Coded Substrate Material Identifier Communication Tool

Abstract

Embodiments disclosed herein include a method of processing a substrate in a tool. In one embodiment, the method of processing a substrate in a tool includes receiving an augmented recipe using a machine learning (ML) and / or artificial intelligence (AI) module. In one embodiment, the augmented recipe includes a recipe for processing the substrate in the tool and a matrix identifier corresponding to one or more substrate properties. In one embodiment, the method further includes obtaining a dataset from a database using the ML and / or AI module, the dataset being associated with the matrix identifier, and modifying the augmented recipe using the ML and / or AI module to form a modified recipe, the modification being dependent on the dataset.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Patent Application No. 17 / 524,677, filed November 11, 2021, the entire contents of which are incorporated herein by reference. [Technical Field]

[0002] Embodiments relate to the field of semiconductor manufacturing, and more particularly to a coded substrate material identifier system used to classify substrates using matrices instead of proprietary information. [Background technology]

[0003] In semiconductor processing, machine learning (ML) and artificial intelligence (AI) are increasingly being used to process substrates (e.g., semiconductor wafers) with improved uniformity and process control. To improve the performance of the ML and AI modules, contextual information from the substrate is provided to the ML and AI modules. For example, the contextual information may include substrate properties (e.g., thickness, reflectivity, resistivity, etc.) or process history. However, providing contextual information to the ML and / or AI modules is not always possible. For example, intellectual property restrictions may require that identifiable contextual information not be provided to the ML and / or AI modules. Correspondingly, limitations on the distribution of existing information may result in suboptimal use of the ML and / or AI modules. Summary of the Invention

[0004] Embodiments disclosed herein include a method for processing a substrate in a tool. In one embodiment, the method for processing a substrate in a tool includes receiving an augmented recipe using a machine learning (ML) and / or artificial intelligence (AI) module. In one embodiment, the augmented recipe includes a recipe for processing the substrate in the tool and a matrix identifier corresponding to one or more substrate characteristics. In one embodiment, the method further includes obtaining, using the ML and / or AI module, a dataset from a database, the dataset being associated with the matrix identifier, and modifying, using the ML and / or AI module, the augmented recipe to form a modified recipe, the modification dependent on the dataset.

[0005] An additional embodiment includes a method of constructing a database used by a machine learning (ML) and / or artificial intelligence (AI) module to process substrates without knowledge of underlying substrate characteristics. In one embodiment, the method includes associating a matrix identifier with a first substrate, the matrix identifier corresponding to one or more substrate characteristics of the first substrate, processing the first substrate in a tool, and storing sensor data from the tool during processing of the first substrate in the database, the sensor data being associated with the matrix identifier.

[0006] Embodiments may also include a semiconductor processing tool. In one embodiment, the tool comprises a host computer and an artificial intelligence (AI) module and / or machine learning (ML) module communicatively coupled to the host computer. In one embodiment, the tool further comprises a database communicatively coupled to the AI ​​and / or ML module and a processing chamber. In one embodiment, the processing chamber operates according to a recipe selected by the host computer and modified by the AI ​​and / or ML module in consideration of a dataset in the database. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram of a processing tool including a host computer in communication with a machine learning (ML) and / or artificial intelligence (AI) module for controlling process recipes executed in the processing tool, according to one embodiment. [Figure 2] FIG. 1 illustrates a cross-sectional view of a processing tool including a control loop sensor and a witness sensor according to an embodiment. [Figure 3] FIG. 1 is a block diagram of a semiconductor processing tool according to one embodiment. [Figure 4] FIG. 10 is a flow diagram of a process for building a database including row and column identifiers for categorizing process data according to one embodiment. [Figure 5] FIG. 1 is a flow diagram of a process for processing a substrate using an ML / AI module that utilizes a matrix identifier to select appropriate reference data, according to one embodiment. [Figure 6] FIG. 1 illustrates a block diagram of an exemplary computer system that may be used with a processing tool, according to one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] The systems described herein include methods and apparatus for measuring radicals within a coded substrate material identifier system used to classify substrates using matrices instead of proprietary information. In the following specification, numerous specific details are set forth to provide a thorough understanding of the embodiments. It will be apparent to those skilled in the art that the embodiments may be practiced without these specific details. In other instances, well-known aspects have not been described in detail so as not to unnecessarily obscure the embodiments. Furthermore, it should be understood that the various embodiments illustrated in the accompanying drawings are illustrative representations and have not necessarily been drawn to scale.

[0009] As described above, machine learning (ML) and / or artificial intelligence (AI) modules function with improved effectiveness when provided with contextual information about the substrates being processed within a semiconductor processing tool. However, such contextual information may not be shared with the ML and / or AI modules due to intellectual property restrictions. For example, a fabrication facility may be a first party, the substrates being processed may be owned by a second party, and the ML and / or AI module may be a third party application. Intellectual property restrictions may require that substrate parameter information be kept between the first and second parties.

[0010] In response, embodiments disclosed herein include constructing a database that includes substrate processing data associated with matrix identifiers instead of basic substrate characteristics. Such a database can be populated by using a matrix generator to assign matrix identifiers to substrate recipes before the recipes are sent to the ML / AI module. Substrate processing data from the processing of substrates can be stored in the database using the matrix identifiers to categorize the processing data.

[0011] After being entered into the database, the ML / AI module can use the data from the database to refine the processing parameters of the recipe to more closely control process uniformity and tolerance. For example, the ML / AI module can receive an augmented process recipe from a host computer. The augmented process recipe includes process recipe steps and matrix identifiers associated with the substrate being processed. The ML / AI module can then query the database for a data set corresponding to the matrix identifier. The ML / AI module can use this information to modify the process recipe steps to improve process uniformity and / or tolerance control. In this way, the ML / AI module can utilize historical data without needing to know the underlying substrate characteristics, which may contain confidential proprietary information.

[0012] 1 , a block diagram of a semiconductor processing tool 100 according to one embodiment is shown. In one embodiment, the processing tool 100 may include a host computer 140. In one embodiment, the host computer 140 may be any computer or server platform within a semiconductor facility. In some embodiments, the host computer 140 is dedicated to controlling a single tool 150. In other embodiments, the host computer 140 may provide control for multiple tools 150.

[0013] In one embodiment, the host computer 140 may include a matrix generator 142. The matrix generator 142 is responsible for associating individual substrates with a matrix identifier. The matrix identifier is a matrix that allows one or more basic substrate characteristics to be classified numerically. To make the appropriate assignment, the matrix generator 142 may have access to the basic substrate characteristics. In some embodiments, the matrix identifier includes a matrix with one row and multiple columns. Each column may refer to a different substrate characteristic. The value of each column may be any number. For example, the value of each column may be between 0 and 1000, although larger values ​​may be used in some embodiments. While ten columns are provided in certain embodiments, it should be understood that any number of columns may be used in various embodiments.

[0014] In one embodiment, the substrate characteristics represented by the matrix identifiers can be any substrate characteristics that, if known, would be useful for processing the substrate. For example, the substrate characteristics can include one or more of the following: material type (e.g., Si, SiO2, SiC, poly-Si, etc.), resistivity, substrate thickness, substrate surface reflectivity, chip layout, chip size, chip uniformity, the number of times the substrate has been processed in a recipe, and the position of the substrate within a lot. Although several substrate characteristics are listed, it should be understood that any number of substrate characteristics can be represented using the matrix identifiers.

[0015] In one embodiment, substrate properties may be categorized as numerical values ​​in a matrix. For example, the resistivity of a substrate may be divided into a series of ranges. Resistivity less than 0.03 Ω·cm may be assigned a value of 1, resistivity between 0.03 Ω·cm and 0.1 Ω·cm may be assigned a value of 2, resistivity between 0.1 Ω·cm and 10 Ω·cm may be assigned a value of 3, and resistivity greater than 10 Ω·cm may be assigned a value of 4. While resistivity is provided as an example, it should be understood that any substrate property may be assigned a number in a similar manner.

[0016] Additionally, substrate characteristics may refer to categorical characteristics instead of ranges of values. For example, a substrate layout may be classified as sparse or dense. If a substrate layer is sparse, a value of 1 may be used in the matrix identifier. Alternatively, if the substrate layout is dense, a value of 2 may be used in the matrix identifier. In this manner, substrate characteristics that are numeric values, or categorical identifiers, may be represented in the matrix identifier.

[0017] In one embodiment, the host computer 140 may also have access to a recipe 144 for processing a substrate. The host computer 140 selects an appropriate recipe 144 for a given substrate, and the host computer 140 uses a matrix generator 142 to generate a matrix identifier that can be added to the recipe 144. The generated matrix identifier corresponds to the substrate characteristics of the substrate being processed using the recipe 144. The resulting recipe may be referred to in some embodiments as an augmented recipe 145 because additional information has been added to the recipe 144 to assist in processing the substrate.

[0018] In one embodiment, the augmented recipe 145 is sent to a machine learning and / or artificial intelligence module 146 (ML / AI module 146 for short). The ML / AI module 146 may not have access to the underlying substrate properties of the substrate being processed. Instead, the ML / AI module 146 uses the matrix identifiers in the augmented recipe 145 to select data sets that are useful for improving the processing of the substrate. The data used by the ML / AI module 146 may be stored in a database 147. The data stored in the database 147 may be sorted using the values ​​of the matrix identifiers instead of the underlying substrate properties. In this way, the ML / AI module 146 can select data sets that correspond to the matrix identifiers provided in the augmented recipe 145.

[0019] In one embodiment, database 147 may be populated with sensor data from the processing of one or more substrates. The sensor data may, in some embodiments, include control loop sensor data. That is, data used in the control loop of the substrate processing may be used. Other data sources may also be used. For example, a witness sensor (e.g., physical or virtual) may be used to provide additional data for database 147. In some embodiments, metrology data obtained after processing a substrate may also be provided in database 147. The process of populating database 147 with data is provided in more detail below.

[0020] The ML / AI module 146 can utilize data from the database 147 to modify the augmentation recipe 145. For example, historical data from substrates with similar (or the same) matrix identifiers can be used to modify processing conditions to provide more uniform process results, tighter tolerance processes, or any other improved substrate results.

[0021] In one embodiment, the ML / AI module 146 may be communicatively coupled to a processing tool 150. The processing tool 150 may be any semiconductor processing tool. For example, the processing tool 150 may include a radical oxidation tool, a plasma tool (e.g., for etching, deposition, or surface treatment), or the like.

[0022] In the particular embodiment shown in FIG. 2 , a schematic diagram of a radical oxidation tool 250 according to one embodiment is shown. Tool 250 illustrates hardware components that may be utilized within one or more of the processing tools 150 described above with respect to FIG. 1 . In the illustrated embodiment, the described tool 250 is a lamp-based chamber for a radical oxidation process. However, it should be understood that processing tool 250 is exemplary in nature, and that embodiments disclosed herein may be suitable for use in combination with other processing tools, such as, but not limited to, heater-based or plasma-based chambers. Those skilled in the art will recognize that the placement of sensors, the number of sensors, and the type of sensors may be modified to track desired processing parameters for various types of processing tools.

[0023] In one embodiment, tool 250 includes chamber 205. Chamber 205 may be a chamber suitable for providing a sub-atmospheric pressure at which a substrate (e.g., a semiconductor wafer) is processed. In one embodiment, chamber 205 may be sized to accommodate one substrate or multiple substrates. Semiconductor substrates suitable for processing in chamber 205 may include silicon substrates or any other semiconductor substrates. Other substrates, such as glass substrates, may also be processed in chamber 205.

[0024] In one embodiment, a gas distribution network supplies gas from one or more gas sources (e.g., Gas 1, Gas 2, Gas n, etc.) to cartridge 210. In certain embodiments, the gas sources may include one or more of oxygen, hydrogen, and nitrogen. While three gas sources are shown in FIG. 2, it is understood that an embodiment may include one or more gas sources. Cartridge 210 may include an inlet for receiving gas from line 211 and an outlet for distributing gas into chamber 210. In the illustrated embodiment, cartridge 210 is shown as supplying gas into the chamber from one side of chamber 210. However, it is understood that cartridge 210 may optionally supply gas into the chamber from above or below the chamber. In some embodiments, cartridge 210 may also be referred to as a showerhead, particularly when the processing tool is a plasma-generating tool.

[0025] In one embodiment, the flow rate of each process gas may be controlled by a separate mass flow controller (MFC) 203. In one embodiment, the MFC 203 may be part of a control loop sensor group. The MFC 203 controls the flow rate of gas entering the input line 211. In one embodiment, a mass flow meter (MFM) 212 is provided upstream of the cartridge 210. The MFM 212 allows for measurement of the actual flow rate from the source gas. A pressure gauge 213 is also included upstream of the cartridge 210. The pressure gauge 213 allows for measurement of the pressure in the input line 211. The MFM 212 and the pressure gauge 213 may be considered witness sensors because they are outside the control loop.

[0026] In one embodiment, a chamber pressure gauge 217 may be provided to measure the pressure in the chamber 205. The chamber pressure gauge 217 may be part of a control loop sensor group. In one embodiment, additional witness sensors are provided along the exhaust line 215 of the processing tool 200. The additional sensors may include a leak detection sensor 216 and additional pressure gauges 218 and 219. The leak detection sensor 216 may include a self-contained optical emission spectroscopy (OES) device to measure oxygen leaking into the chamber 205. The pressure gauges 218 and 219 may be upstream and downstream of the throttle valve 214, respectively.

[0027] In one embodiment, witness sensors (e.g., 212, 213, 216, 218, 219) can be used to provide monitoring for chamber drift. For example, control loop sensors (e.g., 203 and 217) can become miscalibrated during use of processing tool 200. Thus, the readings of control loop sensors 203 and 217 can remain constant while the results on the wafer (e.g., film deposition rate) change. In such an instance, the output of the witness sensor changes, indicating that the chamber has drifted.

[0028] In additional embodiments, witness sensors may be utilized to implement virtual sensors within the chamber 205. A virtual sensor may refer to a sensor that provides a computationally generated output, as opposed to a direct reading of a physical value (as in a physical sensor). Virtual sensors are therefore powerful for determining conditions within the process tool 200 that are difficult or impossible to measure with traditional physical sensors.

[0029] In one embodiment, virtual sensors can be used to determine the flow rate of process gas at the outlet of cartridge 210. Calculating the flow rate at cartridge 210 is an important metric that can be used to control the deposition rate and / or deposition uniformity of a film on a wafer. In certain embodiments, the flow rate at cartridge 210 can be calculated using Bernoulli's equation with variables provided by using the outputs of MFM 212, pressure gauge 213, pressure gauge 217, and the known geometry of cartridge 210. While an example of a flow rate at a cartridge is provided, it should be understood that other unknowns within processing tool 200 can be determined using virtual sensor calculations. For example, unknowns such as, but not limited to, gas composition at various locations within the chamber, deposition rate across the wafer, pressure across the wafer, and film deposition across the wafer can be determined using virtual sensor implementations.

[0030] In one embodiment, one or more temperature sensors 207 are provided within the chamber 205. For example, the temperature sensors 207 may be thermocouples or the like. In one embodiment, the temperature sensors 207 may be provided on a reflector plate (not shown) of the chamber. The temperature sensors 207 may, in some embodiments, be considered witness sensors; that is, the temperature sensors 207 may be outside of the control loop.

[0031] The temperature sensor 207 can provide an additional known variable to enable the implementation of a wider range of virtual sensors. In one embodiment, the temperature sensor 207 can also be used to determine when a steady state is reached within the chamber 205. This is particularly useful when bringing the processing tool 200 up from a cold state, such as starting up the processing tool 250 after a maintenance event. For example, the output of the temperature sensor 207 in combination with one or more pressure gauges 213, 217, 218, and 219, as well as the angle of the throttle valve 214, can be monitored, and the chamber can be ready for use when the steady state of the various sensors is reached. In one embodiment, monitoring when the chamber reaches a steady state is useful because it eliminates wafer scrap and rework that typically occurs within a processing tool due to the first wafer effect.

[0032] As can be appreciated, multiple sensors (e.g., control loop sensors, witness sensors, virtual sensors, etc.) can be used to provide a very detailed picture of process conditions for the processing of various substrates. When substrate characteristics are associated with process conditions using matrix identifiers, a highly detailed database can be created. The database can then be used by an ML / AI module to improve the processing of substrates without having to provide the underlying substrate characteristics to the ML / AI module. In this way, the benefits of the ML / AI module can be obtained without sharing substrate information that may be subject to intellectual property controls that limit the sharing of underlying substrate data.

[0033] 3, a schematic diagram of a processing tool 300 according to one embodiment is shown. As shown, an ML / AI module 320 may be integrated with the processing tool 300. For example, the ML / AI module 320 may be communicatively coupled to a front-end server 360 via a network connection, as indicated by the arrow. However, in other embodiments, the ML / AI module 320 may reside external to the processing tool 300. For example, the ML / AI module 320 may be communicatively coupled to the processing tool 300 via an external network, etc.

[0034] In one embodiment, the ML / AI module 320 may include a hybrid model. The hybrid model may include a physical model 327 and a statistical model 325. The statistical model 325 and the physical model 327 may be communicatively coupled to a database 330 for storing input data (e.g., sensor data, model data, metrology data, etc.) used to build and / or update the statistical model 325 and the physical model 327. In one embodiment, the statistical model 325 may be generated by performing a physical Design of Experiments (DoE) and may use interpolation to provide an extended process space model. In one embodiment, the physical model 327 may be generated using real-world physics and chemistry relationships. For example, the physical model may be constructed using physical and chemical equations for various interactions within a process chamber.

[0035] In certain embodiments, the physical model 327 and the statistical model 325 may be informed by one or more substrate properties. For example, various substrate properties may cause differences in the physical model 327 and / or the statistical model 325. In some embodiments, certain substrate properties (e.g., resistivity, thickness, etc.) are unknown to the physical model 327 and the statistical model 325. Instead, the substrate properties are converted into values ​​in a matrix identifier as described above.

[0036] In one embodiment, the processing tool 300 may include a front-end server 360, a tool control server 350, and tool hardware 340. The front-end server 360 may include a dashboard 365 for the ML / AI module 320. The dashboard 365 provides an interface for process engineers to utilize data modeling to perform various operations, such as augmenting process recipes.

[0037] The tool control server 350 may include a smart monitoring and control block 355. The smart monitoring and control block 355 may include modules for providing diagnostics and other monitoring of the processing tool 300. The modules may include, but are not limited to, health checks, sensor drift, fault recovery, and leak detection. The smart monitoring and control block 355 may receive as input data from various sensors implemented within the tool hardware. The sensors may include standard sensors 347 typically present within a semiconductor manufacturing tool 300 to enable operation of the tool 300. For example, the sensors 347 may include control loop sensors, as described above. The sensors may also include witness sensors 345 added to the tool 300. The witness sensors 345 provide additional information necessary to build highly detailed data models. For example, the witness sensors may include physical sensors and / or virtual sensors. As described above, virtual sensors can utilize data obtained from two or more physical sensors and use computations to provide additional sensor data not available from physical sensors alone. In general, the witness sensors may include any type of sensor, including, but not limited to, pressure sensors, temperature sensors, gas concentration sensors, etc. In one embodiment, the smart monitoring and control block 355 may provide data for use by the ML / AI module 320. In other embodiments, output data from the various witness sensors 345 may be provided directly to the ML / AI module 320.

[0038] 4, a flow diagram of a process 460 for building a database for use by an ML / AI module is shown, according to one embodiment. In one embodiment, process 460 may begin at step 461 with associating a matrix identifier with a first substrate. In one embodiment, the matrix identifier corresponds to one or more substrate characteristics of the first substrate.

[0039] In one embodiment, the matrix identifier may be generated by a matrix generator within a host computer, as described in more detail above, i.e., the host computer may have access to the basic substrate characteristics of the first substrate and may be capable of generating a matrix identifier associated with the basic substrate characteristics of the first substrate.

[0040] In some embodiments, the matrix identifier includes a matrix with one row and multiple columns. Each column may refer to a different substrate characteristic. The value of each column may be any number. For example, the value of each column may be between 0 and 1000, although larger values ​​may be used in some embodiments. In certain embodiments, 10 columns are provided, although it should be understood that any number of columns may be used in various embodiments.

[0041] In one embodiment, the substrate characteristics represented by the matrix identifiers can be any substrate characteristics that, if known, would be useful for processing the substrate. For example, the substrate characteristics can include one or more of the following: material type (e.g., Si, SiO2, SiC, poly-Si, etc.), resistivity, substrate thickness, substrate surface reflectivity, chip layout, chip size, chip uniformity, the number of times the substrate has been processed in a recipe, and the position of the substrate within a lot. Although several substrate characteristics are listed, it should be understood that any number of substrate characteristics can be represented using the matrix identifiers.

[0042] In one embodiment, substrate properties may be categorized as numerical values ​​in a matrix. For example, the resistivity of a substrate may be divided into a series of ranges. Resistivity less than 0.03 Ω·cm may be assigned a value of 1, resistivity between 0.03 Ω·cm and 0.1 Ω·cm may be assigned a value of 2, resistivity between 0.1 Ω·cm and 10 Ω·cm may be assigned a value of 3, and resistivity greater than 10 Ω·cm may be assigned a value of 4. While resistivity is provided as an example, it should be understood that any substrate property may be assigned a number in a similar manner.

[0043] Additionally, substrate characteristics may refer to categorical characteristics instead of ranges of values. For example, a substrate layout may be classified as sparse or dense. If a substrate layer is sparse, a value of 1 may be used in the matrix identifier. Alternatively, if the substrate layout is dense, a value of 2 may be used in the matrix identifier. In this manner, substrate characteristics that are numeric values, or categorical identifiers, may be represented in the matrix identifier.

[0044] In one embodiment, process 460 may proceed to step 462, which includes processing a first substrate in a tool. In one embodiment, the tool may include any tool suitable for processing semiconductor substrates, etc. For example, the tool may be a radical oxidation tool, a plasma tool, etc. In one embodiment, the tool may include a control loop sensor and a witness sensor (e.g., a physical witness sensor or a virtual witness sensor).

[0045] In one embodiment, process 460 may proceed to step 463, which includes storing sensor data from the tool processing the first substrate in a database. In one embodiment, the sensor data is associated with a row and column identifier assigned to the first substrate. In this manner, a unique identifier can be applied to the data for the first substrate without having to share underlying substrate characteristics with the ML / AI module and / or database.

[0046] In one embodiment, steps 461-463 can be repeated with additional substrates. Each additional substrate can also be associated with a matrix identifier. In this manner, many entries in the database can be created. After a database containing sufficient information is built, the ML / AI module can reference the stored data to modify the recipe to improve processing results.

[0047] Although process 460 uses the processing of multiple substrates to build the database, some embodiments may include inputting data from previously processed substrates to populate the database. In yet another embodiment, metrology data obtained after substrate processing may be added to the database as well.

[0048] 5, a flow diagram of a process 570 for processing a substrate using a matrix identifier with an ML / AI module is shown, according to one embodiment. In one embodiment, process 570 begins with step 571, which includes receiving an augmented recipe with the ML / AI module. In one embodiment, the augmented recipe includes a recipe for processing a first substrate and a matrix identifier associated with the first substrate. In one embodiment, the matrix identifier is added to the recipe by a host computer that has access to the underlying substrate properties of the first substrate. The matrix identifier includes one or more entries in a matrix that correspond to the underlying substrate properties. However, the underlying substrate properties themselves are not shared with the ML / AI module.

[0049] In one embodiment, process 570 may proceed to step 572, which may include using an ML / AI module to retrieve a dataset from a database, the dataset being associated with a matrix identifier. In one embodiment, the dataset may be associated with one or more entries of the matrix identifier associated with the first substrate. The dataset may include historical sensor data (e.g., control loop sensor data, witness sensor data, etc.). In some embodiments, the dataset may also include metrology data.

[0050] In one embodiment, process 570 may proceed to step 573, which includes modifying the augmented recipe using an ML / AI module to form a modified recipe. In one embodiment, modifications to the recipe are made in consideration of data from the acquired dataset. For example, the acquired dataset may be used to change the flow rates of various gases, change the temperature, change the pressure, and / or change the times of various processes, to name a few of the many modifications to the recipe.

[0051] In one embodiment, process 570 may proceed to step 574, which includes processing the first substrate associated with the row and column identifier in a processing tool using the modified recipe. In one embodiment, data from the processing tool's sensors may be provided back to the ML / AI module for storage in a database to provide additional data for future processing.

[0052] Referring now to FIG. 6 , a block diagram of an exemplary computer system 600 of a processing tool is shown according to one embodiment. In one embodiment, computer system 600 is connected to the processing tool and controls processing within the processing tool. Computer system 600 may be connected to (e.g., networked with) other machines in a local area network (LAN), an intranet, an extranet, or the Internet. Computer system 600 may operate as a server or client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. Computer system 600 may be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a server, a network router, switch, or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that define operations to be performed by the machine. Furthermore, although only a single machine is shown as computer system 600, the term "machine" is also intended to include any collection of machines (e.g., computers) that individually or jointly execute a set (or sets) of instructions to perform any one or more of the methods described herein.

[0053] The computer system 600 may include a computer program product, or software 622, having a non-transitory machine-readable medium having instructions stored thereon, which may be used to program the computer system 600 (or other electronic device) to perform processes according to embodiments. A machine-readable medium includes any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computer). For example, a machine-readable (e.g., computer-readable) medium may include a machine-readable storage medium (e.g., read-only memory (ROM), random access memory (RAM), magnetic disk storage media, optical storage media, flash memory devices, etc.), a machine-readable (e.g., computer-readable transmission medium (e.g., electrical, optical, acoustic, or other form of propagated signal (e.g., infrared signal, digital signal, etc.)), etc.

[0054] In one embodiment, computer system 600 includes a system processor 602, a main memory 604 (e.g., read only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM)), a static memory 606 (e.g., flash memory, static random access memory (SRAM), etc.), and a secondary memory 618 (e.g., data storage device), which communicate with each other via a bus 630.

[0055] The system processor 602 may refer to one or more general-purpose processing devices, such as a microsystem processor, a central processing unit, or the like. More specifically, the system processor 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 system processor that executes other instruction sets, or a system processor that executes a combination of instruction sets. The system processor 602 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 system processor (DSP), a network system processor, or the like. The system processor 602 is configured to execute processing logic 626 for performing the operations described herein.

[0056] The computer system 600 may further include a system network interface device 608 for communicating with other devices or machines. The computer system 600 may also include a video display unit 610 (e.g., a liquid crystal display (LCD), a light emitting diode display (LED), or a cathode ray tube (CRT)), an alphanumeric input device 612 (e.g., a keyboard), a cursor control device 614 (such as a mouse), and a signal generating device 616 (e.g., a speaker).

[0057] The secondary memory 618 may include a machine-accessible storage medium 632 (or, more specifically, a computer-readable storage medium) having stored thereon one or more sets of instructions (e.g., software 622) that embody any one or more of the methods or functions described herein. This software 622 may also reside, completely or at least partially, within the main memory 604 and / or the system processor 602 while being executed by the computer system 600, with the main memory 604 and the system processor 602 also constituting machine-readable storage media. The software 622 may further be transmitted or received over the network 620 via the system network interface device 608. In one embodiment, the network interface device 608 may operate using RF, optical, acoustic, or inductive coupling.

[0058] While in an exemplary embodiment, machine-accessible storage medium 632 is shown as a single medium, the term "machine-readable storage medium" should be interpreted to include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) on which one or more sets of instructions are stored. The term "machine-readable storage medium" should also be interpreted to include any medium capable of storing or encoding a set of instructions for execution by a machine and causing the machine to perform any one or more of the methods. Correspondingly, the term "machine-readable storage medium" should be interpreted to include, but is not limited to, solid-state memory, optical media, and magnetic media.

[0059] In the foregoing specification, certain exemplary embodiments have been described. It will be apparent that various modifications may be made to the exemplary embodiments without departing from the scope of the following claims. Correspondingly, the specification and drawings should be regarded in an illustrative rather than a restrictive sense.

Claims

1. 1. A method of processing a substrate in a tool, comprising: receiving an augmented recipe by a machine learning (ML) and / or artificial intelligence (AI) module, the augmented recipe comprising: a recipe for processing the substrate in the tool; and Row and column identifiers corresponding to one or more substrate characteristics Receive a reinforcement recipe, including retrieving, by the ML and / or AI module, a dataset from a database, the dataset being associated with the matrix identifier; modifying, by the ML and / or AI module, the augmented recipe to form a modified recipe, wherein the modification is dependent on the dataset; A method comprising:

2. The method of claim 1 , wherein the matrix identifier comprises a matrix including one row and multiple columns, each column representing a different context parameter.

3. The method of claim 2 , wherein the plurality of columns comprises at least 10 columns.

4. The method of claim 2 , wherein an indicator is entered into a plurality of columns of the plurality of columns.

5. 5. The method of claim 4, wherein the indicator is a number between 1 and 1000.

6. 3. The method of claim 2, wherein the matrix identifier provides context parameters for one or more of material type, resistivity, substrate thickness, substrate surface reflectivity, chip layout, chip size, chip uniformity, the number of times the substrate has been processed with the recipe, and the position of the substrate within a lot.

7. The method of claim 1 , wherein the matrix identifiers are generated by a host computer and the underlying substrate data used to generate the matrix identifiers is not accessible to the ML and / or AI modules.

8. The method of claim 1 , wherein the data set has no representation of the one or more substrate properties other than the row and column identifiers.

9. The method of claim 1 , wherein the data set comprises data from sensors in the tool while processing one or more substrates having the same row and column identifiers.

10. The method of claim 9 , wherein the sensors include a closed-loop sensor and a witness sensor.

11. The method of claim 1 , wherein the data set includes data from measurements of one or more substrates having the same row and column identifiers.

12. 1. A method for computationally constructing a database for use by a machine learning (ML) and / or artificial intelligence (AI) module to process substrates without knowledge of underlying substrate properties, comprising: a first step of associating a matrix identifier with a first substrate, the matrix identifier corresponding to one or more substrate characteristics of the first substrate; a second step of processing the first substrate in a tool; a third step of storing sensor data from the tool during processing of the first substrate in the database, the sensor data being associated with the row and column identifiers; A method comprising:

13. 13. The method of claim 12, further comprising repeating the first, second and third steps for a plurality of substrates.

14. The method of claim 12 , wherein the sensor data includes control loop sensor data and witness sensor data.

15. The method of claim 12 , wherein the sensor data further comprises metrology data for the processed first substrate.

16. The method of claim 12 , wherein the matrix identifier comprises a matrix including one row and multiple columns, each column representing a different context parameter.

17. The method of claim 16 , wherein the plurality of rows comprises at least 10 rows.

18. The method of claim 16 , wherein an indicator is entered into multiple columns of the multiple columns.

19. The method of claim 12, wherein the ML and / or AI module modifies a recipe for processing a substrate using stored sensor data.

20. 1. A semiconductor processing tool comprising: A host computer; an artificial intelligence (AI) and / or machine learning (ML) module communicatively connected to the host computer; a database communicatively connected to the AI ​​and / or ML module; a processing chamber operating according to a recipe selected by the host computer and modified by the AI ​​and / or ML module taking into account data sets in the database; Equipped with The semiconductor processing tool, wherein the host computer assigns a matrix identifier to the recipe, and the dataset used by the AI ​​and / or ML module is associated with the matrix identifier.

Images