sensors 140, 142, 144, and 152 and transmit measurements received from the external sensors 140, 142, 144, and 152 to the system controller 128. In one embodiment, the hub 150 has a wired connection (e.g., an Ethernet connection, a parallel connection, a serial connection, a Modbus connection, etc.) to the system controller 128 and transmits measurements to the system controller 128 via the wired connection. In one embodiment, the hub 150 is connected to one or more external sensors via a wired connection.

[0037] In some embodiments, hub 150 is connected to network devices connected to a local area network (LAN). System controller 128 and the network devices can each connect to the LAN via a wireless connection and can be wirelessly connected to each other via the LAN. External sensors 140, 142, 144, and 152 may not support all communication types supported by the network devices. For example, external sensor 140 may support Zigbee, and external sensor 142 may support Bluetooth. To enable such devices to connect to the LAN, hub 150 may function as a gateway device that connects to the network devices (not shown) via one of the connection types supported by the network devices (e.g., via Ethernet or Wi-Fi). The gateway device may also support other communication protocols, such as Zigbee, PLC, and / or Bluetooth, and can translate between supported communication protocols.

[0038] System controller 128 may be connected to a wide area network (WAN). The WAN may be a private WAN (e.g., an intranet) or a public WAN such as the Internet, or may include a combination of private and public networks. In an embodiment, system controller 128 may be connected to a LAN that includes a router and / or modem (e.g., a cable modem, a direct serial link (DSL) modem, a Worldwide Interoperability for Microwave Access (WiMAX®) modem, a long term evolution (LTE®) modem), etc., that provides connectivity to the WAN.

[0039] The WAN may include or be connected to one or more server computing devices (not shown). The server computing devices may include physical machines and / or virtual machines hosted by the physical machines. The physical machines may be rack-mounted servers, desktop computers, or other computing devices. In one embodiment, the server computing devices include virtual machines managed and provided by a cloud provider system. Each virtual machine provided by a cloud service provider may be hosted on a physical machine configured as part of the cloud. Such physical machines are often located within a data center. The cloud provider system and cloud may be provided as an Infrastructure as a Service (IaaS) layer. One example of such a cloud is Amazon's Elastic Compute Cloud (EC2).

[0040] The server computing device can host one or more services, which can be web-based services and / or cloud services (e.g., web-based services hosted within a cloud computing platform). The services can maintain sessions (e.g., via continuous or intermittent connections) with system controller 128 and / or system controllers of other manufacturing systems in the same location (e.g., within a manufacturing facility or fab) and / or at various locations. Alternatively, the services can periodically establish sessions with the system controllers. Through the sessions with system controller 128, the services can receive status updates from eco-efficiency module 129 running on system controller 128. The services can aggregate data and provide a graphical user interface (GUI) accessible via any device connected to a WAN (e.g., a mobile phone, a tablet computer, a laptop computer, a desktop computer, etc.).

[0041] An eco-efficiency module 129 executing on the system controller 128 may process first sensor data from integrated sensors in one or more processing chambers 114, 116, 118 and second sensor data from external sensors 140, 142, 144, 152 to determine environmental resource usage data reflecting environmental resource consumption, such as water consumption, gas consumption, and electricity consumption. Processing that may be performed by the eco-efficiency module 129 is described below with reference to the remaining figures.

[0042] In some embodiments, the eco-efficiency module can predict the eco-efficiency of a process recipe executed in one of the process chambers 114, 116, 118. In some examples, utilizing one or more machine learning models, the eco-efficiency module 129 can predict the environmental resource consumption of a process step. Using multiple process recipes as input, the eco-efficiency module 129 can determine the environmental resource consumption of each recipe. In some embodiments, the system controller can output a recommendation for substrate processing after comparing the environmental resource consumption associated with each recipe execution. The recommendation can be for executing the most eco-efficient recipe. In some embodiments, the recommendation can include a modification of one of the process recipes to make the process recipe more eco-efficient. In some examples, the eco-efficiency module 129 can update the process recipe based on the modification included in the recommendation.

[0043] 2A is a block diagram illustrating a logical view of an exemplary eco-efficiency platform 200A, according to one embodiment. Eco-efficiency platform 200A, in an embodiment, may execute on system controller 201. In one embodiment, system controller 201 corresponds to system controller 128 of FIG. 1, and eco-efficiency platform 200A is provided by eco-efficiency module 129 of FIG. 1.

[0044] The eco-efficiency platform 200A can receive first sensor data 270 from tool sensors 202, which in some embodiments can be integrated sensors of the process chambers 114, 116, 118 of FIG. 1 . The eco-efficiency platform 200A can further receive second sensor data 272 from a hub 206, where the hub 206 receives second sensor data from one or more external sensors 204, which in some embodiments can correspond to the external sensors 140, 142, 144, 152 of FIG. 1 . In some embodiments, the hub 206 provides the second sensor data to a server 207, which can run on one or more computing devices (e.g., in a cloud environment). The server (e.g., an IoT platform) 207 can compile the second sensor data into aggregated second sensor data 274 and transmit the aggregated second sensor data 274 to the eco-efficiency platform 200A. Such aggregated second sensor data 274 may be provided to eco-efficiency platform 200A instead of or in addition to second sensor data 272.

[0045] In some embodiments, historical data 208 (e.g., historical sensor data) may be stored in a data store, such as a database. Such historical data 208 may, in some embodiments, be additionally provided to eco-efficiency platform 200A. In some embodiments, historical data 208 may be used to train one or more machine learning models to predict eco-efficiency data as described herein.

[0046] At block 230, eco-efficiency platform 200A collects first sensor data 270, second sensor data 272, aggregated second sensor data 274, and / or historical data 208. At block 232, eco-efficiency platform 200A may pre-process some or all of the received data. Pre-processing may include normalizing the data, changing the units of the data, adding timestamps to the data, synchronizing the data based on the timestamps, and adding labels to the data.

[0047] At block 234, eco-efficiency platform 200A performs data processing on the received data (e.g., first sensor data 270 and second sensor data 272). This may include inputting the data into one or more data processing algorithms or functions, inputting the data into one or more physics-based models (e.g., digital twins, etc.), inputting the data into one or more trained machine learning models, etc. At block 236, output is generated by the one or more models, data processing algorithms, functions, etc. The output may include physical conditions and / or environmental resource usage data associated with the manufacturing process performed on the processing chamber. The output may be stored in a local data store, such as database 210.

[0048] A client computing device running a web client 220 or other client application including a graphical user interface (GUI) 222 or other type of user interface can be connected to eco-efficiency platform 200A. Web client 220 can send requests 212 to eco-efficiency platform 200A and receive responses 214. Requests 212 can include, for example, requests for environmental resource usage data, such as for one or more processing chambers and for manufacturing systems including multiple processing chambers, recipes executed on processing chambers, etc. Such requests can include requests to present environmental resource usage data in graphs, tables, etc.

[0049] In some embodiments, eco-efficiency platforms 200A of multiple system controllers 201 connect (e.g., via a WAN) with a remote computing device 250. The remote computing device 250 can include a remote server that aggregates data from the multiple eco-efficiency platforms and stores the aggregated data in a data store, such as a database 255. A web client 220 (or other client application) can connect to the remote server of computing device 250 to access environmental resource usage data for multiple manufacturing systems within a fab, multiple fabs, etc.

[0050] 2B is a simplified block diagram illustrating a logical view of an exemplary eco-efficiency prediction platform, according to some embodiments of the present disclosure. Eco-efficiency prediction platform 200B may, in some embodiments, execute on a system controller (e.g., system controller 201). Alternatively, eco-efficiency prediction platform 200B may execute on a server computer, which may or may not execute in a cloud environment. In one embodiment, eco-efficiency prediction platform 200B is provided by eco-efficiency module 129 of FIG. 1.

[0051] The eco-efficiency prediction platform 200B may receive one or more process targets 278 (e.g., a set of process targets) from the substrate processing tool 268 or another source. In one embodiment, a user (e.g., an engineer) inputs the one or more process targets via a user interface (e.g., a graphical user interface) of the eco-efficiency prediction platform 200B. In some embodiments, the one or more process targets 278 include target substrate process data indicating a target substrate state for a processed substrate. For example, the process target 278 may indicate a target processed substrate outcome and / or a target substrate specification. In some examples, the process target 278 may indicate that the processed substrate should have one or more features (e.g., an etch feature, a deposition feature, a coating feature, a film thickness, etc.). The process target may be received by the substrate processing tool 268 via user input (e.g., via a GUI).

[0052] In some embodiments, process targets 278 are input to process model 262. Process model 262 can be a physics-based model, a statistical model, a trained machine learning model (e.g., one or more trained machine learning models), or a hybrid model (e.g., a combination of one or more model types). For example, process model 262 can be a trained machine learning model based on physics information. In some embodiments, process model 262 is trained to output multiple process recipes (e.g., process recipes 280(1)-280(n)) based on the input process targets. In some embodiments, process model 262 represents a substrate manufacturing process. Process model 262 can be trained using historical data 208, including historical process targets, historical process recipes, and / or historical eco-efficiency data. Process model 262 can be trained using training input data, including historical target substrate processing data corresponding to process targets for various substrate processing steps. Historical target substrate processing data can be collected over time while substrates are processed and / or while new process targets are received. The process model 262 may be trained using training target data, including historical process recipes (e.g., historical process recipe setpoint data) corresponding to historical target substrate processing data. For example, the process model 262 may be trained using process recipe data (e.g., training target outputs) and corresponding process target data (e.g., training inputs) that the process recipes were to achieve. The historical process recipes may be collected over time during and / or before the processing steps of new substrates are performed.

[0053] The output from the process model 262 is a plurality of process recipes 280(1)-280(n). In some examples, the process model 262 outputs n process recipes. Each process recipe 280 may indicate settings (e.g., process recipe settings, control knob settings, etc.) for one or more process recipe steps. In some embodiments, each of the process recipes 280(1)-280(n), when executed, produces a processed substrate that meets the process objectives 278 (e.g., meets target specifications). Each of the process recipes 280(1)-280(n) may have different process settings, such as different temperatures, gas flow rates, gas delivery times, and pressures. Furthermore, each of the process recipes 280(1)-280(n) may use different amounts of environmental resources, such as process gases and / or power. Correspondingly, each of the process recipes 280(1)-280(n) may have different environmental efficiencies.

[0054] In some embodiments, the process recipes 280(1)-280(n) are input to one or more chamber models 264. The chamber models 264 can serially or parallelly receive and / or make eco-efficiency predictions based on the process recipes 280(1)-280(n). In some embodiments, multiple chamber models 264 are used, where some chamber models receive the process recipes and outputs of other chamber models and generate outputs based on such inputs. In some embodiments, multiple chamber models are "daisy chained," where the first model(s) in the chain can output readings and / or resource consumption predictions that have a direct and understandable correlation to sensor readings and / or recipe settings of the process recipe. Subsequent models may have less understandable or less direct correspondences to recipe settings and / or sensor readings. However, there may be a correspondence between a first reading / resource consumption output by a first model and a subsequent reading / resource consumption output by a second model. Each chamber model 264 may be a physics-based model, a statistical model, a trained machine learning model (e.g., one or more trained machine learning models), or a hybrid model (e.g., a combination of one or more model types). For example, chamber model 264 may be a trained machine learning model based on physics information.

[0055] The chamber model 264 may be a model representing a processing chamber. For example, the chamber model 264 may be a digital twin of the processing chamber. In some embodiments, the chamber model 264 is trained to output predicted environmental data (e.g., eco-efficiency data 282(1)-282(n)) corresponding to an input process recipe. The chamber model 264 may be trained using historical data 208. The chamber model 264 may be trained using training input data including historical process recipe data collected over time. For example, the chamber model 264 may be trained using process recipe data corresponding to process recipe steps performed in the corresponding processing chamber. In some embodiments, the chamber model 264 includes two or more trained machine learning models. In some examples, a first machine learning model is trained using historical process recipe data and historical eco-efficiency data. The first machine learning model may be trained to output predicted measurements (e.g., predicted sensor measurements such as temperature, power, flow rate, and / or other data related to environmental resource consumption) based on an input process recipe. The second machine learning model can be trained using output (e.g., predicted measurements) from the first machine learning model, historical process recipes, and / or historical eco-efficiency data. The second machine learning model can be trained to output predicted eco-efficiency data based on the input process recipes.

[0056] In some embodiments, the chamber model 264 is trained using additional training input data including sensor data associated with substrate processing (e.g., historical sensor data) received from sensors in the corresponding processing chamber. In some embodiments, a second machine learning model of the chamber model 264 is trained using predicted sensor data output by the first machine learning model of the chamber model 264. The sensor data may include the first sensor data and / or the second sensor data, as described herein. For example, the chamber model 264 may be trained with data including measurements of current, voltage, power, flow rate (e.g., of one or more gases, CDA, water, etc.), pressure, concentration (e.g., of one or more gases), velocity (e.g., of one or more moving parts, gases, etc.), acceleration (e.g., of one or more moving parts, gases, etc.), or temperature (e.g., of a substrate being processed and at various locations within the processing chamber). In some embodiments, sensor data is collected over time and stored in a database for later training of the chamber model 264. In some embodiments, the sensor data is used to provide information to the chamber model 264. For example, providing historical sensor data to chamber model 264 can result in chamber model 264 being a physics-informed trained machine learning model. Providing information to chamber model 264 can provide constraints for chamber model 264, thus improving the accuracy of the model.

[0057] In some embodiments, predicted eco-efficiency data 282 is output from the chamber model 264. The eco-efficiency data 282 may indicate the environmental resource usage (e.g., consumption) of the process recipe 280. In some examples, the eco-efficiency data 282 includes predicted time-series data reflecting the predicted behavior of the processing chamber (e.g., energy usage over time, gas consumption over time, etc.). FIG. 9B illustrates an example of predicted time-series resource consumption data. In some examples, the eco-efficiency data 282 indicates time-series data of predicted environmental resource consumption associated with substrate processing over time, such as predicted power consumption, predicted gas consumption, and predicted water consumption. For each process recipe 280 input into the chamber model 264, a corresponding set of eco-efficiency data 282 is output. For example, based on the process recipe 280(1) input into the chamber model 264, the chamber model 264 outputs eco-efficiency data 282(1). Similarly, based on the process recipe 280(n) input to the chamber model 264, the chamber model 264 outputs eco-efficiency data 282(n). The chamber model 264 may output each set of eco-efficiency data 282(1)-282(n) sequentially or in parallel. In some embodiments, a data analyzer 266 receives the eco-efficiency data 282. The data analyzer 266 may perform data analysis operations on the eco-efficiency data 282. In some examples, the data analyzer 266 may compare each set of eco-efficiency data 282 to determine the most eco-efficient corresponding process recipe 280.

[0058] In some embodiments, the data analyzer 266 outputs a recommendation 284 to the substrate processing tool 268. The recommendation 284 may be associated with processing a substrate in a processing chamber according to one of the process recipes 280(1)-280(n). For example, based on (e.g., in response to) eco-efficiency data 282(2) indicating that process recipe 280(2) is the most eco-efficient process recipe among the process recipes 280(1)-280(n), the data analyzer 266 may recommend to the substrate processing tool 268 that process recipe 280(2) be executed to process the substrate in order to meet the process target 278. In some embodiments, the recommendation 284 includes a modification to one of the process recipes 280 and / or a modification to the process target 278 to increase the eco-efficiency of the process recipe. In some embodiments, the modification includes one or more additional targets, one or more constraints for the process recipe (e.g., maximum temperature, minimum temperature, etc.). In one example, the data analyzer 266 may determine that a previously predicted process recipe should be modified to increase the eco-efficiency of one of the process recipes 280. The data analyzer 266 may instruct the substrate processing tool 268 of the modification via the recommendation 284. In some embodiments, the modification to the process recipe is to form a modified process recipe. The modified process recipe may have reduced environmental resource consumption (e.g., be more eco-efficient) compared to the unmodified process recipe. In some embodiments, the data analyzer 266 utilizes one or more trained machine learning models that are trained to output the recommendation 284 based on the input eco-efficiency data 282.

[0059] In some embodiments, the recommendations 284 are received by the process model 262. The recommendations 284 may be used by the process model 262 to predict additional process recipes 280(1)-280(n) that have lower resource consumption and / or satisfy one or more updated process goals and / or newly added constraints.

[0060] In some examples, the process model 262 is further trained using training inputs including past recommendations 284. The process model 262 can use the recommendations 284 indicating modifications to the process recipe to output additional predicted process recipes 280(1)-280(n) that are more eco-efficient than the previously predicted process recipe. In this manner, an iterative cycle can be established. For each cycle, the process targets 278 can be updated based on the recommendations 284, the process model 262 can output an updated process recipe based on the updated process targets, the chamber model 264 can output updated eco-efficiency data based on the updated process recipe, the data analyzer 266 can output updated recommendations based on the updated eco-efficiency data, and so on. In some examples, the process model 262 can output the more efficient process recipe 280 indicated by the eco-efficiency data 282. The data analyzer 266 may then output recommendations 284 for further modifications to the determined most eco-efficient process recipe and / or modifications to the process targets 278 to further increase the eco-efficiency of the process recipe or recipes. The substrate processing tool 268 may cause substrate processing in a processing chamber based on the process targets 278 and / or the recommendations 284. For example, the substrate processing tool 268 may initialize substrate processing using the process indicated by the recommendations 284 to meet the process targets 278.

[0061] 3 is a block diagram illustrating an example system architecture 300 in which embodiments of the present disclosure may function. As shown in FIG. 3 , the system architecture 300 includes a manufacturing system 302, a data store 312, a server 320, a client device 350, and / or a machine learning system 370. The machine learning system 370 may be part of the server 320. In some embodiments, one or more components of the machine learning system 370 may be fully or partially integrated into the client device 350. The manufacturing system 302, the data store 312, the server 320, the client device 350, and the machine learning system 370 may each be hosted by one or more computing devices, including a server computer, a desktop computer, a laptop computer, a tablet computer, a notebook computer, a personal digital assistant (PDA), a mobile communication device, a mobile phone, a handheld computer, an augmented reality (AR) display and / or headset, a virtual reality (VR) display and / or headset, a mixed reality (MR) display and / or headset, or a similar computing device. In this specification, the term "server" may refer to a server, but may also include an edge computing device, an on-premise server, a cloud, and the like.

[0062] The manufacturing system 302, the data store 312, the server 320, the client device 350, and the machine learning system 370 may be connected to each other via a network (e.g., to perform methods described herein). In some embodiments, the network 340 is a private network that provides each element of the system architecture 300 with access to each other and to other privately available computing devices. The network 340 may include one or more wide area networks (WANs), local area networks (LANs), wired networks (e.g., Ethernet networks), wireless networks (e.g., 802.11 networks or Wi-Fi networks), cellular networks (e.g., Long Term Evolution (LTE) networks), cloud networks, cloud services, routers, hubs, switches, server computers, and / or any combination thereof. Alternatively or additionally, any of the above elements of the system architecture 300 may be integrated or otherwise connected together without using the network 340.

[0063] Client device 350 can be or include any personal computer (PC), laptop, mobile phone, tablet computer, netbook computer, network-connected television ("smart TV"), network-connected media player (e.g., Blu-ray player), set-top box, over-the-top (OTT) streaming device, operator box, etc. Client device 350 can include a browser 352, an application 354, and / or other tools as described above executed by other systems in system architecture 300. In some embodiments, client device 350 can access manufacturing system 302, data store 312, server 320, and / or machine learning system 370 and can communicate (transmit and / or receive) one or more predicted environmental resource consumption (e.g., environmental resource consumption) and / or predicted environmental impact and / or predicted eco-efficiency indications, including inputs and outputs of various processing tools (e.g., component integration tool 322, digital replica tool 324, optimization tool 326, recipe builder tool 328, and resource consumption tool 330) at various stages of processing of system structure 300 as described herein.

[0064] 3, the manufacturing system 302 includes a machine 304, a system controller 306, a process recipe 308, and a sensor 310. The machine 304 may be an ion implanter, an etch reactor (e.g., a processing chamber), a photolithography tool, a deposition tool (e.g., for performing chemical vapor deposition (CVD), physical vapor deposition (PVD), ion-assisted deposition (IAD), etc.), or any other combination of manufacturing tools.

[0065] The process recipe 308, also referred to as a manufacturing recipe or manufacturing process instructions, includes a sequence of machine operations that, when applied in a specified order, comprise the execution of a process that produces a manufactured sample (e.g., a substrate having predetermined properties or meeting predetermined specifications). In some embodiments, the process recipe is stored in a data store, or alternatively or additionally, is stored to produce a table of data describing the processing of the manufacturing process. Each step can be associated with known environmental resource usage data. Alternatively or additionally, each processing step can be associated with parameters that describe the physical conditions of the processing step (e.g., target pressure, temperature, exhaust, energy throughput, etc.).

[0066] The system controller 306 may include software and / or hardware components capable of executing the steps of the process recipe 308. The system controller 306 may monitor the manufacturing process through sensors 310. The sensors 310 may measure process parameters to determine whether process criteria (e.g., target process criteria) are met. The process criteria may be associated with process parameter value windows. The sensors 310 may include various sensors that can be used to measure consumption (e.g., power, current, etc.) (either explicitly or as a measure of consumption). The sensors 310 may include physical sensors, integrated sensors that are components of the process chamber, external sensors, Internet-of-Things (IoT) sensors, and / or virtual sensors (e.g., based on virtual measurements based on models that estimate parameter values ​​rather than physical sensors), etc.

[0067] Additionally or alternatively, the system controller 306 can monitor eco-efficiency by measuring resource consumption (e.g., emissions, energy consumption, process component consumption, etc.) of various process steps. In some embodiments, the system controller 306 determines the eco-efficiency of the associated machinery 304. Based on the predicted and / or determined eco-efficiency model (e.g., including determined and / or predicted modifications to the process recipe 308), the system controller 306 can also adjust settings associated with the manufacturing equipment 304 to optimize the eco-efficiency of that equipment 304 given the current manufacturing conditions.

[0068] In one embodiment, the system controller 306 may include main memory (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM), static random access memory (SRAM), etc.) and / or secondary memory (e.g., a data storage device such as a disk drive) (e.g., data store 312 or cloud data). The main memory and / or secondary memory may store instructions (e.g., process recipes 308) for performing various types of manufacturing processes.

[0069] In one embodiment, the system controller 306 can determine an actual eco-efficiency characteristic value associated with the manufacturing equipment 304 based on first utility usage data associated with the manufacturing equipment 304 and first usage data associated with the manufacturing equipment 304. The first utility usage data and the first usage data can be determined, for example, by the system controller 306. In other embodiments, the first utility usage data and the first usage data are received from an external source (e.g., the server 320, a cloud service, and / or a cloud data store). The system controller 306 can compare the actual eco-efficiency characteristic value to the first eco-efficiency characteristic value (e.g., a first estimated eco-efficiency characteristic value) and / or predicted eco-efficiency associated with the manufacturing equipment 304. The eco-efficiency characteristic value can differ when values ​​of the utility usage data and usage data that differ from the actual values ​​associated with the operating manufacturing equipment 304 are used to calculate the first eco-efficiency characteristic value.

[0070] In one embodiment, the system controller 306 may determine the predicted eco-efficiency characteristic as being more eco-efficient than the actual eco-efficiency characteristic, indicating that settings on the manufacturing equipment 304 may be adjusted to better optimize the manufacturing equipment 304 for eco-efficiency. In some embodiments, the manufacturing equipment 304 may control and / or adjust settings of its subcomponents to better optimize eco-efficiency.

[0071] The system controller 306 may also determine that the actual utility usage data or actual utilization data is not the same as the predicted utility usage data and utilization data associated with the first eco-efficiency characteristic value. This may be the case when nominal or estimated data values ​​are used to determine the first eco-efficiency characteristic value and different actual recorded data values ​​are used while the manufacturing equipment 304 is in operation. In such a scenario, adjustments to one or more settings associated with the manufacturing equipment 304 may be beneficial to optimize the eco-efficiency of the manufacturing equipment.

[0072] The data store 312 can be a memory (e.g., random access memory), a drive (e.g., a hard drive, a flash drive), a database system, or any other type of component or device capable of storing data, such as a store provided by a cloud server and / or a processor. The data store 312 can store one or more historical sensor data. The data store 312 can store one or more of eco-efficiency data 314 (e.g., including historical, predicted, and / or current eco-efficiency data), sensor and process recipe data 316 (e.g., including historical, predicted, and / or current sensor and process recipe data 316), correction and optimization data (e.g., including historical, predicted, and / or current correction and optimization data 318), and digital replica data 319. The sensor and process recipe data 316 can include various process steps, process parameter windows, alternative process steps, process queuing instructions, etc. for performing multiple processes on overlapping manufacturing equipment. The sensor and process recipe data 316 may be linked or otherwise associated with the eco-efficiency data 314 to track and / or predict eco-efficiency across various process steps, recipes, etc. The modification and optimization data 318 may include past modifications made to previous process recipes (including adjustments to individual process steps or multiple process recipes) and the associated eco-efficiency changes resulting from the modifications.

[0073] Eco-efficiency data 314 may include various resource consumption data used in characterizing and / or predicting eco-efficiency. In one embodiment, eco-efficiency data 314 incorporates one or more of water usage, emissions, electrical energy usage, and any combination thereof. In other embodiments, eco-efficiency data 314 may include other categories of resource consumption, such as gas usage, heavy metal usage, and eutrophication potential.

[0074] Digital replica data 319 may include data associated with the digital replica. Digital replica data 319 may include data associated with the digital twin. As used herein, a digital twin may include a digital replica of a physical asset, such as manufacturing equipment 304. The digital twin includes properties of the physical asset at each stage of the manufacturing process, including, but not limited to, coordinate axis dimensions, weight properties, material properties (e.g., density, surface roughness), electrical properties (e.g., conductivity), optical properties (e.g., reflectivity), etc.

[0075] As previously described, a digital replica may include a physics-based model of one or more physical assets of a substrate manufacturing system. The digital replica data 319 may encapsulate relationships, parameters, specifications, etc. associated with one or more aspects of the physics-based model. For example, a physics-based model may describe the relationship between the size and geometry of a substrate processing chamber and environmental resource consumption. A physics-based model may describe the relationship between the type of purge gas used in a substrate manufacturing system and environmental resource consumption.

[0076] Server 320 may include a component integration tool 322, a digital replica tool 324, an optimization tool 326, a recipe builder tool 328, a resource consumption tool 330, and / or a research tool. Component integration tool 322 may determine cumulative consumption per equipment (e.g., per individual manufacturing equipment). The various tools of server 320 may communicate data between each other to perform their respective functions, as described herein.

[0077] The component integration tool 322 can receive manufacturing data (e.g., recipes, recipe selections, manufacturing equipment, inter-recipe and intra-recipe processes, etc.) and perform eco-efficiency analysis across various portions of the data. In some embodiments, the component integration tool 322 can determine eco-efficiency characteristics across multiple processing steps from individual process recipes. For example, the component integration tool 322 can determine eco-efficiency characteristics across all steps of a substrate manufacturing process from start to finish. For example, each manufacturing step can include one or more manufacturing steps (e.g., hundreds of manufacturing steps), each with its own eco-efficiency projection and collectively with its own collective eco-efficiency projection. In other examples, a selection from the manufacturing processes can be used to determine the eco-efficiency of a subset of manufacturing processing steps.

[0078] In another embodiment, the component integration tool 322 can perform inter-recipe process eco-efficiency predictions. For example, an eco-efficiency prediction may be associated with a manufacturing tool (e.g., in the manufacturing system 302) that performs multiple different process steps from multiple different manufacturing processes (e.g., process recipes 308). In another example, the order of various process steps (e.g., within a recipe or across recipes) may affect the overall eco-efficiency. The component integration tool 322 can perform an overall eco-efficiency prediction across a system of manufacturing tools and / or a series of processes. For example, the component integration tool 322 can perform an eco-efficiency comparison between subcomponents (e.g., multiple process chambers) that perform similar functions.

[0079] In an illustrative example, each process step, such as epitaxial deposition or etching, may be performed by a processing chamber. Each of the process steps is performed using a process recipe. Various process recipes may exist to perform a process, such as epitaxial deposition. For example, a process recipe may include multiple steps, such as: 1) purging the chamber, 2) pumping, 3) flowing with gas, 4) heating the chamber, etc. These steps may be associated with one or more process recipes.

[0080] In another embodiment, the component integration tool 322 can perform eco-efficiency predictions, including eco-efficiency of auxiliary equipment. Auxiliary equipment may include equipment not directly used in manufacturing but that assists in the execution of various process recipes. For example, auxiliary equipment may include substrate transport systems designed to move wafers between various manufacturing tools. In other examples, auxiliary equipment may include heat sinks, shared exhaust ports, power supply systems, etc. The component integration tool 322 can consider the resource consumption of the auxiliary equipment and combine it with the manufacturing resource consumption to predict the resource consumption of a process recipe (e.g., a subset or the entire recipe) or a combination of recipes (e.g., a subset or the entire recipe).

[0081] In other embodiments, the component integration tool 322 can perform eco-efficiency predictions that consider a sequence of processes or recipes. For example, processing step A followed by processing step B can result in a first resource consumption, and processing step B followed by processing step A can result in a second resource consumption that is different from the first resource consumption. The component integration tool 322 integrates eco-efficiency across multiple machines and / or processing steps and considers the sequence of processing steps in a processing recipe (e.g., a subset or the entire recipe) or combination of recipes (e.g., a subset or the entire recipe).

[0082] In some embodiments, there are different manufacturing tools for each process step. For example, a film on a wafer may have multiple layers. A first machine may perform a first process (e.g., deposition), a second machine may perform a second process (e.g., etch), a third machine may perform a third process (e.g., deposition), and so on. The component integration tool 322 may direct the resource consumption tracker to track multiple process steps across multiple machines and generate a data stash report. As previously mentioned, a consumption report including wafer life from start to finish may be plotted for a selection of process recipes.

[0083] In some embodiments, the component integration tool 322 can perform a comparison of environmental resource consumption between chambers. The component integration tool can utilize the digital replica tool 324 to provide one or more physical data that provides a rationale for the difference in eco-efficiency between the two chambers.

[0084] The digital replica tool 324 receives manufacturing data from the manufacturing system 302 and / or the client device 350 and generates a digital replica associated with the manufacturing data. The manufacturing data may include machine 304 selection and process steps for the process recipe 308. The digital replica tool 324 generates a digital twin of the physical system structure of the manufacturing system, or a virtually input system (e.g., generated by a user on the client device 350).

[0085] The digital replica generated by the digital replica tool 324 may include one of a physical model, a statistical model, and / or a hybrid model. The physical model may include physics-based constraints and control algorithms designed to estimate the physical conditions of the input manufacturing data (e.g., exhaust temperature, power supply requirements, and / or other conditions indicative of the physical environment associated with environmental resource consumption). For example, a user may generate a processing recipe on the client device 350. The digital replica may include process or recipe parameters and instructions for using machinery in a specific manner. The digital replica tool 324 may obtain the manufacturing data and determine the physical constraints of the system (e.g., operating temperature, pressure, exhaust parameters, etc.). For example, the physical model may identify the physical conditions of the system based on the chamber hardware configuration (e.g., use of type A equipment materials vs. type B equipment materials) and / or recipe parameters. In another example, the physical conditions may be determined from associated machinery components that affect water, heat loss to air, and / or heating, ventilation, and air conditioning (HVAC) equipment. The digital replica tool 324 can work in conjunction with other tools (e.g., the component integration tool 322 and / or the resource consumption tool 330) to develop a prediction of the eco-efficiency of the received manufacturing data. Note that the digital replica tool 324 can predict the eco-efficiency of a manufacturing process and the selection of manufacturing equipment without receiving empirical data from the execution of the process recipe by the manufacturing equipment 304. Thus, the digital replica of the manufacturing equipment can be used to predict the eco-efficiency of an equipment design and / or process recipe without actually building the particular equipment design or executing the particular process recipe.

[0086] In some embodiments, the digital replica tool 324 can operate in conjunction with a digital twin. As used herein, a digital twin is a digital replica of a physical asset, such as a manufactured part. The digital twin includes properties of the physical asset at each stage of the manufacturing process, including, but not limited to, coordinate dimensions, weight properties, material properties (e.g., density, surface roughness), electrical properties (e.g., conductivity), and optical properties (e.g., reflectivity), among others.

[0087] In some embodiments, the physical models used by the digital replica tool 324 may include fluid flow modeling, gas flow and / or consumption modeling, chemistry-based modeling, heat transfer modeling, electrical energy consumption modeling, plasma modeling, and the like.

[0088] In some embodiments, the digital replica tool 324 can utilize statistical modeling to predict the eco-efficiency of the manufacturing data. Statistical models can be used to process manufacturing data based on previously processed historical eco-efficiency data (e.g., eco-efficiency data 314) using statistical operations to validate, predict, and / or transform the manufacturing data. In some embodiments, statistical models are generated using statistical process control (SPC) analysis to determine control limits for the data and identify whether the data is highly or less reliable based on the control limits. In some embodiments, statistical models are associated with univariate and / or multivariate data analysis. For example, various parameters can be analyzed using statistical models to determine patterns and correlations through statistical procedures (e.g., range, minimum, maximum, quartiles, variance, standard deviation, etc.). In other examples, regression analysis, path analysis, factor analysis, multivariate statistical process control (MCSPC), and / or multivariate analysis of variance (MANOVA) can be used to identify relationships between multiple variables.

[0089] The optimization tool 326 can receive the process recipe 308 and machine equipment 304 selections and identify modifications to the selections to improve eco-efficiency (e.g., reduce resource consumption, resource cost consumption, and / or environmental impact (e.g., gaseous or particulate species entering the atmosphere)). The optimization tool 326 can incorporate the use of one or more machine learning models (e.g., model 390 of machine learning system 370). In some examples, a first machine learning model can receive as input a target substrate output associated with a target outcome for a processed substrate. A second machine learning model can receive as input (e.g., output from the first machine learning model) the process recipe selections and determine eco-efficiency data corresponding to each recipe in the process recipe selections. In some embodiments, the machine learning model can determine one or more modifications to the selections that improve the overall eco-efficiency of the selections when executed by the manufacturing system 302. In some embodiments, the machine learning model can use a digital replica tool to generate synthetic manufacturing data for training. Alternatively or additionally, the machine learning model can use historical data (e.g., eco-efficiency data 314, sensor and process recipe data 316, and / or correction and optimization data 318) to train the machine learning model.

[0090] The modifications identified by the optimization tool 326 may include changes to the process steps, changes to the order of processes, changes to parameters performed by a piece of machinery, changes to the interaction between the first and second process recipes (e.g., order, simultaneous processing, delay times, etc.), etc. In some embodiments, the optimization tool 326 may send instructions to the manufacturing system 302 to perform the optimization directly. However, in other embodiments, the optimization tool may display the modifications on a graphical user interface (GUI) for an operator to act upon. For example, the digital replica tool 324 may send one or more modifications to the client device 350 for display in the browser 352 and / or application 354.

[0091] In some embodiments, optimization tool 326 can adjust hyperparameters of the digital twin model generated by digital replica tool 324. As described in later embodiments, optimization tool 326 can incorporate reinforcement learning and / or deep learning by running simulated modifications on the digital replica and evaluating the eco-efficiency results output from the digital replica.

[0092] In some embodiments, the optimization tool 326 can perform eco-efficiency predictions and optimizations that prioritize one or more types of environmental resources. For example, as previously described, eco-efficiency predictions can be based on various resource consumptions, such as water consumption, gas consumption, and energy consumption. The optimization tool 326 can perform optimizations that prioritize a first resource consumption (e.g., water consumption) over a second resource consumption (e.g., gas consumption). In some embodiments, the optimization tool 326 can perform optimizations using a weighted priority system. For example, when optimizing eco-efficiency and / or identifying eco-efficiency modifications to a manufacturing process, weights can be assigned to one or more resource consumptions, with the weights indicating the optimization priority of the associated per-unit resource consumption.

[0093] The recipe builder tool 328 can receive a manufacturing process and / or machinery selection and dynamically determine and predict eco-efficiency step-by-step after each addition, deletion, and / or modification to the virtual manufacturing process and / or equipment selection. The recipe builder tool 328 can dynamically update the determined eco-efficiency using other tools (e.g., the component integration tool 322, the digital replica tool 324, the optimization tool 326, and the resource consumption tool 330) when the manufacturing recipe is updated. For example, a user can generate a manufacturing recipe. The recipe builder tool 328 can output a current eco-efficiency for the current iteration of the process recipe. The recipe builder tool 328 can receive updated process recipe modifications for the current iteration. The recipe builder tool 328 can output a prediction of the updated eco-efficiency. In some embodiments, the recipe builder tool 328 uses one or more models to predict a substrate processing recipe that meets threshold criteria. (See FIG. 2B and related discussion.)

[0094] In some embodiments, the recipe builder tool 328 and the optimization tool 326 can be used to identify one or more recipes as being more eco-efficient than other recipes. For example, the recipe builder tool 328 can cause one or more (e.g., the top three) of the most energy-efficient recipes associated with a processing tool to be presented or otherwise provided on a GUI (e.g., on the client device 350). The recipe builder tool 328 can use the digital replica tool 324 to provide details that provide the rationale for why one or more energy-efficient recipes are performing with a correspondingly high eco-efficiency.

[0095] The resource consumption tool 330 can track various resource consumptions (e.g., projected resource consumption). For example, as previously described, eco-efficiency projections can be based on broader resources, such as energy consumption, gas emissions, water usage, etc. However, the resource consumption tool 330 can track resource consumption more specifically. In some embodiments, a selection of a process recipe and / or manufacturing equipment is received by the resource consumption tool 330. The resource consumption tool 330 can determine lifespan data for components associated with the selection of manufacturing equipment and / or process recipe. For example, manufacturing equipment wears out over time and, in some cases, requires remedial action, such as part replacement and / or repair. The remedial action is also associated with an environmental consumption (e.g., projected resource consumption to implement the remedial action). The resource consumption tool 330 can track lifespan data for components individually based on the remedial action projected to be implemented in the future, and provide per-unit environmental resource consumption and / or environmental impact.

[0096] In some embodiments, environmental resource consumption may be predicted, monitored, tracked, and / or otherwise determined across various failures. In some embodiments, the resource consumption tool 330 is capable of predicting resource consumption based on a selected process recipe. In some embodiments, the resource consumption tool 330 may perform live monitoring of energy, gas, and water consumption. The resource consumption tool 330 may determine consumption at the chamber level (e.g., per wafer, per day, per week, per year, etc.), including determining the total electricity, gas, and water consumption of the chamber. The resource consumption tool 330 may determine consumption at the tool level (e.g., per day, per week, per year, etc.), including determining the total electricity, gas, and water consumption of the tool. The resource consumption tool 330 may determine individual gas consumption, including determining a breakdown of individual gas consumption (e.g., per wafer, per day, per week, per year, etc.). The resource consumption tool 330 can generate standard reports that include energy, gas, and water consumption at the chamber and tool level.

[0097] In some embodiments, the resource consumption tool 330 can determine the total electricity, gas, and water consumption of all sub-fab components (e.g., per day, per week, per year, etc.). The resource consumption tool can determine recipe-level consumption, including the total electricity, gas, and water consumption of every recipe executed in the corresponding chamber and / or tool. The resource consumption tool can determine component-level consumption, including a breakdown of the energy consumption of all energy-consuming components in a chamber. The resource consumption tool 330 can perform on-demand customized reports, including determining on-demand customized information and on-demand customized eco-efficiency reports. The resource consumption tool 330 can perform comparisons between energy consumption for various recipes and / or time points, including quantifying energy savings and energy savings opportunities using recipe optimization (e.g., using the optimization tool 326).

[0098] The research tool 332 can communicate with the digital replica tool 324 in determining the impact of one or more updates to the manufacturing equipment 304. The research tool 332 can utilize the digital replica tool 324 to generate a digital replica that includes a digital replica of the substrate manufacturing system (e.g., the manufacturing equipment 304). The research tool can receive updates to the manufacturing equipment and enable a user to investigate various alternative configurations for the equipment used, equipment settings, and process parameters associated with equipment performance, among other things. The research tool 332 can use the resource consumption tool 330 to determine environmental resource usage data corresponding to the substrate manufacturing system's execution of one or more processing steps that incorporate the updates described herein. The environmental resource usage data can be provided for display on a graphical user interface (GUI) (e.g., on the client device 350).

[0099] In some embodiments, the environmental resource usage data determined and / or predicted by other tools on the server may include a predicted environmental resource consumption and / or a predicted environmental impact associated with one of a replacement procedure or a maintenance procedure for a consumable part of the first manufacturing equipment. In some embodiments, the optimization tool 326 may determine a modification to the manufacturing process, which may include performing a corrective action associated with a component of a machine (e.g., machine 304).

[0100] A research tool 332 can perform a cost of ownership analysis associated with the manufacturing system. The cost of ownership analysis can include a comprehensive analysis of the interworking of the manufacturing system to calculate the total cost of owning and / or operating the system. The research tool 332 can calculate the cost for a customer to perform a particular manufacturing procedure. The research tool 332 can determine the cost of wafers, corresponding costs for gases used by the system, costs associated with tools in use (e.g., life degradation data), and power for the manufacturing system to perform one or more processing procedures. The cost of ownership can be calculated on a per unit basis (e.g., per wafer).

[0101] In some embodiments, machine learning system 370 further includes server machine 372, server machine 380, and / or server machine 392. Server machine 372 includes dataset generator 374, which can generate datasets (e.g., a set of data inputs and a set of target outputs) for training, validating, and / or testing machine learning model 390.

[0102] The server machine 380 includes a training engine 382, ​​a validation engine 384, and / or a test engine 386. An engine (e.g., the training engine 382, ​​the validation engine 384, and / or the test engine 386) may refer to hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, a processing unit, etc.), software (e.g., instructions executed on a processing unit, a general-purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof. The training engine 382 may be capable of training a machine learning model 390 using one or more sets of features associated with a training set from the dataset generator 374. The training engine 382 may be capable of generating one or more trained machine learning models 390, where each trained machine learning model 390 may be trained based on a distinct set of features of the training set and / or a distinct set of labels of the training set. For example, a first trained machine learning model may have been trained using resource consumption data output by the digital replica tool 324, a second trained machine learning model may have been trained using historical eco-efficiency data (e.g., eco-efficiency data 314), and so on.

[0103] The validation engine 384 may be able to validate the trained machine learning model 390 using the validation set from the dataset generator 374. The test engine 386 may be able to test the trained machine learning model 390 using the test set from the dataset generator 374.

[0104] Machine learning model 390 may refer to one or more trained machine learning models generated by training engine 382 using a training set including data inputs and, in some embodiments, corresponding target outputs (e.g., correct answers for each training input). Patterns that cluster and / or map data inputs to target outputs (correct answers) may be found in a dataset, and mappings incorporating such patterns are provided to and / or learned by machine learning model 390. Machine learning model 390 may include artificial neural networks, deep neural networks, convolutional neural networks, recurrent neural networks (e.g., long short-term memory (LSTM) networks, convLSTM networks, etc.), and / or other types of neural networks. Additionally or alternatively, machine learning model 390 may include other types of machine learning models, such as those using one or more of linear regression, Gaussian regression, random forests, support vector machines, etc.

[0105] One type of machine learning model that can be used to perform some or all of the above tasks is an artificial neural network, such as a deep neural network. Artificial neural networks typically include a feature representation component, which includes a classifier or recurrent layer that maps features to a desired output space. A convolutional neural network (CNN), for example, hosts multiple layers of convolutional filters. Lower layers perform pooling and can address nonlinearities, and multilayer perceptrons are typically added above the lower layers, with the top layer mapping features extracted by the convolutional layers to produce a decision (e.g., a classification output). Deep learning is a class of machine learning algorithms that uses a cascade of multiple layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. Deep neural networks can be trained in a supervised (e.g., classification) and / or unsupervised (e.g., pattern analysis) manner. Deep neural networks contain a hierarchy of layers, with different layers learning different levels of representation corresponding to different levels of abstraction. In deep learning, each level learns to transform its input data into a slightly more abstract, composite representation. For example, in an image recognition application, the raw input may be a matrix of pixels; the first representation layer may abstract the pixels and encode edges; the second layer may composite and encode the edge configuration; the third layer may encode higher-level shapes (e.g., teeth, lips, gums); and the fourth layer may recognize trajectories. It is noteworthy that the deep learning process can learn on its own which features are optimally placed at which levels. The "deep" in "deep learning" refers to the number of layers transforming data. More precisely, deep learning systems have a fairly deep credit assignment path (CAP). A CAP is a chain of transformations from input to output. CAP describes the potential causal relationships between input and output.For feedforward neural networks, the depth of the CAP may be the depth of the network, which may be the number of hidden layers plus 1. For recurrent neural networks, where a signal may propagate through one layer multiple times, the depth of the CAP is potentially unlimited.

[0106] Training of machine learning models can be broadly categorized as supervised learning and unsupervised learning. In embodiments, both techniques may be used to train a machine learning model. In one embodiment, training of a neural network may be achieved through a form of supervised learning, which involves feeding a training dataset consisting of labeled inputs through the network, observing its outputs, defining an error (by measuring the difference between the output and the label values), and using techniques such as deep gradient descent and backpropagation to adjust the network's weights across all its layers and nodes to minimize the error. In many applications, repeating this process across multiple labeled inputs in the training dataset results in a network that can generate correct outputs even when presented with inputs different from those present in the training dataset. In high-dimensional settings, such as large images, this generalization is achieved when a sufficiently large and diverse training dataset is provided.

[0107] For model training, a training dataset including hundreds, thousands, tens of thousands, hundreds of thousands, or even more data inputs should be used to train the dataset. In embodiments, up to thousands, tens of thousands, hundreds of thousands, or millions of historical data cases (e.g., of processes performed in processing chambers and associated labels of resource consumption) may be available to form the training dataset, where each case may include various labels of one or more types of useful information. Each case may include, for example, data indicative of processing chambers, recipes, and utilization of various resources. The above data may be processed to generate one or more training datasets for training one or more machine learning models. The machine learning models are trained based on the input processing chamber, recipe, and / or processing target information, for example, to predict processing recipes, estimate resource consumption and / or environmental efficiency, suggest modifications to recipes and / or processing chambers, etc. Such trained machine learning models can be added to an eco-efficiency dashboard and applied before, during, and / or after a process is run in a processing chamber to provide detailed information about resource consumption and eco-efficiency and ways to reduce resource consumption and / or improve eco-efficiency.

[0108] The processing logic may collect a training data set that includes past process run information with one or more associated labels (e.g., resource consumption, eco-efficiency values, recommendations for improving process recipe parameters, process recipe parameters, etc.). The training data set may additionally or alternatively be augmented. Training large neural networks typically uses tens of thousands of inputs, which are not easily obtained in many real-world applications. Data augmentation may be used to artificially increase the effective sample size.

[0109] To perform the training, processing logic inputs a training dataset into one or more untrained machine learning models. Before inputting the first input to the machine learning models, the machine learning models can be initialized. Processing logic trains the untrained machine learning models based on the training dataset to generate one or more trained machine learning models that perform the various operations described above.

[0110] Training may occur by feeding one or more data inputs to the machine learning model one at a time. Each input may include data from a past process run on a training data item from a training dataset. The machine learning model processes the inputs to generate an output. The artificial neural network includes an input layer consisting of values ​​at data points (e.g., intensity values ​​and / or height values ​​of pixels in a height map). The next layer is called the hidden layer, and each node in the hidden layer receives one or more input values. Each node contains parameters (e.g., weights) to apply to the input values. Thus, each node essentially inputs the input values ​​into a multivariate function (e.g., a nonlinear mathematical transform) to generate an output value. The next layer may be another hidden layer or an output layer. In either case, the nodes in the next layer receive the output values ​​from the nodes in the previous layer, and each node applies a weight to the value before generating its own output value. This occurs at each layer. The final layer is the output layer, where there is one node for each class, prediction, and / or output that the machine learning model may generate. Thus, the output may include predicted or estimated resource consumption for one or more resources, may include eco-efficiency values, and the like.

[0111] The processing logic can then compare the generated output with known labels that were included in the training data items. The processing logic determines an error (i.e., classification error) based on the difference between the output and the provided label. The processing logic adjusts the weights of one or more nodes in the machine learning model based on the error. An error term, or delta, can be determined for each node in the artificial neural network. Based on the error, the artificial neural network adjusts one or more of its parameters (weights for one or more inputs of the node) for one or more of its nodes. The parameters can be updated in a back-propagation manner, such that nodes in the top layer are updated first, followed by nodes in the next layer, and so on. An artificial neural network includes multiple layers of "neurons," each layer receiving values ​​as inputs from neurons in the previous layer. The parameters for each neuron include weights associated with values ​​received from each neuron in the previous layer. Adjusting the parameters can therefore include adjusting the weights assigned to each input for one or more neurons in one or more layers in the artificial neural network.

[0112] Once the model parameters are optimized, model validation may be performed to determine whether the model has improved and to determine the current accuracy of the deep learning model. After one or more rounds of training, the processing logic may determine whether a stopping criterion is met. The stopping criterion may be a target level of accuracy, a target number of images processed from the training dataset, a target amount of change to the parameters relative to one or more previous data points, combinations thereof, and / or other criteria. In one embodiment, the stopping criterion is met when at least a minimum number of data points have been processed and at least a threshold accuracy has been achieved. The threshold accuracy may be, for example, 70%, 80%, or 90% accuracy. In one embodiment, the stopping criterion is met when the accuracy of the machine learning model no longer improves. If the stopping criterion is not met, further training is performed. If the stopping criterion is met, training may be completed. Once the machine learning model is trained, a reserved portion of the training dataset may be used to test the model.

[0113] The modification identification component 394 can provide current data to the trained machine learning model 390 and execute the trained machine learning model 390 on the input to obtain one or more outputs. The modification identification component 394 can make decisions and / or perform actions from the output of the trained machine learning model 390. The output of the ML model can include confidence data indicating a confidence that the output of the ML model (e.g., modifications and optimization parameters) corresponds to a modification that, when applied, improves the overall eco-efficiency of the manufacturing process and / or manufacturing equipment selection. In some embodiments, the modification identification component 394 can perform a modification to the process recipe based on the ML model output. The modification identification component 394 can provide the output of the ML model to one or more tools on the server 320.

[0114] The confidence data may include or indicate a confidence that the output of the ML model is correct (e.g., the output of the ML model corresponds to a known label associated with a training data item). In one example, the confidence is a real number between 0 and 1, where 0 indicates no confidence that the output of the ML model is correct and 1 indicates absolute confidence that the output of the ML model is correct. In response to confidence data indicating a confidence below a threshold for a predetermined number of instances (e.g., a percentage of instances, a frequency of instances, a total number of instances, etc.), the server 320 may retrain the trained machine learning model 390.

[0115] For purposes of explanation and not limitation, aspects of the present disclosure describe training a machine learning model using process target data and / or process recipe data, inputting the current selection of manufacturing processes and / or manufacturing equipment into the trained machine learning model to determine the output of the machine learning model (predicted eco-efficiency data based on the process target, such as predicted resource consumption). In other embodiments, a heuristic or rule-based model is used to determine the output (e.g., without using a trained machine learning model).

[0116] In some embodiments, the functionality of manufacturing system 302, client device 350, machine learning system 370, data store 312, and / or server 320 may be provided by fewer machines. For example, in some embodiments, server machines 372 and 380 may be combined into a single machine, and in other embodiments, server machine 372, server machine 380, and server machine 392 may be combined into a single machine. In some embodiments, server 320, manufacturing system 302, and client device 350 may be combined into a single machine.

[0117] In general, functions described in one embodiment as being performed by manufacturing system 302, client device 350, and / or machine learning system 370 may also be performed on server 320, as appropriate, in other embodiments. Additionally, functions attributed to a particular component may be performed by a different component or by multiple components operating together. For example, in some embodiments, server 320 may receive manufacturing data and perform machine learning processing. In other examples, client device 350 may perform processing of manufacturing data based on output from a trained machine learning model.

[0118] Additionally, the functionality of a particular component may be performed by different or multiple components operating together, and may be accessed as a service provided to other systems or devices by one or more of server 320, manufacturing system 302, or machine learning system 370 via an appropriate application programming interface (API).

[0119] In some embodiments, a "user" may be represented as a single individual. However, other embodiments of the present disclosure encompass a "user" being an entity controlled by multiple users and / or automated sources. For example, a collection of individual users formed as a group of administrators may be considered a "user."

[0120] FIG. 4 illustrates an exemplary digital replica according to some embodiments of the present disclosure. The digital replica 400 can include a digital twin of a selected manufacturing system, e.g., a digital replica of a manufacturing system including the same chambers, valves, gas supply lines, materials, and chamber components. The digital replica 400 can receive manufacturing equipment process data as input, which can include first sensor data 402-404 output by integrated sensors in the processing chambers and second sensor data 406-408 output by external sensors that are not components of the processing chambers. The input can further include a process recipe for the manufacturing system including the processing chambers, and / or can output physical conditions of the manufacturing system. In some embodiments, the digital replica 400 includes physics-based models that can incorporate various physical relationships, such as thermodynamics, fluid dynamics, energy conservation, gas laws, mechanical systems, transportation, and delivery. The digital replica 400 processes the input data and generates output 410. The output can include one or more physical conditions of the processing chambers and / or other systems or devices. The output may additionally or alternatively include environmental resource usage data.

[0121] In one example, the digital replica 400 may receive as input a first gas flow of a first gas, a second gas flow of a second gas, and a third gas flow of a third gas, as well as a first process recipe. The digital replica may use a physics-based model to estimate the amount of energy leaving the chamber due to the gas flows. For example, the model may determine the temperature of the exhaust and the total energy flow rate through the exhaust. In another example, the same digital replica 400 may output predicted eco-efficiency data, such as the predicted consumption of environmental resources. The digital replica may identify associated resource consumption and identify optimization suggestions to improve energy conservation.

[0122] In some embodiments, the digital replica 400 can determine exhaust for one or more gas panels or gas boxes containing gases used at one or more locations throughout the manufacturing system. For example, each gas box can use a dedicated exhaust at negative pressure to efficiently exhaust gases (e.g., to prevent toxins from entering undesired locations in the manufacturing facility or manufacturing system), such as in the event of a gas line leak or more general failure. The digital replica can be part of a digital twin that utilizes information about the likely types and quantities of gases in the gas box to determine adjustments to the exhaust flow necessary for proper final disposal of the gas (e.g., venting leaks). The exhaust flow rate can be determined taking into account exhaust pressure and flow. The exhaust flow can include determining related parameters to optimize eco-efficiency while maintaining minimum safety thresholds and / or standards.

[0123] In some embodiments, the digital replica 400 can indicate the temperature of the exhaust and the total energy flow rate through the exhaust based on the heating within the processing chamber. For example, a processing chamber can include one or more processing devices, such as a substrate pedestal, during a substrate processing procedure. Excess heat from within the chamber can be removed through the exhaust. The operation of the pedestal can be modified to reduce the heat lost through the exhaust. Several methods have been reported for controlling heat transfer within a heat transfer assembly, such as a pedestal for supporting a substrate. The pedestal includes both a heating element and a cooling element that removes excess heat by circulating a cooling medium, such as a gas or liquid coolant, within the pedestal or between the substrate and the pedestal. When the temperature of the substrate rises above a set range during processing, the heating element is turned off and the cooling element is activated to remove the excess heat, thereby controlling the temperature. One or more parameters associated with the process are used as inputs to the digital replica 400 to determine how much excess heat is lost through the exhaust.

[0124] In some embodiments, the digital replica 400 may show the energy flow and / or chemicals, including lost precursors or reaction by-products, exiting an abatement or scrubber system. For example, gaseous exhaust streams from the manufacture of electronic materials, devices, products, solar cells, and memory products (hereinafter "electronic devices") may contain a variety of chemical compounds, organic compounds, oxidizers, decomposition products of photoresists and other reagents, and other gases and airborne particulates that may be desirable to remove from the exhaust stream before the exhaust stream is vented to the atmosphere from the processing facility.

[0125] The effluent stream to be abated may include chemical species generated by the electronic device manufacturing process and / or chemical species transferred to the electronic device manufacturing process and passed through the processing chambers unchanged. As used herein, the term "electronic manufacturing process" is intended to be broadly interpreted to include any and all process and unit operations in the manufacture of electronic devices, as well as any operations involving the processing or fabrication of materials used in or produced by electronic device and / or LCD manufacturing equipment, and any operations performed in connection with electronic device and / or LCD manufacturing equipment that do not involve active manufacturing (e.g., conditioning processing equipment, purging chemical supply lines in preparation for operation, etching cleansing processing tool chambers, abatement of toxic or harmful gases from effluents generated by electronic device and / or LCD manufacturing equipment, etc.).

[0126] In some embodiments, digital replica 400 accounts for the exhaust flow of leaked gases or as part of a cleaning procedure. For example, gases are periodically flushed from a manufacturing asset to extend the asset's life, improve product performance, or prepare the product for a different function it is tasked to perform. Digital replica 400 can determine the environmental consumption (e.g., energy consumption, gas consumption) associated with performing this purge procedure. For example, digital replica 400 can show the energy and / or gas consumption used to flush the system (e.g., constantly supplying a gas flow to the system to maintain dynamic gas movement within the system). Digital replica 400 can show how adjusting one or more gas flow rates (e.g., purge gas) within the processing system changes the energy and / or gas consumption.

[0127] In some embodiments, digital replica 400 can utilize a process recipe to determine which gases are entering the process chamber, which reactions are occurring on a substrate disposed in the process chamber, and how the gases are utilized by the substrate reactions. Digital replica 400 can also determine which gases and how much of each remain after reactions occur on the substrate surface. Digital replica 400 can also determine the amount and type of gas lost during abatement. Digital replica 400 can also determine the end by-products of the abatement and their overall impact on the environment.

[0128] In some embodiments, one or more substrate processing procedures may require a constant gas flow into and / or out of the processing chamber to process a substrate that meets target process result conditions. The substrate processing system can perform a stable gas flow procedure by implementing one or more transitions from flow-to-vent to flow-to-chamber to reduce transient airflow due to turning airflow into and off the chamber. For example, an initial gas flow can be set to an initial value and vented, and once the gas flow stabilizes, the vented air can be introduced into the processing chamber to provide a stable gas flow to the chamber. The digital replica 400 can determine gas consumption (e.g., gas lost through venting) as a result of the above process. For example, the digital replica can identify the transition time and the amount of gas lost through venting during the transition period of starting or stopping the gas flow. The digital replica can determine optimization for the transition between venting and introducing gas into the chamber. Optimizing the transition time can reduce gas loss through venting while identifying the time for the gas to reach a steady state. In some embodiments, the gas flow transition frequency can be determined based on process result requirements, for example, the gas flow transition times can be determined (e.g., optimized) to include flow rates that do not adversely affect process results in the corresponding processing chamber.

[0129] Digital replica 400 can be used to predict eco-efficiency data associated with one or more operating states of a physical asset of a manufacturing system. As an example, digital replica 400 can receive data associated with one or more operating states of a physical asset of a manufacturing system. For example, digital replica 400 can receive reduced power data, sleep mode data, shared operating mode data, and / or process recipe data that indicates one or more process steps to be performed by the manufacturing system, which data are represented by digital replica 400.

[0130] Energy savings can occur when one or more physical assets operate in various operating states during processing and idle times. For example, various elements of sub-fab equipment may not be needed during various steps of a manufacturing process and, therefore, may be in a sleep, idle, hibernate, or off state depending on how soon the elements are likely to be needed. Examples of power-saving, low-power states include idle, sleep, and hibernation states. The primary differences between the three power-saving states are in duration and energy consumption. Deeper levels of idle mode energy savings, such as sleep or hibernation, require longer times to return from a power-saving mode to full production without impacting the quality or yield of the manufacturing process. Restoring processing chambers and associated sub-fab equipment to BKM (best known method) temperatures and pressures can take seconds, minutes, or hours depending on the degree of deviation from BKM chamber conditions associated with the power-saving state of the sub-fab equipment and processing chambers. The idle state typically lasts for a few seconds, the sleep state typically lasts for a few minutes, and the hibernation state typically lasts for a few hours.

[0131] Digital replica 400 can identify one or more operational / power states of the physical assets of a manufacturing system to determine the effect of using that power state in a given scenario (e.g., system hardware configuration, subsystem hardware configuration, processing of one or more process recipes, execution of a particular scheduled process, etc.) For example, digital replica 400 can be part of a digital twin that determines the effect of such power states and scenarios for idle or full power or modulation before actually performing power adjustments on the manufacturing system.

[0132] Processing tools and associated fabrication systems may have a variety of different power settings based on operational needs. For example, there may be a power setting in which the processing tool is in an "off" state while various airflow and abatement systems are fully operational and perform shutdown operations after completing a fabrication step. For purposes of this application, the term "low power configuration" refers to any state in which one or more elements of a processing tool and / or fabrication system sub-fab are instructed by one or more controllers to operate in a power-saving mode, for example, at various levels of energy consumption between steps of a particular processing recipe, in a non-productive idle mode such as the idle, sleep, and hibernation states described above, or in an off state.

[0133] In some embodiments, one or more support assets can provide support functions to one or more other physical assets. For example, pumping two processing chambers can be performed by one pump. Utilizing support assets to alternate operations between two physical assets can save energy and reduce overall environmental costs.

[0134] Digital replica 400 can identify environmental resource consumption data associated with one or more physical assets operating in one or more corresponding operational modes. Digital replica 400 can provide recommendations for reducing environmental consumption costs by recommending that one or more physical assets enter a reduced power state, a sleep mode state, a hibernation state, and / or utilize shared operational mode data during periods when the corresponding tools are idle or in a state when demand on the physical asset is low.

[0135] In another example, digital replica 400 may be configured to determine eco-efficiency data associated with performing preventative maintenance (PM) and / or cleaning of physical assets of a manufacturing system. Digital replica 400 may receive purge gas data, cleaning process data, preventative maintenance data, chamber recovery data, and / or process recipes to determine environmental resource consumption.

[0136] Substrate processing can include a series of steps to create electrical circuits in semiconductors, e.g., silicon wafers, according to a circuit design. These processes can be performed in a series of chambers. Successful operation of modern semiconductor tooling can facilitate a steady flow of wafers moving from one chamber to another during the formation of electrical circuits in the wafers. In many substrate processing processes, processing chamber conditions can deteriorate, resulting in processed substrates that do not meet desired conditions or processing results (e.g., critical dimensions, process uniformity, thickness dimensions, etc.).

[0137] The cleaning process data may indicate one or more parameters associated with the cleaning process, such as cleaning duration, frequency, and / or etchant flow. The cleaning process may utilize specific environmental resources, such as cleaning materials, precursors, etchants, and / or other substances utilized to perform the cleaning procedure. For example, the cleaning procedure may be performed at a certain cadence or frequency (e.g., after a certain amount of wafers have been processed) for process results for future substrates to meet threshold conditions (e.g., process uniformity, critical dimensions, etc.). The frequency of processing chamber cleaning may be adjusted (e.g., optimized) to identify a cleaning frequency where the results of substrates processed by a chamber operating on that cleaning frequency schedule still meet threshold conditions (e.g., minimum process result requirements). For example, multi-wafer cleaning procedures may be implemented that conserve environmental resources, such as cleaning materials, precursors, etchants, and / or other substances utilized to perform the cleaning procedure. The digital replica 400 may receive the cleaning data to determine cleaning optimizations, such as updating cleaning duration, frequency, cleaning agent usage, etc.

[0138] The preventive maintenance data indicates one or more of the type, frequency, duration, etc. of one or more preventive maintenance procedures associated with one or more physical assets of the manufacturing system. The preventive maintenance procedures (e.g., cleaning the chamber) are often used as part of a chamber recovery process to return the processing chamber to a state suitable for entering a substrate processing production mode (e.g., processing a large volume of substrates). A recovery procedure is often used following the preventive maintenance procedure to prepare the chamber (e.g., to "warm up" the chamber) for the production mode.

[0139] Chamber recovery data indicates one or more of the type, frequency, duration, etc., of one or more chamber recovery procedures associated with one or more physical assets of a manufacturing system. A common recovery procedure conventionally utilized is process chamber seasoning. Chamber seasoning is a procedure that involves processing a series of substrates (e.g., blank silicon wafers) to restore chamber conditions (e.g., coating the chamber walls) suitable for production substrate processing (e.g., substrates processed in the chamber having process results that meet desired threshold criteria). After chamber seasoning, the chamber can be operated in a production mode for a specified period of time until re-performed preventive maintenance and further chamber seasoning are required or otherwise recommended to restore the processing chamber conditions.

[0140] The purge gas data may indicate the type, amount, frequency, flow rate, and duration of the purge gas. The digital replica can determine the effect of changing one or more process parameters related to the purge gas utilized. For example, the digital replica 400 can predict environmental resource consumption updates based on switching to a purge procedure using alternative purge gas species, such as H2, N2, and CDA.

[0141] In another example, digital replica 400 may be configured to determine eco-efficiency data associated with one or more operating conditions of physical assets of a manufacturing system. Digital replica 400 may receive coolant loop configuration data, process chilled water (PCW) data, ambient air data, and / or process recipes from which to determine environmental resource consumption data.

[0142] Processing chambers used during substrate processing typically contain several internal components that are repeatedly heated and cooled during and after processing. In some instances, such components are cooled to approximately room temperature, for example, when periodic inspection or maintenance is required after a process has been performed in the processing chamber. For example, in a temperature-controlled component, such as a processing chamber showerhead having coolant channels, to cool the component from its typical operating temperature (e.g., about 90 degrees Celsius), a heat source heating the component can be turned off and coolant can be flowed through the coolant channels to remove heat from the component.

[0143] The coolant loop configuration data indicates one or more configurations of one or more coolant loops configured to remove heat from one or more physical assets of the manufacturing system. One or more coolant loops can operate in parallel, with multiple loops cooling a shared area of ​​the physical asset. One or more coolant loops can sequentially cool multiple physical assets. The process cooling water (PCW) data indicates one or more parameters of the coolant (e.g., process cooling water (PCW)), such as the type, flow rate, and temperature of the coolant. The digital replica can include a heat flow model that indicates where energy is transferred within the environment of the manufacturing system utilizing one or more coolant loops. The digital replica 400 can identify modifications to physical assets (e.g., chambers, chamber walls, chamber systems) that direct heat to the cooling loops and the associated environmental efficiencies achieved by directing heat to the cooling loops. The digital replica 400 can further determine the impact of PCW adjustments on process results while adjustments are being made. PCW adjustments can include changing the flow rate within the cooling loops to change the heat exchange within the physical assets of the manufacturing system.

[0144] 5 is an exemplary diagram of process parameter limits 500 for a manufacturing process, according to some embodiments of the present disclosure. Various manufacturing process steps can include process parameter limits 500, which indicate a process parameter window 510, or a set of values ​​(e.g., a combination of values) for a corresponding set of parameters, that, when met, achieve a result that meets a threshold condition (e.g., a minimum quality condition, a target condition, etc.). For example, the process parameter window 510 can include a first parameter 502 (e.g., a first flow rate of a first gas) and a second parameter 504 (e.g., a temperature of the gas). To perform a manufacturing process and meet a threshold condition (e.g., a minimum quality standard, a statistical process control (SPC) limit, a specification limit, a substrate processing goal, etc.), a process parameter value window 510 is determined that identifies a combination of parameter values ​​that will result in a product that is likely to meet the threshold condition. As shown in FIG. 5, the process parameter window 510 includes a lower limit 506B and an upper limit 506A for the first parameter 502, and a lower limit 508B and an upper limit 508A for the second parameter.

[0145] Optimizations identified by the manufacturing process system (e.g., using recommendations 284, data analyzer 226, etc.) can include determining an eco-optimized process parameter window 512 within the process parameter window 510 that results in a manufacturing process consuming a reduced amount of resources compared to process parameter values ​​outside the eco-optimized process parameter window 512.

[0146] 5 illustrates a simplified process parameter window 510 and an eco-optimized process parameter window 512 that depends on only two parameters 502, 504. Both the process parameter window 510 and the eco-optimized process parameter window 512 form simple rectangles. The process parameter windows may include more than two parameters and may include more diverse parameter dependencies. For example, non-linear, physically-based, statistical, and / or empirical relationships between parameters may result in non-linear and eco-optimized process parameter windows.

[0147] 6 is a flowchart of a method 600 for generating a training dataset for training a machine learning model to perform cooling parameter estimation, according to an embodiment of the present disclosure. Method 600 is performed by processing logic, which may include hardware (circuitry, dedicated logic, etc.), software (such as running on a general-purpose computer system or a dedicated machine), firmware, or some combination thereof. In one embodiment, method 600 may be performed by a computer system, such as computer system architecture 300 of FIG. 3. In other or similar embodiments, one or more steps of method 600 may be performed by one or more other machines not depicted in the figure. In some aspects, one or more steps of method 600 may be performed by dataset generator 374 of machine learning system 370 described with respect to FIG. 3.

[0148] For ease of explanation, method 600 is shown and described as a series of acts. However, acts in accordance with the present disclosure may occur in various orders and / or concurrently, with other acts not shown or described herein. Moreover, not all of the acts shown may be performed to implement method 600 in accordance with the disclosed subject matter. Furthermore, those skilled in the art will understand and appreciate that method 600 could alternatively be represented as a series of interrelated states or events via a state diagram.

[0149] At block 610, processing logic initializes a training set T to an empty set (e.g., {}). At block 612, processing logic acquires substrate processing recipe data (e.g., process recipe setpoint data, process knob setpoint data, process pressure setpoint data, process temperature setpoint data, etc.) associated with processing substrates in a processing chamber of the manufacturing system. The process recipe data may include and / or supplement historical process recipe data (e.g., recipe data collected over time). In some embodiments, processing logic further acquires sensor data (e.g., temperature sensor data, pressure sensor data, energy sensor data, etc.) and / or predicted sensor data (e.g., output from one or more additional models) associated with processing substrates in the processing chamber according to the process recipe and / or according to other process recipes.

[0150] At block 614, processing logic obtains environmental resource usage information. The environmental resource usage information may include information associated with consumption of resources such as chemical precursors, gases, water, energy, etc. The environmental resource usage information may include and / or supplement historical environmental resource usage information (e.g., historical consumption data, etc.).

[0151] At block 616, processing logic generates training inputs based on the process recipe data and / or sensor data obtained at block 612. In some embodiments, the training inputs may include a normalized set of recipe data.

[0152] At block 618, processing logic may generate a target output based on the environmental resource usage information obtained at block 614. The target output may correspond to environmental resource usage metrics for a process recipe being executed in the process chamber.

[0153] At block 620, processing logic generates an input / output mapping, which refers to training inputs that include or are based on process recipe data and target outputs for the training inputs, where the target outputs specify predicted environmental resource consumption, and the training inputs are associated with (or mapped to) the target outputs. At block 622, processing logic adds the input / output mapping to a training set T.

[0154] At block 624, processing logic determines whether training set T includes a sufficient amount of training data to train the machine learning model. In some embodiments, the sufficiency of training set T may be determined solely based on the number of input / output mappings in the training set; however, in some other embodiments, the sufficiency of training set T may be determined based on one or more other criteria (e.g., a measure of the diversity of the training examples) in addition to or instead of the number of input / output mappings. In response to determining that training set T includes a sufficient amount of training data to train the machine learning model, processing logic provides training set T for training the machine learning model. In response to determining that the training set does not include a sufficient amount of training data to train the machine learning model, method 600 returns to block 612.

[0155] In block 626, processing logic provides a training set T for training the machine learning model. In some embodiments, the training set T is provided to the machine learning system 370 and / or the training engine 382 of the server machine 380 to perform the training. In the case of a neural network, for example, input values ​​(e.g., recipe data and / or cooling parameter data) of a predetermined input / output mapping are input to the neural network, and output values ​​of the input / output mapping are stored in output nodes of the neural network. The connection weights, layers, and / or hyperparameters in the neural network are then adjusted according to a learning algorithm (e.g., backpropagation, etc.), and the procedure is repeated for other input / output mappings in the training set T. After block 626, the machine learning model 390 can be used to provide predicted environmental resource usage (e.g., predicted data indicative of resource consumption) for process recipe steps performed in the process chamber.

[0156] 7 is a flow chart illustrating an embodiment of a method 700 for training a machine learning model to estimate cooling parameter values ​​for a process recipe executed in a processing chamber, according to aspects of the present disclosure. Method 700 may be performed by processing logic, which may include hardware (circuitry, dedicated logic, etc.), software (such as running on a general-purpose computer system or a dedicated machine), firmware, or some combination thereof. In one embodiment, method 700 may be performed by a computer system, such as computer system architecture 300 of FIG. 3. In other or similar embodiments, one or more steps of method 700 may be performed by one or more other machines not depicted in the figures. In some aspects, one or more steps of method 700 may be performed by training engine 382 of machine learning system 370 described with respect to FIG. 3.

[0157] For ease of explanation, method 700 is shown and described as a series of acts. However, acts in accordance with the present disclosure may occur in various orders and / or simultaneously, with other acts not shown or described herein. Moreover, not all of the acts shown may be performed to implement method 700 in accordance with the disclosed subject matter. Furthermore, those skilled in the art will understand and appreciate that method 700 could alternatively be represented as a series of interrelated states or events via a state diagram.

[0158] At block 702 of method 700, processing logic collects a training data set, which may include data from multiple substrate processing recipes (e.g., process recipe settings, process control knob settings, etc.). The training data set may further include sensor data associated with the performance of the substrate processing recipes. Each data item in the training data set may include one or more labels. The data items in the training data set may include input-level labels that indicate environmental resource usage associated with the substrate processing recipe. For example, some data items may include labels for resource usage (e.g., usage of one or more resources, such as chemical precursors, gases, energy, etc.) associated with the process recipe.

[0159] At block 704, data items from the training dataset are input to an untrained machine learning model. At block 706, the machine learning model is trained based on the training dataset to generate a trained machine learning model that estimates environmental resource usage (e.g., resource consumption, etc.) for processing a substrate in a processing chamber according to a process recipe. The machine learning model may also be trained to output one or more other types of predictions, classifications, decisions, etc.

[0160] In one embodiment, at block 710, training data items are input to the machine learning model. The input may include substrate processing recipe data (e.g., a substrate processing recipe) indicating one or more process recipe settings. In some embodiments, the data may be input as a feature vector. At block 712, the machine learning model processes the input to generate an output. The output may include environmental resource usage (e.g., consumption of one or more resources, etc.). The environmental resource usage may be an estimated environmental resource usage for processing a substrate according to the process recipe.

[0161] At block 714, processing logic compares the output predicted environmental resource usage data to known environmental resource usage associated with the input. At block 716, processing logic determines an error based on the difference between the output and the known environmental resource usage. At block 718, processing logic adjusts weights of one or more nodes in the machine learning model, one or more layers in the machine learning model, and / or one or more hyperparameters of the machine learning model based on the error.

[0162] At block 720, processing logic determines whether a stopping criterion is met. If the stopping criterion is not met, the method returns to block 710, where another training data item is input to the machine learning model. If the stopping criterion is met, the method proceeds to block 725, where training of the machine learning model is complete.

[0163] In one embodiment, one or more ML models are trained for application across multiple processing chambers (which may be processing chambers of the same type or model). The trained ML models can then be further fine-tuned for use with a specific instance of a processing chamber. Further tuning can be performed using additional training data items, including substrate processing recipes executable in the processing chamber of interest. Such tuning can account for inconsistencies between chambers and / or the specific hardware process kits of some processing chambers. Additionally, in some embodiments, after processing chamber maintenance and / or one or more changes to the processing chamber hardware, further training is performed to adjust the ML models for the processing chamber.

[0164] FIG. 8A is a flow diagram of a method 800A for obtaining recommendations for processing a substrate, according to some embodiments of the present disclosure. Method 800A may be performed by processing logic, which may include hardware (circuitry, dedicated logic, etc.), software (such as running on a general-purpose computer system or a dedicated machine), firmware, or some combination thereof. In one embodiment, method 800A may be performed by a computer system, such as computer system architecture 300 of FIG. 3. In other or similar embodiments, one or more steps of method 800A may be performed by one or more other machines not depicted in the figures. In some aspects, one or more steps of method 800A may be performed by eco-efficiency module 129 described with respect to FIG. 1. In some aspects, one or more steps of method 800A may be performed by one or more components of server 320 described with respect to FIG. 3.

[0165] For ease of explanation, method 800A is shown and described as a series of operations. However, operations in accordance with the present disclosure may occur in various orders and / or concurrently, with other operations not shown or described herein. Moreover, not all of the illustrated operations may be performed to implement method 800A in accordance with the disclosed subject matter. Furthermore, those skilled in the art will understand and appreciate that method 800A may alternatively be represented as a series of interrelated states or events via a state diagram.

[0166] At block 802, processing logic receives a process recipe including process recipe setting data. The process recipe may be for processing a substrate in a processing chamber of the manufacturing system. In some embodiments, processing logic receives multiple process recipes, each process recipe including recipe setting data. For example, a first set of process recipe setting data may indicate settings for a first process recipe, and a second set of process recipe setting data may indicate settings for a second process recipe. Processing logic may receive both a first data set and a second data set. Similarly, a third set of process recipe setting data may indicate settings for a third process recipe. Processing logic may receive the first set, the second set, and the third set. In some embodiments, the process recipe setting data includes predicted setting data output from a model (e.g., process model 262 of FIG. 2B ) configured to predict recipe setting data based on input process targets. The recipe setting data may indicate one or more process recipes that can be executed (e.g., in a processing chamber) to process a substrate that meets the process targets.

[0167] The processing logic may optionally receive sensor data associated with substrate processing in the processing chamber. In some embodiments, the sensor data is indicative of conditions (e.g., temperature, pressure, precursor flow rates, gas flow rates, etc.) within the processing chamber during substrate processing. In some examples, the sensor data is indicative of conditions across an operating range of the processing chamber. The sensor data may be associated with processing a first substrate according to a first process recipe, processing a second substrate according to a second process recipe, and / or processing a third substrate according to a third process recipe, etc. In some embodiments, the sensor data is indicative of physical limits of processing chamber conditions and / or process recipes executed in the processing chamber. For example, the sensor data may be indicative of normal ranges of conditions within the processing chamber (e.g., normal temperature ranges, normal pressure ranges, etc.) during substrate processing according to one or more process recipes.

[0168] At block 804, processing logic inputs the process recipe received at block 802 into one or more machine learning models. In some embodiments, the one or more machine learning models are trained to predict eco-efficiency data (e.g., chamber model 264 of FIG. 2B). In some embodiments, the one or more machine learning models are trained using training input data that includes historical process recipe data (e.g., recipe setpoint data, recipe setpoints, etc.) and training target output data that includes historical environmental resource usage data (e.g., resource consumption data, etc.).

[0169] In some embodiments, the one or more machine learning models include a “chain” of machine learning models. For example, a first machine learning model may be trained to output first predicted data, and a second machine learning model may be trained using the first predicted data to output second predicted data. In some embodiments, the first machine learning model is trained using training input data including past process recipes and training target output data including past sensor data (e.g., past sensor measurement data associated with substrate processing in a process chamber). At block 805A, one or more process recipes may be input to the first machine learning model to obtain predicted measurements (e.g., predicted measurement data, predicted sensor measurement data, etc.) corresponding to the input process recipes.

[0170] In some embodiments, a second machine learning model is trained using the predicted measurements output from the first machine learning model. The second machine learning model can be further trained using training input data including historical processing recipes and training output data including historical eco-efficiency data. The second machine learning model can be trained to output predicted eco-efficiency data (e.g., predicted environmental resource usage data). At block 805B, one or more processing recipes and the predicted measurements output from the first machine learning model can be input to the second machine learning model to determine the predicted environmental resource usage data.

[0171] In some embodiments, at block 804, multiple process recipes are input to one or more machine learning models, each process recipe including a corresponding set of recipe setting data.

[0172] In some embodiments, one or more trained machine learning models are trained to output predicted environmental resource usage data. The predicted environmental resource usage data may indicate environmental resource consumption associated with processing a substrate in a processing chamber according to a process recipe. For example, the predicted environmental resource usage data may indicate predicted consumption of one or more resources used to process a substrate according to a process recipe. In some embodiments, the predicted environmental resource usage data may indicate consumption of a particular resource (e.g., chemical precursor, water, etc.) and / or consumption of multiple resources. In some embodiments, the predicted environmental resource usage data includes multiple sets of environmental resource usage data. For example, the one or more trained machine learning models may output a set of environmental resource usage data for each corresponding process recipe input into the one or more models. In such an example, the set of environmental resource usage data may indicate that the corresponding process recipe is more environmentally efficient than other process recipes. In particular, the set of environmental resource usage data may indicate that the corresponding process recipe uses fewer resources to execute the recipe compared to other recipes input into the one or more models. In some examples, the one or more models output predicted first environmental resource usage data corresponding to the first process recipe and predicted second environmental resource usage data corresponding to the second process recipe. In some embodiments, the environmental resource usage data may include time series data from which resource consumption can be determined.

[0173] At block 808, processing logic determines a recommendation associated with processing a substrate according to a process recipe based on a comparison of the predicted first environmental resource usage data and the predicted second environmental usage data. In some embodiments, processing logic compares predicted resource consumption associated with the process recipe (e.g., as indicated by the first environmental resource usage data) with predicted resource consumption associated with other process recipes (e.g., as indicated by the second environmental resource usage data). The predicted resource consumption may indicate that one process recipe is more eco-efficient than other process recipes. This may be determined by the comparison. In some embodiments, processing logic compares multiple sets of environmental resource usage data and determines the most eco-efficient process recipe. For example, processing logic may determine the most eco-efficient process recipe from a selection of multiple process recipes based on corresponding resource consumption predicted (e.g., by a model).

[0174] In some embodiments, the recommendation indicates that the most eco-efficient process recipe should be implemented to process the substrate to meet the process goals (e.g., goals for the substrate being processed). For example, the recommendation may be a selection of the most eco-efficient process recipe selected from a plurality of process recipes (e.g., whose set points are received in data block 802). In some embodiments, the recommendation indicates a modification of the process recipe to make it more eco-efficient. For example, the recommendation may indicate a change to recipe set points to reduce the resource consumption of the process recipe. In some examples, the processing logic may determine the modification using predicted resource consumption data corresponding to other process recipes. The recommendation may be to optimize the eco-efficiency of the process recipe (e.g., as indicated by a lower predicted resource consumption) by changing recipe set points to more closely match other process recipes with higher predicted eco-efficiency.

[0175] At block 810, the processing logic outputs a recommendation associated with the process recipe. In some embodiments, the recommendation is output to a system controller for implementation in processing the substrate. For example, the system controller may execute the process recipe indicated by the recommendation (e.g., the most eco-efficient process recipe as a result of the comparison) for processing the substrate in the processing chamber. In other examples, the system controller may modify the process recipe in accordance with the recommendation to create a more eco-efficient process recipe and / or to improve the eco-efficiency of the process recipe.

[0176] FIG. 8B is a flow diagram of a method 800B for obtaining predicted process recipe settings data according to some embodiments of the present disclosure. Method 800B may be performed by processing logic, which may include hardware (circuitry, dedicated logic, etc.), software (such as running on a general-purpose computer system or a dedicated machine), firmware, or some combination thereof. In one embodiment, method 800B may be performed by a computer system, such as computer system architecture 300 of FIG. 3. In other or similar embodiments, one or more steps of method 800B may be performed by one or more other machines not depicted. In some aspects, one or more steps of method 800B may be performed by eco-efficiency module 129 described with reference to FIG. 1. In some aspects, one or more steps of method 800B may be performed by one or more components of server 320 described with reference to FIG. 3. In some embodiments, method 800B is performed in conjunction with method 800A.

[0177] For ease of explanation, method 800B is shown and described as a series of acts. However, acts in accordance with the present disclosure may occur in various orders and / or concurrently, with other acts not shown or described herein. Moreover, not all of the acts shown may be performed to implement method 800B in accordance with the disclosed subject matter. Moreover, those skilled in the art will understand and appreciate that method 800B may alternatively be represented as a series of interrelated states or events via a state diagram.

[0178] At block 822, processing logic receives target data including a target substrate condition associated with the processed substrate. In some embodiments, the target condition indicates one or more characteristics of the target processed substrate (e.g., surface features, coatings, etc.). In some examples, the target substrate processing data indicates specifications for the processed substrate. The specifications may indicate thresholds for acceptable processed substrates.

[0179] At block 824, processing logic inputs the target conditions into a model (e.g., process model 262 of FIG. 2B ). The model may include one or more additional models in addition to the models described with respect to method 800A of FIG. 8A . In some embodiments, the model is an additional trained machine learning model. The additional machine learning model may be trained (to form the additional trained machine learning model) using training inputs including historical process target data, including historical target conditions, and training target output data, including historical process recipes (e.g., including historical process recipe setpoint data). In some examples, the historical target conditions include multiple historical target conditions. The historical conditions may represent one or more historical characteristics of previously targeted processed substrates. In some examples, the historical target conditions represent historical specifications for previously processed substrates. The historical specifications may represent historical thresholds for acceptable previously processed substrates. In some embodiments, the additional machine learning model is trained to output a predicted process recipe associated with the process targets input into the model. For example, the process targets can be input into an additional trained machine learning model, which outputs one or more predicted process recipes that will produce processed substrates that meet the process targets.

[0180] At block 826, the processing logic receives a first process recipe and a second process recipe as output from the model. The first process recipe and / or the second process recipe may each correspond to the target substrate processing data received at block 822. In some embodiments, the model outputs additional process recipes (e.g., additional sets of process recipe settings data). Each of the process recipes output by the model, when executed, is capable of producing a substrate that meets the target substrate state. For example, the model may output a first process recipe and a second process recipe. When a substrate is processed in a processing chamber according to either the first process recipe or the second process recipe, the processed substrate will meet the target conditions indicated by the target conditions. In some embodiments, the one or more process recipes received at block 826 correspond to the process recipe received at block 802 of method 800A.

[0181] FIG. 8C is a flow diagram of a method 800C for obtaining predicted process recipe setpoint data according to some embodiments of the present disclosure. Method 800C may be performed by processing logic, which may include hardware (circuitry, dedicated logic, etc.), software (such as running on a general-purpose computer system or a dedicated machine), firmware, or some combination thereof. In one embodiment, method 800C may be performed by a computer system, such as computer system architecture 300 of FIG. 3. In other or similar embodiments, one or more steps of method 800C may be performed by one or more other machines not depicted. In some aspects, one or more steps of method 800C may be performed by eco-efficiency module 129 described with reference to FIG. 1. In some aspects, one or more steps of method 800C may be performed by one or more components of server 320 described with reference to FIG. 3. In some embodiments, method 800C is performed in conjunction with method 800A.

[0182] For ease of explanation, method 800C is shown and described as a series of operations. However, operations in accordance with the present disclosure may occur in various orders and / or concurrently, with other operations not shown or described herein. Moreover, not all of the illustrated operations may be performed to implement method 800C in accordance with the disclosed subject matter. Furthermore, those skilled in the art will understand and appreciate that method 800C may alternatively be represented as a series of interrelated states or events via a state diagram.

[0183] At block 832, processing logic trains a first machine learning model to output predicted measurement data based on the process recipe input to the first machine learning model. In some embodiments, the first machine learning model is the first model in a “chain” of machine learning models. The first machine learning model may be trained using training input data including past process recipes (e.g., past process recipe setpoint data). The first machine learning model may be further trained using training target output data including past measurement data. The past measurement data may include sensor data collected during processing of substrates in one or more process chambers. In some examples, the past measurement data may include measurements of current, voltage, power, flow rate, pressure, concentration, velocity, acceleration, and / or temperature. Similarly, the predicted measurement data that the first machine learning model is trained to output may include predicted measurement values ​​of current, voltage, power, flow rate, pressure, concentration, velocity, acceleration, and / or temperature. In some embodiments, the predicted measurement values ​​include predicted time series data of the measurement values.

[0184] At block 834, processing logic trains a second machine learning model to output predicted environmental resource usage data (e.g., predicted eco-efficiency data). In some embodiments, the second machine learning model is the second model in a "chain" of machine learning models. In some embodiments, the second machine learning model is trained using training input data including predicted measurements output from the first machine learning model. In some embodiments, the second machine learning model is trained using predicted time series measurements output from the first machine learning model. The training input data may further include historical processing recipes used to train the first machine learning model. In some embodiments, the second machine learning model is trained using training target output data including historical environmental resource usage data (e.g., historical eco-efficiency data). Training the second machine learning model using the output of the first machine learning model can improve the accuracy of the predicted environmental resource data output by the second machine learning model. In some examples, using intermediate output (e.g., predicted measurement data) from the first machine learning model to train a second machine learning model can provide greater accuracy in the final output (e.g., predicted environmental resource usage data) from the second machine learning model compared to predicting the final output using a single model. In some embodiments, the first machine learning model represents processing chamber behavior that closely tracks changes in process recipe set points, and the second machine learning model additionally represents processing chamber behavior that does not closely track changes in process recipe set points.

[0185] Optionally, in some embodiments, the processing logic trains a third machine learning model to output further predicted environmental resource usage data (e.g., further predicted eco-efficiency data). In some embodiments, the third machine learning model is a third model in a "chain" of machine learning models. In some embodiments, the third machine learning model is trained using training input data including predicted measurements output from the first machine learning model. In some embodiments, the third machine learning model is trained using predicted time series measurements output from the first machine learning model. The training input data may further include historical processing recipes used to train the first machine learning model. In some embodiments, the third machine learning model is trained using training target output data including historical environmental resource usage data (e.g., historical eco-efficiency data) and predicted environmental resource usage data output from the second machine learning model. By training a second machine learning model using the output of the first machine learning model and / or the output of the second machine learning model, the accuracy of the predicted environmental resource data output by the third machine learning model can be increased (e.g., the further predicted environmental resource data output by the third machine learning model can be more accurate than the predicted environmental resource data output by the second machine learning model).

[0186] At block 836, processing logic inputs the process recipe (e.g., data indicative of the process recipe, such as process recipe settings) into a second trained machine learning model, which can predict environmental resource usage data based on (e.g., corresponding to) the process recipe.

[0187] At block 838, processing logic receives predicted environmental resource usage data output from the second machine learning model. In some embodiments, the environmental resource usage data indicates environmental resource consumption associated with processing a substrate according to a process recipe. In some embodiments, the predicted environmental resource usage data is time series data indicating resource consumption over time. For example, the second machine learning model may predict power consumption of one or more components (e.g., heater, etc.) of a processing chamber when a substrate is processed according to a recipe input to the second machine learning model.

[0188] FIG. 9A illustrates a chart showing predicted environmental resource consumption data versus observed environmental resource consumption, according to some embodiments of the present disclosure. The chart illustrated in FIG. 9A may show predicted environmental resource consumption from a regression model (e.g., a machine learning model trained using one or more regression methods). In some embodiments, the predicted environmental resource consumption for a particular process recipe falls within a threshold range bounded by an upper limit 912 and a lower limit 914. Data points 908 (some are shown) may represent actual observed resource consumption versus predicted resource consumption (e.g., via one or more trained machine learning models described herein). When the predicted and actual resource consumption match, the data point will fall on dashed line 910, signifying that the predicted and actual resource consumption are equal. If the data point falls within the threshold range bounded by 914 and 912, one or more machine learning models described herein may have sufficient accuracy to predict environmental resource consumption. In some embodiments, a particular process recipe may have performance variations even when executed in the same processing chamber. Therefore, the predicted environmental resource consumption may correspond to the average environmental resource consumption.

[0189] FIG. 9B illustrates a chart showing predicted and actual time-series environmental resource consumption data 950, according to some embodiments of the present disclosure. A solid line 952 may represent actual resource consumption data, and a dashed line 954 may represent predicted resource consumption data. In some embodiments, the predicted time-series environmental resource consumption data corresponding to the dashed line 954 is output from a trained machine learning model. In some embodiments, a trained machine learning model (e.g., one or more trained machine learning models, multiple trained machine learning models connected in a "daisy chain," etc.) is trained using the actual resource consumption data as represented by the solid line 952. In some embodiments, physical constraints are used to inform the trained machine learning model. Additional machine learning models can predict additional resource consumption data based on the predicted time-series environmental resource consumption data represented by the dashed line 954. The consumption data may be predicted based on various inputs, such as substrate targets, process recipes, and / or historical training data. In some embodiments, eco-efficiency may be determined using the time-series environmental resource consumption data 950. For example, if energy consumption is represented by data 950, the total energy and / or power consumption over time for a processing recipe can be calculated (e.g., the area under the curve), and eco-efficiency data can be determined from the energy and / or power consumption.

[0190] 10 illustrates a block diagram of an exemplary computing device that operates in accordance with one or more aspects of the present disclosure. In various illustrative examples, various components of computing device 1000 may represent various components such as system controller 128, computing device 250, a device running web client 220, etc.

[0191] The exemplary computing device 1000 may be connected to other computing devices (e.g., using a cloud environment, cloud technology, and / or edge computing) in a LAN, an intranet, an extranet, and / or the Internet. The computing device 1000 may operate as a server in a client-server network environment. The computing device 1000 may be a personal computer (PC), a set-top box (STB), a server, a network router, switch, or bridge, or any device capable of executing (sequentially or otherwise) a set of instructions that specify operations to be performed by that device. Furthermore, while only a single exemplary computing device is illustrated, the term "computer" should also be understood to include any collection of computers that individually or jointly execute an instruction set (or multiple instruction sets) to perform any one or more of the methods described herein.

[0192] The exemplary computing device 1000 may include a processing unit 1002 (also referred to as a processor or CPU), a main memory 1004 (e.g., read only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM)), a static memory 1006 (e.g., flash memory, static random access memory (SRAM)), etc.), and a secondary memory (e.g., data storage device 1018), which may communicate with each other via a bus 1030.

[0193] The processing unit 1002 represents one or more general-purpose processing units, such as a microprocessor or central processing unit. More specifically, the processing unit 1002 may be a complex instruction set computing (CISC) microsystem processor, a reduced instruction set computing (RISC) microsystem processor, a very long instruction word (VLIW) microsystem processor, a processor that executes other instruction sets, or a processor that executes a combination of instruction sets. The processing unit 1002 may also be one or more special-purpose processing units, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), or a network processor. According to one or more aspects of the present disclosure, the processing unit 1002 may be configured to execute instructions that implement the methods 600-800B illustrated in FIGS. 6-8B.

[0194] The exemplary computing device 1000 may further include a network interface device 1008, which may be communicatively connected to a network 1020. The exemplary computing device 1000 may further include a video display 1010 (e.g., a liquid crystal display (LCD), a touch screen, or a cathode ray tube (CRT)), an alphanumeric input device 1012 (e.g., a keyboard), a cursor control device 1014 (e.g., a mouse), and an audio signal generating device 1016 (e.g., a speaker).

[0195] The data storage device 1018 may include a machine-readable storage medium (or, more specifically, a non-transitory machine-readable storage medium) 1028 having one or more sets of executable instructions 1022 stored thereon. For example, the data storage may be on-premise physical storage or may be remote storage, such as a cloud storage environment. According to one or more aspects of the present disclosure, the executable instructions 1022 may include executable instructions associated with performing method 800A of FIG. 8A and / or method 800B of FIG. 8B. In one embodiment, the instructions 1022 include instructions for the eco-efficiency module 129 of FIG. 1.

[0196] The executable instructions 1022 may also reside, completely or at least partially, within the main memory 1004 and / or within the processing unit 1002 during execution by the exemplary computing device 1000, with the main memory 1004 and the processing unit 1002 also constituting computer-readable storage media. The executable instructions 1022 may also be transmitted or received over a network via the network interface device 1008.

[0197] While Figure 10 depicts computer-readable storage medium 1028 as a single medium, the term "computer-readable storage medium" should be understood to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store one or more sets of processing instructions. The term "computer-readable storage medium" should also be taken to include any medium capable of storing or recording a set of instructions for execution by a machine, causing the machine to perform any one or more of the methods described herein. Accordingly, the term "computer-readable storage medium" should be taken to include, but is not limited to, solid-state memory, optical media, and magnetic media.

[0198] Some portions of the detailed description are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here, and generally, conceived to be a self-consistent sequence of operations leading to a desired result. The operations require physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.

[0199] It should be noted, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to such quantities. Unless otherwise specifically stated, throughout this specification, descriptions utilizing terms such as "providing," "determining," "storing," "adjusting," "causing," "receiving," "comparing," "creating," "stopping," "loading," "copying," "throwing," "replacing," or "performing," or "outputting," etc., should be understood to refer to the actions and processing of a computer system or similar electronic computing device that manipulates data represented as physical quantities (e.g., electrical quantities) in the computer system's registers and memory, and converts that data into other data that are similarly represented as physical quantities in the computer system's memory or registers, or other storage, transmission, or display device for such information.

[0200]

[0010] Embodiments of the present disclosure also relate to apparatus for performing the methods described herein. The apparatus may be specially configured for the required purposes, or may be a general-purpose computer system selectively programmed by a computer program stored on the computer system. Such computer program may be stored on a computer-readable storage medium, such as any type of disk, including optical disks, CD-ROMs, and magneto-optical disks, read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic disk storage media, optical storage media, flash memory devices, other types of machine-accessible storage media, or any type of medium suitable for storing electronic instructions, each connected to a computer system bus.

[0201] The methods and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may be used with programs in accordance with the teachings herein, or it may prove more convenient to construct a more specialized apparatus to perform the required method steps. The required structure for a variety of such systems will be apparent from the description that follows. Moreover, the scope of the present disclosure is not limited to any particular programming language. It will be understood that a variety of programming languages ​​can be used to implement the teachings herein.

[0202] It should be understood that the foregoing specification is intended to be illustrative, and not limiting. Many other embodiments will become apparent to those skilled in the art upon reading and understanding the foregoing specification. While particular examples have been described in this disclosure, it will be understood that the systems and methods of the present disclosure are not limited to the examples described herein, but can be practiced with modification within the scope of the appended claims. Correspondingly, the specification and drawings should be regarded in an illustrative, and not a limiting, sense. The scope of the present disclosure should, therefore, be defined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

1. 1. A method comprising: receiving a first process recipe including first process recipe settings data; inputting the first process recipe into one or more trained machine learning models that output predicted first environmental resource usage data indicative of first environmental resource consumption associated with processing a substrate in a processing chamber according to the first process recipe; outputting a recommendation associated with the first processing recipe based at least in part on the predicted first environmental resource usage data; A method comprising:

2. determining the recommendation based on a comparison of the predicted first environmental resource usage data and predicted second environmental resource usage data, the predicted second environmental resource usage data indicating a second environmental resource consumption associated with processing the substrate in the processing chamber according to a second process recipe; The method of claim 1 further comprising:

3. receiving target data indicative of a target substrate state for the processed substrate; inputting the target data into one or more additional models; receiving the first process recipe and the second process recipe as output from the one or more additional models; The method of claim 2 further comprising:

4. The method of claim 3 , wherein the one or more additional models include a second trained machine learning model.

5. predicting one or more first measurements corresponding to the first process recipe via a first additional model of the one or more additional models; predicting, via a second additional model of the one or more additional models, one or more second measurement values ​​based on the first process recipe and the one or more first measurement values ​​output from the first additional model; The method of claim 3 further comprising:

6. 6. The method of claim 5, wherein the one or more first measurements and the one or more second measurements comprise predicted measurements of at least one of current, voltage, power, flow rate, pressure, concentration, velocity, acceleration, or temperature.

7. 2. The method of claim 1, wherein the predicted first environmental resource usage data comprises predicted time series data associated with predicted behavior of the processing chamber while executing the first process recipe.

8. 10. The method of claim 1, wherein the recommendation includes modifying the first process recipe to form a modified first process recipe, wherein processing the substrate according to the modified first process recipe consumes less environmental resources than processing the substrate according to the first process recipe.

9. The method of claim 1 , wherein the environmental resource usage data comprises time series data for at least one of energy consumption, gas consumption, or water consumption associated with substrate processing in the processing chamber.

10. 1. A system comprising: one or more processing chambers configured to process a substrate, the one or more processing chambers including a plurality of sensors; a system controller for controlling the one or more processing chambers, receiving a first process recipe including first process recipe settings data; inputting the first process recipe into one or more trained machine learning models that output predicted first environmental resource usage data indicative of first environmental resource consumption associated with processing a substrate in a first processing chamber according to the first process recipe; outputting a recommendation associated with the first processing recipe based at least in part on the predicted first environmental resource usage data; a system controller that performs the above steps; A system equipped with

11. The system controller determining the recommendation based on a comparison of the predicted first environmental resource usage data and predicted second environmental resource usage data, the predicted second environmental resource usage data indicating a second environmental resource consumption associated with processing the substrate in the first processing chamber according to a second process recipe; The system of claim 10 further comprising:

12. The system controller receiving target data indicative of a target substrate state for the processed substrate; inputting the target data into one or more additional models; receiving the first process recipe and the second process recipe as output from the one or more additional models; The system of claim 11 further comprising:

13. The system of claim 12 , wherein the one or more additional models include a second trained machine learning model.

14. The system controller predicting one or more first measurements corresponding to the first process recipe via a first additional model of the one or more additional models; predicting, via a second additional model of the one or more additional models, one or more second measurement values ​​based on the first process recipe and the one or more first measurement values ​​output from the first additional model; The system of claim 12 further comprising:

15. 11. The system of claim 10, wherein the recommendation includes modifying the first process recipe to form a modified first process recipe, wherein processing the substrate according to the modified first process recipe consumes less environmental resources than processing the substrate according to the first process recipe.

16. A non-transitory machine-readable storage medium containing instructions that, when executed by a processing device, cause the processing device to: training a first machine learning model to form a first trained machine learning model, the first trained machine learning model being trained to output predicted measurement data based on a process recipe input to the first trained machine learning model; training a second machine learning model using training data including the predicted measurement data output from the first trained machine learning model to form a second trained machine learning model, the second trained machine learning model being trained to output predicted first environmental resource usage data indicative of environmental resource consumption associated with processing a substrate in a processing chamber according to the process recipe input to the second trained machine learning model; A non-transitory machine-readable storage medium that causes

17. The processing device includes: training a third machine learning model using training data including the predicted measurement data output from the first trained machine learning model and the predicted first environmental resource usage data output from the second machine learning model to form a third trained machine learning model, the third machine learning model being trained to output predicted second environmental resource usage data indicative of the environmental resource consumption associated with processing a substrate in a processing chamber according to the process recipe input to the trained third machine learning model.

17. The non-transitory machine-readable storage medium of claim 16, further comprising:

18. The processing device includes: training an additional machine learning model using training input data including historical process target data and training target output data including historical process recipes to form an additional trained machine learning model, wherein the additional trained machine learning model is trained to output one or more predicted process recipes associated with the process targets input to the additional trained machine learning model; 17. The non-transitory machine-readable storage medium of claim 16, further comprising:

19. The processing device includes: receiving measurement data associated with a plurality of process recipes, the measurement data including measurements of at least one of current, voltage, power, flow rate, pressure, concentration, velocity, acceleration, or temperature; receiving environmental resource usage data corresponding to the plurality of processing recipes, the environmental resource usage data indicating environmental resource consumption associated with the plurality of processing recipes; training one or more of the first machine learning model or the second machine learning model using one or more of the measurement data or the environmental resource usage data; 17. The non-transitory machine-readable storage medium of claim 16, further comprising:

20. The processing device includes: receiving a first process recipe including first process recipe settings data; inputting the first process recipe into the second trained machine learning model; outputting a recommendation associated with the first process recipe based at least in part on predicted first environmental resource usage data associated with the first process recipe; 17. The non-transitory machine-readable storage medium of claim 16, further comprising:

Images