Method for managing airborne pollutants

Abstract

A method for managing airborne contaminants includes: generating a contaminant distribution map by sampling the environment of a cleanroom; selecting a first fabrication tool for the cleanroom by comparing the contaminant distribution map with at least one diffusion image in a first database; comparing parameters of the first fabrication tool with process utility information in a second database; and taking at least one measure when the parameters are consistent with the process utility information. The at least one measure may include: moving a cleaning tool to a location associated with the contaminant concentration of the contaminant distribution map; turning on a fan of the cleaning tool; stopping pod conveyance to the first fabrication tool; and terminating generation through the first fabrication tool.

Description

Technical Field

[0001] The techniques described in embodiments of the present invention generally relate to contaminant management and, more particularly, to methods for airborne contaminant management. Background Art

[0002] The semiconductor integrated circuit (IC) industry has experienced exponential growth. Technological advances in IC materials and design have given rise to several generations of ICs, each having smaller and more complex circuits than the previous generation. In the course of IC evolution, while the geometric size (i.e., the smallest components (or lines) that can be formed using a fabrication process) has decreased, the functional density (i.e., the number of interconnected devices per chip area) has generally increased. Such scaling down of processes generally provides benefits by increasing production efficiency and reducing associated costs. Such scaling down has also increased the complexity of processing and manufacturing ICs.

[0003] The cleanrooms in which ICs are manufactured face increasingly stringent tolerances for particulates and contaminants (such as chemicals that may leak from fabrication tools or attachments affixed to fabrication tools). When fabricating smaller and more complex circuits, cleanrooms with relatively low particulate and / or contaminant levels may be associated with higher yields. Summary of the Invention

[0004] Embodiments of the present invention provide a method for airborne contaminant management, including: generating a contaminant distribution map by sampling the environment of a cleanroom; selecting a first fabrication tool of the cleanroom by comparing the contaminant distribution map with at least one diffusion image in a first database; comparing parameters of the first fabrication tool with process utility information in a second database; and when the parameters are consistent with the process utility information, taking at least one of the following measures: moving a cleaning tool to a position associated with the contaminant concentration of the contaminant distribution map; turning on a fan of the cleaning tool; stopping pod conveyance to the first fabrication tool; and terminating generation through the first fabrication tool.

[0005] Embodiments of the present invention provide a method for airborne contaminant management, including: generating cleanroom contaminant data by sampling cleanroom contaminants using a sampling system; generating a first image based on the cleanroom contaminant data by a time-of-flight mass spectrometer; selecting a first fabrication tool based on a prediction using the first image and at least one other cleanroom diffusion image; and reducing the contaminant concentration near the first fabrication tool by an automated guided vehicle dispatched by an automated guided vehicle controller.

[0006] An embodiment of the present invention provides a method for managing airborne contaminants, including: positioning a wafer in a fabrication tool; detecting a peak concentration level of the contaminants above a first threshold; predicting that the fabrication tool is the source of the contaminants; stopping the delivery of additional wafers to the fabrication tool; completing the processing of the wafer by the fabrication tool; removing the wafer from the fabrication tool; reducing the peak concentration level by repairing the fabrication tool; and resuming the delivery of the additional wafers to the fabrication tool when the fabrication tool is repaired and the peak concentration level is below a baseline threshold, the baseline threshold being lower than the first threshold. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Aspects of the present disclosure are best understood when read in conjunction with the following detailed description. It should be noted that, in accordance with standard practice in the industry, various features are not drawn to scale. In fact, the dimensions of the various features may be arbitrarily increased or reduced for clarity of discussion.

[0008] Figures 1A to 1D is a view of a contaminant management system in accordance with an embodiment of the present disclosure.

[0009] Figures 2A to 2H is a view of a contaminant management process in accordance with various aspects of the present disclosure.

[0010] Figures 3A to 3B is a flowchart showing a method for processing a wafer in accordance with various aspects of the present disclosure.

[0011] Figures 4A to 4B is a view showing a process for predicting a contaminant source in accordance with various aspects of the present disclosure. DETAILED DESCRIPTION

[0012] The following disclosure provides many different embodiments or examples for implementing different features of the provided subject matter. Specific examples of components and arrangements are set forth below to simplify the present disclosure. Of course, these are merely examples and are not intended to be limiting. For example, in the following description, forming a first feature on or above a second feature may include embodiments in which the first feature and the second feature are formed in direct contact, and may also include embodiments in which additional features may be formed between the first feature and the second feature such that the first feature and the second feature are not in direct contact. Additionally, the present disclosure may reuse reference numerals and / or letters in various examples. Such reuse is for the purpose of simplicity and clarity and does not in itself indicate a relationship between the various embodiments and / or configurations discussed.

[0013] In addition, for ease of description, spatially relative terms such as "beneath", "below", "lower", "above", "upper", etc. may be used herein to describe the relationship of one element or feature shown in the figures to another (other) element or feature. The spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. The device may have other orientations (rotated 90 degrees or at other orientations), and the spatially relative descriptive terms used herein may be interpreted accordingly as well.

[0014] For ease of description, terms such as "about", "substantially", "essentially", etc. may be used herein. Those of ordinary skill in the art will be able to understand and deduce the meanings of these terms. For example, "about" may indicate a variation in size of 20%, 10%, 5%, and similar values, but other values may also be used as appropriate. Features as large as the longest dimension of a semiconductor fin may have a variation of less than 5%, while very small features such as the thickness of an interface layer may have a variation of up to 50%, and both types of variations may be represented by the term "about". "Substantially" is generally more stringent than "about", such that a variation of 10%, 5%, or less may be appropriate, but is not limited thereto. A feature that is "substantially flat" may have a deviation within 10% or less than 10% relative to a straight line. A material having a "substantially constant concentration" may have a concentration variation within 5% or less than 5% along one or more dimensions. Similarly, those of ordinary skill in the art will be able to understand and deduce the appropriate meanings of these terms based on industry knowledge, current fabrication techniques, etc.

[0015] Semiconductor fabrication generally involves forming an electronic circuit by performing multiple depositions, etchings, annealings, and / or implantations of material layers, thereby forming a stacked structure including many semiconductor devices and the interconnects between the semiconductor devices. Scaling (downscaling) is a technique for fitting a larger number of semiconductor devices within the same area. However, at advanced technology nodes, scaling is becoming increasingly difficult. As the dimensions are scaled down, the tolerance for cleanroom contaminants becomes increasingly stringent to ensure sufficient yield.

[0016] Airborne molecular contamination (AMC) is over-represented in out-of-control (OOC) and out-of-specification (OOS) events. A variety of tools and methods can be used to detect an increase in contaminant levels in a cleanroom, but generally are insufficient in many respects. Gas chromatography mass spectrometer (GC-MS) provides online global monitoring of a large spectrum of contaminants, but generally is very slow and may only be able to accommodate a single sampling point at each processing section. Patrol operators may also carry or push offline contaminant sampling devices, but these devices may only detect a small fraction of contaminants (such as acids, bases, and / or hydrochlorofluorocarbons (HCFCs)) and are unable to detect total volatile organic compounds (TVOCs). As such, detection and response to an increase in AMC generally are completed with significant delay, which can increase the likelihood that an OOC event becomes an OOS event, which may result in halting production while identifying, diagnosing, and repairing the tool or tool attachment that caused the increase in AMC.

[0017] In an embodiment of the present disclosure, a time-of-flight mass spectrometer (TOF-MS) is employed. The TOF-MS measures various AMC parameters by combining refrigerant by-products (such as chlorofluorocarbons, hydrofluorocarbons, perfluorocarbons, or the like), isopropyl alcohol (IPA), acetone, and general TVOC detection on a qualitative and / or quantitative basis with a precision of several hundred parts per billion (ppb) within a few minutes, and can utilize rapid identification of chemicals to reflect sudden leakage events.

[0018] The pollutant distribution and the peak concentration level position can be calculated by computational fluid dynamics (CFD) technology according to the tool layout, and the results are recorded in a database. When an AMC leak occurs in the cleanroom, the sampler array collects the pollutants and the TOF-MS analyzes the pollutants to generate an AMC concentration distribution map. The comparison between the AMC concentration distribution map and the calculation results can be used to predict the location (or tool) as the source of the leak. The predicted location (or tool) can be further compared with the tool information to ensure an accurate prediction. The tool information can include signals indicating the supply send of acids and / or solvents from the tool, and the signals can be stored in the plant-side supervisory control and data acquisition (SCADA) through a system integration (SI) network. The tool information can also include the tool operating status from a fault detection system and / or the tool utility information from an "electronic bluebook" (e-Bluebook). After comprehensive comparison and judgment (e.g., through artificial intelligence (AI) / machine learning (ML)), the predicted AMC source can be identified and confirmed. If the predicted result is inconsistent with the advanced comparison including the tool information just mentioned, the next most likely location / tool can be used as the source of the leak, and the same advanced comparison can be repeated until a match is found.

[0019] In this way, quasi-real-time detection and response can be achieved. The described embodiments can be deployed in sensitive process sections (such as etching, electrochemical plating (ECP) / seed deposition, bench, or similar sections) for local monitoring. In some embodiments, the AMC real-time information can be integrated into a centralized system (such as a supervisory control and data acquisition (SCADA) system) to automatically generate contour maps and animations and issue alarms in the AMC hot sections.

[0020] Figure 1AFIG. 0 is a schematic diagram of a system 100 with AMC management in accordance with various embodiments of the present disclosure. The system 100 may be configured to monitor contaminant levels in a cleanroom, dispatch an automated cleaning unit to the source of a contaminant leak, activate the automated cleaning unit to reduce the contaminant level, terminate the transfer of a front opening unified / universal pod (FOUP) to a tool if the tool is the source, and stop generation through the tool.

[0021] The system 100 includes a foundry 120, a sampling system 130, an analysis system 140, and a control center 150, which interact with each other in manufacturing and / or services related to manufacturing IC devices. Entities in the system 100 are connected via a communication network. In some embodiments, the communication network is a single network. In some embodiments, the communication network is a variety of different networks (e.g., an intranet and the Internet). The communication network includes wired communication channels and / or wireless communication channels. In some embodiments, the communication network includes short-range asset tracking hardware and software, such as radio-frequency identification (RFID), Bluetooth Low Energy (BLE), wireless-fidelity (WiFi), ultra-wideband (UWB), or the like. In some embodiments, the communication network includes wide-range asset tracking hardware and software, such as a low-power wide-area network (LPWAN), Long-Term Evolution (LTE), 5th generation mobile network (5G), Global Positioning System (GPS), or the like. Each entity interacts with one or more of the other entities and provides services to and / or receives services from one or more of the other entities. In some embodiments, one or more of the foundry 120, the sampling system 130, the analysis system 140, and the control center 150 are owned by a single larger company. In some embodiments, one or more of the foundry 120, the sampling system 130, the analysis system 140, and the control center 150 coexist in a common facility and use common resources.

[0022] The foundry 120 includes a wafer fabrication tool 122 (hereinafter referred to as "fabrication tool 122"), which is configured to perform various manufacturing operations on a semiconductor wafer to fabricate IC devices. In various embodiments, the fabrication tool 122 includes one or more of a wafer stepper, an ion implanter, a photoresist coater, a process chamber (such as a chemical vapor deposition (CVD) chamber or a low-pressure CVD (LPCVD) furnace, a chemical-mechanical planarization (CMP) system, a plasma etching system, a wafer cleaning system, or other manufacturing equipment capable of performing one or more suitable manufacturing processes discussed herein).

[0023] The fabrication tool 122 is located in a cleanroom 121. The cleanroom 121 provides an environment designed to maintain extremely low levels of particulates (such as dust, airborne organisms, or vaporized particles), which can be quantified as the number of particles per cubic meter. As the IC feature size shrinks, the acceptable number and size of particulates decrease accordingly. In addition, the cleanroom 121 may have restrictions on AMC (such as HCFCs, acids, bases, and other chemicals). Therefore, the cleanroom 121 may include many design features for reducing particulate and contaminant levels, such as filtration units, dedicated lighting, temperature and humidity control, air locks, pressure control, and the like.

[0024] In addition to the above, the cleanroom 121 also includes a cleaning tool 123. The cleaning tool 123 is configured to move to a location in the cleanroom 121, locally filter the air through an on-board pump / fan and filter to reduce the contaminant level, and return to an electrical charging station without manual intervention. In some embodiments, the cleaning tool 123 is an automated guided vehicle (AGV). Refer to Figure 1D The cleaning tool 123 will be described in more detail. Refer to Figure 1B The cleanroom 121 will be described in more detail.

[0025] A sampling system 130 collects samples of air from the cleanroom 121 at a large number of locations distributed throughout the cleanroom 121. The sampling system 130 may include a multi-channel sampling system 132 that collects the samples and is in fluid communication with a TOF-MS 142 of an analysis system 140. In some embodiments, the multi-channel sampling system 132 includes a multi-channel sampler, a manifold pipe sampler, a rotary valve sampler, or the like. Refer to Figure 1CDescribe the multi-channel sampling system 132 in more detail.

[0026] The analysis system 140 receives a sample of air from the sampling system 130 at the TOF-MS 142. The TOF-MS 142 can perform qualitative analysis and / or quantitative analysis on the chemicals in the sample. In some embodiments, the TOF-MS 142 measures the mass and concentration level of molecular ions in the air of the clean room 121. For example, the TOF-MS 142 can detect the concentration levels of refrigerant by-products, IPA, acetone, and / or general TVOCs in the sample with an accuracy better than about 0.05 ppb (e.g., about 0.02 ppb). The processing of the detections performed by the TOF-MS 142 can be faster than about 2 minutes / sample (e.g., about 1 minute / sample), but also includes other slower or faster detection processing times, e.g., due to the relative simplicity (faster) or complexity (slower) of detecting various chemical concentrations.

[0027] The control center 150 receives data from the TOF-MS 142 and stores the data in the database 152. The AMC supervisory control and data acquisition (SCADA) 151 can receive measurements (e.g., concentration levels) from the TOF-MS 142 and write the measurements to the database 152. In some embodiments, the AMC SCADA 151 receives the measurements via a network to which the TOF-MS 142 and the AMC SCADA 151 are each connected. In some embodiments, the AMC SCADA 151 is a control system including at least one computer. The AMC SCADA 151 can send information about the acid and / or solvent supply corresponding to each of the storage and fabrication tools 122, for example, via a system integration network. In some embodiments, the AMC SCADA 151 can further control the operation of the TOF-MS 142, which includes activation / deactivation (power on / off) of the TOF-MS 142, the order and / or type of measurements performed by the TOF-MS 142, and other suitable operating parameters of the TOF-MS 142. In some embodiments, the AMC SCADA 151 performs data processing on the measurements received from the TOF-MS 142, which can include writing the measurements to the database 152 as described above, and can also include other types of data processing (e.g., compression / decompression, noise reduction, smoothing, filtering, peak finding, and similar processing). In some embodiments, the AMC SCADA 151 performs visual graphics generation and display for the clean room 121 or one or more sections of the clean room 121.

[0028] The database 152 can be local, cloud-based, or any combination thereof. In some embodiments, the integration system 153 (which can be part of the AMC SCADA 151) can read the measurements recorded in the database 152 and can generate a contour map (e.g., a global map or a sectional map) corresponding to the layout of the clean room 121 based on the measurements. The AGV controller 154 of the control center 150 can remotely control the cleaning tool 123 (e.g., via the network device 155) to, for example, reposition to a contaminant hot spot, activate the on-board fan, deactivate the on-board fan, and return to the charging station. In some embodiments, the network device 155 includes at least one of a wired communication channel and / or a wireless communication channel, such as Wi-Fi, ultra-wideband (UWB), low-power wide area network (LPWAN), long-term evolution (LTE), fifth-generation mobile network (5G), or the like. Referring to Figures 2A to 2D Elaborate in more detail the details related to the control of the cleaning tool 123.

[0029] Figure 1B is a schematic diagram showing a partial floor plan of the clean room 121 according to various embodiments. The fabrication tools 122 include various fabrication tools 122A to 122J respectively arranged in a plurality of sections 121A to 121J. In Figure 1B Ten sections 121A to 121J are shown, however fewer or more sections can also be included in the clean room 121. The fabrication tools 122 can include any of deposition (e.g., CVD, physical vapor deposition (PVD), atomic layer deposition (ALD)), plating (e.g., electroless copper), lithography, etching, cleaning, and planarization (e.g., CMP, polishing) tools, or other suitable tools for manufacturing IC devices. Each of the sections 121A to 121J can include one or more types of tools (e.g., only an etching tool, or a plating tool and a planarization tool). The aisle 1210 (or walkway) can extend between the sections 121A to 121C and the sections 121D to 121F and can end adjacent to the section 121H. Generally, the aisle 1210 does not have fabrication tools 122. Some of the sections (e.g., sections 121G, 121I, 121J) may not be adjacent to the aisle 1210. The cleaning tool 123 (which can also be referred to as an automated guided vehicle (AGV)) is located in each of the sections 121A to 121J throughout the clean room 121. Some sections (e.g., section 121A) can have more than one cleaning tool 123 (e.g., two cleaning tools 123). Some other sections (e.g., sections 121I, 121J) may not have a cleaning tool 123. Some sections (e.g., sections 121G, 121H) can share a cleaning tool 123.

[0030] Figure 1C FIG. 1 is a schematic partial side view showing section 121A according to various embodiments. Fabrication tool 122A and cleaning tool 123 are placed on floor 1200, which in some embodiments may be a raised floor. Tool attachment 1220 in communication with fabrication tool 122A (e.g., in fluid, electrical, data, or other ways) may be located beneath floor 1200. In some embodiments, tool attachment 1220 is a refrigerator / freezer that can be used to cool components of fabrication tool 122A and / or materials consumed by fabrication tool 122A.

[0031] In some embodiments, multi-channel sampling system 132 is located beneath floor 1200. Sampling manifold 1310 is connected to a plurality of sampling ports 1321, 1322. Sampling port 1322 may be directly connected to a port (e.g., an exhaust port) of tool attachment 1220. Sampling port 1321 may be connected to intake port 1323. Each set of one sampling port 1321 connected to intake port 1323 may be collectively referred to as sampling unit 1325. In some embodiments, a single sampling port 1321 may be connected to more than one intake port 1323.

[0032] In Figure 1B , one or more sampling units 1325 may be located in each of sections 121A to 121J. Some of sections 121A to 121J may include sampling units 1325 with a higher density than other sections of sections 121A to 121J. For example, section 121A may include sampling units 1325 with a higher density than section 121B. Although each of sections 121A to 121J is shown as having at least one of sampling units 1325, in some other embodiments, some of sections 121A to 121J may not have sampling units 1325. In some embodiments, sampling units 1325 may be arranged in a uniform two-dimensional array. As shown in Figure 1B , the arrangement of sampling units 1325 may be non-periodic above cleanroom 121 and / or within each of sections 121A to 121J. For example, section 121A may have a higher density than sections 121B to 121J, and within section 121A, high-density sampling area 1327 may include a higher density than other areas within section 121A. High-density sampling area 1327 may correspond to areas within section 121A that are more sensitive to contaminants, more likely to release contaminants (e.g., “hot spots”), or contain more dangerous (e.g., for yield, safety, or the like) contaminants. The density of sampling units 1325 may be measured as the ratio of the number of sampling units 1325 per one hundred square meters (m 2 ) to the area of cleanroom 121. In some embodiments, the density is between about 1 / m2 to about 50 / m 2 (or about 10 / m 2 to about 30 / m 2 ). Below about 1 / m 2 the sensitivity to contaminant leakage may not be sufficient to identify the manufacturing tool 122 or tool attachment 1220 of the leaked contaminant.

[0033] Figure 1D is a schematic block diagram of an AGV 123 according to various embodiments. The drive system 123A of the AGV 123 is configured to move the AGV 123 in at least a two-dimensional space. The drive system 123A may include one or more motors, one or more wheels, and one or more axles. The drive system 123A receives power (e.g., electricity) from a power system 123B. The drive system 123A also receives data (e.g., control signals for controlling motor rotation speed, wheel / axle rotation angle, and similar parameters) from a control system 123F via a data connection, which may include one or more electrical signal wirings and / or optical signal wirings or one or more wireless network ICs.

[0034] The filtration system 123C of the AGV 123 is configured to remove contaminants from the air in the clean room 121. The filtration system 123C may include one or more filters 123C2, and the one or more filters 123C2 may include chemical filters, high-efficiency particulate air (HEPA) filters, combinations thereof, or the like. The filters 123C2 are in fluid communication with one or more fans 123C1, and the one or more fans 123C1 may be circular fans, blowers, exhaust fans, or the like. In some embodiments, the fans 123C1 are in fluid communication with the filters 123C2 through one or more ducts 123C3. In some embodiments, the fans 123C1 are in fluid communication with the filters 123C2 through the housing of the filtration system 123C, and there are no additional ducts 123C3 in the filtration system 123C. The power system 123B provides power to the filtration system 123C, for example, to drive the rotation of the fans 123C1. In some embodiments, the filtration system 123C has an on-board power system (not shown separately), and the on-board power system may at least include a power monitoring / regulation IC. In such embodiments, the filtration system 123C may receive alternating current (AC) power from the power system 123B, and the on-board power system of the filtration system 123C may convert the AC power into direct current (DC) power to drive the fans 123C1. Such a configuration may provide improved modularity of the AGV 123, such that the filtration system 123C can be easily removed from the AGV 123, enabling the AGV 123 to be used for other purposes in the clean room 121.

[0035] The filtration system 123C of the AGV 123 can be connected to the control system 123F through a data connection. In some embodiments, the control system 123F of the AGV 123 can control the parameters of the filtration system 123C (e.g., power on / off, fan speed, or similar parameters) through a data connection. In some embodiments, the filtration system 123C reports operating parameters (e.g., filter condition / age) to the control system 123F through a data connection.

[0036] The network system 123D of the AGV 123 is configured to transmit data to and / or receive data from an external network, for example, via the network device 155 of the control center 150. In some embodiments, the network system 123D includes a wireless network device (such as one or more ICs and one or more antennas), and the wireless network device can achieve data communication through the network using protocols such as WiFi, the fourth generation mobile network (4G) / 5G, or other suitable data communication protocols. In some embodiments, the network system 123D communicates with the network device 155. The network system 123D also communicates with the control system 123F, and the control system 123F can receive commands from the AGV controller 154 via the network system 123D. In some embodiments, the data communication between the network system 123D and the control system 123F is carried out through one or more wiring lines (such as electrical wiring lines or optical wiring lines), or through the air via two or more communication ICs (such as Bluetooth, NFC, or other similar ICs) respectively located in the network system 123D and the control system 123F. Refer to Figure 2A To elaborate more deeply on the details of the commands and operations of the AGV 123.

[0037] In some embodiments, the AGV 123 further includes a sensor system 123E, and the sensor system 123E has one or more sensors, and the one or more sensors are configured to provide at least information about the environment in which the AGV 123 operates to the control system 123F. In some embodiments, the one or more sensors include at least one of an optical sensor (such as a camera, an infrared sensor, or the like), a vibration sensor, an air quality sensor, or other suitable sensors. The sensors can enhance the ability of the AGV 123 to navigate and detect obstacles, or provide real-time local information about the air quality in the surrounding area of the AGV 123.

[0038] The control system 123F may include a microcontroller unit (MCU), a computer processing unit (CPU), a graphic processing unit (GPU), an input / output (I / O) IC, a bus, a memory, a data storage (e.g., a solid state drive, a hard disk drive, etc.), and the like. The I / O IC and the bus are generally configured to enable one-way or two-way data communication between the control system 123F and the drive system 123A, the power system 123B, the filtration system 123C, the network system 123D, and the sensor system 123E, and the drive system 123A, the power system 123B, the filtration system 123C, the network system 123D, and the sensor system 123E may be collectively referred to as "systems 123A to 123E". For example, the control system 123F may communicate with systems 123A to 123E (e.g., transmit data to and / or receive data from systems 123A to 123E) via one or more universal serial bus (USB), peripheral component interconnect (PCI), and / or serial advanced technology attachment (SATA) buses. In some embodiments, the control system 123F communicates with systems 123A to 123E via an optical data bus (e.g., via an optical fiber). In such embodiments, similar buses exist in systems 123A to 123E to enable data communication with the control system 123F.

[0039] Figure 2A is a schematic diagram showing a process 20 for managing contaminants in a clean room 121 according to various embodiments. The process 20 includes operations 200A to 200G, 220A to 200D, with certain operations thereof illustrated in more detail with reference to Figures 2B to 2D which are shown in more detail. The process 20 will be further elaborated based on one or more embodiments. It should be noted that the operations of the process 20 may be rearranged or otherwise modified within the scope of the various aspects. It should also be noted that additional processes may be provided before, during, and after the process 20, and some other processes may only be briefly elaborated herein.

[0040] In operation 200A, the AMC in-line detection system (which can be system 100) scans cleanroom 121, which can include scanning the surrounding environment of cleanroom 121 and scanning fabrication tool 122. When scanning the surrounding environment and / or fabrication tool 122, system 100 can scan chemical compounds, which can include contaminants (e.g., TVOC) and / or process chemicals (e.g., acids, bases, solvents, or the like).

[0041] In some embodiments, scanning fabrication tool 122 includes determining the process chemicals associated with fabrication tool 122. In some embodiments, the process chemicals are recorded in a database (e.g., database 152). As Figure 2A shown, database 152 can include one or more separate databases, such as image database 152A and tool database 152B. Tool database 152B can store information from one or more sources (e.g., SCADA system 1521 (e.g., AMC SCADA 151) that records the acids and / or solvents used by fabrication tool 122). The sources can also include a fault detection system 1522, a tool utility information log 1523 (e.g., an electronic blue book), a tool chemical information record 1524, or the like.

[0042] In some embodiments, scanning the surrounding environment includes one or more separate operations, such as sampling the surrounding environment by a multi-channel sampling system 132 (as referenced Figure 1C above) and analyzing the samples obtained from multi-channel sampling system 132 by TOF-MS 142. The output of scanning the surrounding environment can include the levels of various elements (e.g., carbon, chlorine, fluorine, and the like) in parts per billion or another suitable metric.

[0043] The scanning of the surrounding environment can be implemented in a variety of different ways. In some embodiments, reference Figure 1B, the multi-channel sampling system 132 can output samples from the sampling units 1325 in a specific order. For example, the sampling units 1325 can be sorted by sections 121A to 121J such that the multi-channel sampling system 132 outputs samples from all the sampling units 1325 in a specific section (e.g., section 121A) before outputting samples from a different section (e.g., section 121B). In some embodiments, the multi-channel sampling system 132 can output samples based on a priority list. For example, a first subset of the sampling units 1325 can be associated with a high incidence of spikes in the contaminant level, a second subset of the sampling units 1325 can be associated with a medium incidence of spikes, and a third subset of the sampling units 1325 can be associated with a low incidence of spikes. In this way, samples can be extracted from the sampling units 1325 in the first subset before the sampling units 1325 in the second subset, and then samples can be extracted from the sampling units 1325 in the second subset before samples from the sampling units 1325 in the third subset. In some embodiments, the priority list is generated based on historical sampling data collected over days, months, or years during which the sampling units 1325 are sampled. In some embodiments, a machine learning algorithm is used to generate the priority list, which can be similar to the process described in more detail with reference to Figures 4A to 4B Once all the sampling units 1325 have been sampled, the multi-channel sampling system 132 can be restarted.

[0044] In some embodiments, the multi-channel sampling system 132 outputs samples from the sampling units 1325 in a random order or a pseudo-random order, and the random order or pseudo-random order can be combined with any of the above techniques. For example, the multi-channel sampling system 132 can randomly sample the sampling units 1325 in a certain section (e.g., section 121A) until all the sampling units 1325 in that section have been sampled, and then it can move to a different section (e.g., section 121B) to randomly sample the sampling units 1325 in that section until all the sampling units 1325 in the different section have been sampled. In some embodiments, a first subset of the sampling units 1325 can be randomly sampled, then a second subset can be randomly sampled, and then a third subset can be randomly sampled.

[0045] In some embodiments, with Figure 2AThe operation 200B described above corresponds to generating a real-time pollutant distribution map for various AMC pollutants. A distribution map can be generated for each AMC pollutant, and / or a single distribution map can be generated that includes concentration level information associated with multiple AMC pollutants. The distribution map can be an image file, where each pixel of the image file corresponds to a physical location (e.g., a zone of unit area) in the clean room 121 and has a color and / or brightness level corresponding to the measured level of one or more pollutants. An image file of the type just described can be referred to as a heat map. In some embodiments, the sampling unit 1325 and / or the fabrication tool 122 are mapped to the pixels of the image file by location. In some embodiments, a single sampling unit 1325 can correspond to one or more pixels, which may depend on the resolution (width * length) of the image file. Similarly, each fabrication tool 122 can correspond to a number of pixels. In some embodiments, each fabrication tool 122 is represented as one or more point sources in the image file. For example, the fabrication tool 122A can include one or more zones (from a top view orientation) that are more prone to leakage, such as input / output ports, exhaust ports, seals, or other similar zones. In such an instance, the fabrication tool 122A can be represented in the image file as one or more different point sources corresponding to the zones that are more prone to leakage, with each point source mapped to one or more pixels of the image file. Based on the color and / or brightness level at each mapped pixel, further interpolation can be performed at pixels that do not correspond to any sampling unit 1325 or fabrication tool 122, such that each pixel in the image file can contain an amplitude (color and / or brightness level) corresponding to either the measured level or the interpolated level.

[0046] In operation 200C, one or more real-time pollutant distribution maps are compared with the AMC images stored in a database (e.g., the image database 152A). In some embodiments, the AMC images stored in the image database 152A include, for example, AMC diffusion images generated by computational fluid dynamics (CFD) techniques for different compounds at different concentrations and different times. In some embodiments, CFD calculations are performed according to the layout of the fabrication tool 122, and the CFD calculations are stored in the image database 152A. In some embodiments, when performing CFD calculations for a particular fabrication tool 122 (e.g., the fabrication tool 122A), it is assumed that the fabrication tool 122A is the source of pollutant leakage, and the pollutant distribution and the location of the peak concentration level are calculated by CFD techniques. As Figure 2AAs shown, for each fabrication tool 122, the image database 152A may store a plurality of contamination profile images 210A through 210N, which may correspond to contaminant distributions calculated for the fabrication tool 122 using various conditions (such as compounds, concentrations, times, and the like). The peak concentration level positions may be stored in the contamination profile images 210A through 210N, or the peak concentration level positions may be stored as a separate file (such as a text file).

[0047] Figure 2B FIG. 210 shows a diffusion image 210 according to various embodiments. The illustrated diffusion image 210 includes seven rows and five columns, but may include more or fewer rows and / or columns in various embodiments. Each column includes seven regions. For example, the first column includes regions 230A1 through 230A7, the second column includes regions 230B1 through 230B7, the third column includes regions 230C1 through 230C7, the fourth column includes regions 230D1 through 230D7, and the fifth column includes regions 230E1 through 230E7. Fabrication tools 222A1 through 222A6 (which may be fabrication tool 122) may be mapped onto the diffusion image 210. In some embodiments, the fabrication tools 222A1 through 222A6 do not exist as data in the diffusion image 210, which may be indicated by the dashed lines in Figure 2B FIG. 210. Each of the fabrication tools 222A1 through 222A6 corresponds to one or more of the regions 230A1 through 230E7. For example, fabrication tool 222A1 corresponds to six regions 230A1 through 230A3, 230B1 through 230B3, fabrication tool 222A2 corresponds to three regions 230C1 through 230C3, fabrication tool 222A3 corresponds to six regions 230D1 through 230D3, 230E1 through 230E3, fabrication tool 222A4 corresponds to four regions 230A5, 230A6, 230B5, 230B6, fabrication tool 222A5 corresponds to two regions 230C5, 230C6, and fabrication tool 222A6 corresponds to four regions 230D5, 230D6, 230E5, 230E6. Depending on the size of each of the fabrication tools 222A1 through 222A6 and the size of each region, each of the fabrication tools 222A1 through 222A6 may correspond to one region or more than six regions.

[0048] Figure 2CShown are contamination distribution images 210A to 210F respectively associated with fabrication tools 222A1 to 222A6. Each of the contamination distribution images 210A to 210F includes a peak concentration level (shown by hatching) in one of regions 230A1 to 230E7. The peak concentration level may correspond to the peak concentration of the one or more compounds. For example, the contamination distribution image 210A corresponding to fabrication tool 222A1 includes a peak concentration level at region 230B3, the diffusion image 210B corresponding to fabrication tool 222A2 includes a peak concentration level at region 230D3, the diffusion image 210C corresponding to fabrication tool 222A3 includes a peak concentration level at region 230E3, the diffusion image 210D corresponding to fabrication tool 222A4 includes a peak concentration level at region 230B7, the diffusion image 210E corresponding to fabrication tool 222A5 includes a peak concentration level at region 230C7, and the diffusion image 210F corresponding to fabrication tool 222A6 includes a peak concentration level at region 230E7.

[0049] Generally, the peak concentration level exists in only one region of each of the contamination distribution images 210A to 210F. In some embodiments, each of the contamination distribution images 210A to 210F may include a probability associated with each of the regions 230A1 to 230E7. The probability may be the probability that a peak concentration level will occur in a particular one of the regions 230A1 to 230E7 given the conditions (compound, concentration, time) of a leak of contaminants from fabrication tools 222A1 to 222A6. In such a configuration, the shaded regions in the contamination distribution images 210A to 210F may represent the regions with the highest probability of having a peak concentration level for the given conditions. Refer to operation 200D and Figure 2C provides a further illustration of the probability.

[0050] As previously described, in operation 200C, one or more real-time contaminant distribution maps are compared with the AMC images stored in a database (such as image database 152A), as Figure 2C shown. A distribution map 240 is shown, which has a peak concentration level at location 241 indicated by a star. The distribution map 240 may be determined as described with reference to Figures 1A to 1D above.

[0051] Refer to Figure 2D, to determine which of the fabrication tools 222A1 to 222A6 is most likely the source of the peak concentration level at location 241, the contamination distribution images 210A to 210F that most closely match the distribution map 240 can be selected. In some embodiments, the selection is made by determining the regions 230A1 to 230E7 in which location 241 lies. For example, in the distribution map 240, location 241 may have coordinates 241C, such as "(1343, 384)". Continuing the example, using the region table 230T, the regions 230A1 to 230E7 corresponding to the coordinates 241C can be determined to be region 230D3 (see the boxed row in region table 230T). The determination of the contamination distribution images 210A to 210F corresponding to the coordinates 241C can also be made by looking up the regions 230A1 to 230E7 (e.g., region 230D3) corresponding to location 241 in the image table 210T. Using the image table 210T and again continuing the example, the diffusion image 210B can be selected from the image table 210T in the row corresponding to region 230D3. In this way, the diffusion image 210B can be selected as the one that most closely matches the coordinates 241C of location 241 presenting the peak concentration level of the contaminant.

[0052] In operation 200D, referring again to Figure 2A , having selected the contamination distribution images 210A to 210F that best match the distribution map 240, the highest probability fabrication tools 222A1 to 222A6 can be set as the primary suspects 251 (see Figure 2C , Figure 2F ) for causing the peak concentration level at location 241 in the distribution map 240. In some embodiments, setting the primary suspects 251 can be based on the AMC leakage probability of each of the fabrication tools 222A1 to 222A6 relative to the contamination distribution images 210A to 210F that best match the distribution map 240. Continuing the above example and referring to Figure 2C , a probability map 250B corresponding to the diffusion image 210B can be generated. In some embodiments, the probability map 250B can be or include a probability table 250BT, as shown in Figure 2E . In some embodiments, the percentages listed in the probability table 250BT are prediction / confidence scores generated by machine learning / artificial intelligence techniques or an inverse method, for example, as described in Figures 4A to 4B . In such embodiments, the percentages listed for each of the contamination distribution images 210A to 210F can total more than 100%. Similarly, the total percentages for all of the contamination distribution images 210A to 210F of each of the fabrication tools 222A1 to 222A6 can total more than 100%. As in Figure 2EAs shown, in operation 200D, fabrication tool 222A2 has the highest percentage associated with the diffusion image 210B. Thus, in operation 200D, fabrication tool 222A2 is set as the primary suspect 251.

[0053] Referring Figures 2B to 2E The above description of operations 200C and 200D can correspond to a simplified embodiment of process 20, where a single peak concentration level of a single contaminant is identified for profile 240, and the single peak concentration level of the single contaminant is compared with contamination distribution images 210A to 210F to find the contamination distribution image 210A to 210F that best matches the peak concentration level in profile 240. In some embodiments, a machine learning / artificial intelligence (ML / AI) process (e.g., referring Figures 4A to 4B as described) can be used to perform more complex comparisons that can incorporate larger and more diverse sets of inputs, including peak concentration levels and gradients / distributions of multiple contaminants and / or TVOCs, information from tool database 152B, and other suitable inputs. The ML / AI process can also be configured to identify at least two simultaneous and distinct contaminant leaks originating from two or more fabrication tools 222A1 to 222A6. Referring Figures 4A to 4B provides a further description of the ML / AI process.

[0054] Referring Figure 2C , after identifying and setting fabrication tool 222A2 as the primary suspect 251 in operation 200D, advanced verification is performed in operation 200E. In some embodiments, the advanced verification includes comparing the primary suspect 251 (e.g., fabrication tool 222A2) with various process utility information. In some embodiments, the process utility information is stored in tool database 152B. The process utility information can include information from one or more sources 1521 to 1524.

[0055] As Figure 2C shown, the first source 1521 can include information about the supply of one or more acids and / or solvents recorded by SCADA, where SCADA can be a plant-side SCADA independent of the AMC SCADA 151. In some embodiments, the advanced verification is or includes determining whether an acid and / or solvent was sent to fabrication tool 222A2 during the generation of profile 240. If fabrication tool 222A2 was not supplied with acid / solvent during (part or all of) the generation of profile 240, a first mismatch flag can be set, indicating a mismatch between fabrication tool 222A2 and profile 240. If an acid / solvent was supplied during the generation of profile 240, a first match flag (or simply no flag) can be set, indicating a potential match between fabrication tool 222A2 and profile 240.

[0056] The second source 1522 may include information from a fault detection system. In some embodiments, the advanced verification is or includes comparing the fabrication tool 222A2 with the information generated by the fault detection system. For example, the fault detection system may indicate the operating state (e.g., active, stopped, under maintenance, under repair, or a similar state) of each of the fabrication tools 222A1 to 222A6. If the information in the tool database 152B from the second source 1522 indicates that the fabrication tool 222A2 is inactive (e.g., stopped, under maintenance, under repair, or a similar state) for at least a portion (or all) of the time when the distribution map 240 is being generated, then a second mismatch flag may be set, indicating a mismatch between the fabrication tool 222A2 and the distribution map 240. If the fabrication tool 222A2 is active (as indicated by the tool database 152B) for a portion (or all) of the time during which the distribution map 240 is being generated, then a second match flag (or simply no flag) may be set, indicating a potential match between the fabrication tool 222A2 and the distribution map 240.

[0057] The third source 1523 may include information from a utility information log, which may be referred to as electronic utility information (e.g., an electronic blue book or "e - bluebook"). In some embodiments, the advanced verification is or includes comparing the contaminants of the distribution map 240 with the materials used by the fabrication tool 222A2 when the distribution map 240 is being generated. If the contaminant that is the primary suspect 251 for leakage for the fabrication tool 222A2 is not associated with the materials listed in the third source 1523 in the tool database 152B, then a third mismatch flag may be set, indicating a mismatch between the materials used by the fabrication tool 222A2 and the contaminants that the fabrication tool 222A2 is suspected of leaking. If the contaminant is associated with the materials listed in the tool database 152B, then a third match flag (or simply no flag) may be set, indicating a potential match between the materials used by the fabrication tool 222A2 and the contaminants of the distribution map 240.

[0058] The fourth source 1524 may include information from a tool chemical log. In some embodiments, the advanced verification is or includes comparing the chemicals used within fabrication tool 222A2 with the contaminants detected when generating the profile map 240. For example, the tool chemical log may record the gas precursors used by certain fabrication tools 222A1 to 222A6 (e.g., for chemical vapor deposition (CVD) tools, atomic layer deposition (ALD) / atomic layer etching (ALE) tools, or similar tools). If the information in the tool database 152B from the fourth source 1524 indicates a chemical that is not associated with the contaminants detected when generating the profile map 240, then a fourth mismatch flag may be set, indicating a mismatch between fabrication tool 222A2 and the profile map 240. If fabrication tool 222A2 is indicated as using a chemical that is associated with the contaminants detected when generating the profile map 240, then a fourth match flag (or simply no flag) may be set, indicating a potential match between fabrication tool 222A2 and the profile map 240. The first through fourth mismatch flags may be collectively referred to as "mismatch flags", and the first through fourth match flags may be collectively referred to as "match flags".

[0059] As Figure 2F shown, the primary suspect 251 may be compared against the four sources 1521 to 1524. If no mismatch flag is set after the comparison, then fabrication tool 222A2 may be identified as the source of the contaminant leak, corresponding to the "yes" branch of decision block 200F. If any of the mismatch flags are set in operation 200E, then process 20 may return to operation 200A corresponding to the "no" branch of decision block 200F. Optionally, process 20 may return to operation 200D, and a fabrication tool 222A1 to 222A6 having the second highest percentage in the probability table 250BT may be selected. Continuing the previous example, if it is found that the initially selected fabrication tool 222A2 in operation 200D has not been supplied with acid / solvent (first mismatch flag set), is inactive (second mismatch flag set), the materials used do not match the detected contaminants (third mismatch flag set), and / or the chemicals used do not match the detected contaminants (fourth mismatch flag set), then in a subsequent iteration of operation 200D after the "no" branch of operation 200F, fabrication tool 222A1 may be selected as the primary suspect 251 because fabrication tool 222A1 has the second highest percentage in the row associated with the diffusion image 210B.

[0060] Once the primary suspect 251 is identified in operation 200F, one or more actions 220A to 220D may be taken to resolve the contaminant leak, corresponding to Figure 2AThe operation shown is 200G. In some embodiments, under the control of the AGV controller 154 (see Figure 1A ), actions 220A to 220D are taken based on the peak concentration level. Actions 220A to 220D may correspond to one or more thresholds. In the configuration shown in Figure 2A , the first threshold corresponds to the first action 220A, the second threshold corresponds to the second action 220B, the third threshold corresponds to the third action 220C, and the fourth threshold corresponds to the fourth action 220D.

[0061] In some embodiments, the first threshold may be a baseline threshold. The baseline threshold may be a concentration level below which the AGV 123 should not take any action. In some embodiments, the baseline threshold is a fixed concentration level in the range of from about 0.04 ppb to about 1 ppm, but other ranges may be suitable depending on, for example, the process node or safety considerations. In some embodiments, the baseline threshold is dynamic and may be adjusted according to one or more process parameters (such as ambient temperature, production load (e.g., the number of fabrication tools 122 operating simultaneously), process node (e.g., certain nodes that are more sensitive to contaminants than other nodes), or other suitable process parameters). In some embodiments, the baseline threshold is set by the AMCSCADA 151. When the peak concentration level of the profile 240 is below the first threshold, the scanning of the clean room 121 (e.g., the update / generation of the profile 240) continues. As long as the peak concentration level is below the first threshold, generally the AGV 123 is not activated to clean the air specifically, and the AGV 123 is located at the charging station. During the first action 220A, the AGV 123 may be periodically activated for inspection or other purposes.

[0062] The second threshold may be a warning (e.g., "above warning") threshold. In some embodiments, the second threshold is a concentration level that is substantially equal to the first threshold. When the peak concentration level of the profile 240 is above the second threshold, the second action 220B is taken. In some embodiments, the second action 220B includes the AGV controller 154 transmitting one or more commands from the network device 155 to the network system 123D of the AGV 123. The commands may be received by the control system 123F and may include activation (e.g., power-on) commands and / or dispatch commands.

[0063] The dispatch command may include a location, which may be a code corresponding to a destination (e.g., a manufacturing tool 122 or a waypoint near the manufacturing tool 122). In some embodiments, the control system 123F may include or have access to an internal database (inside the AGV 123) and may look up route instructions, coordinates, or other relevant information for navigating to the manufacturing tool 122 or the waypoint. In some embodiments, the dispatch command itself includes the relevant information (e.g., the coordinates of the manufacturing tool 122 or the waypoint) or its route indication, such that the control system 123F may not have a navigation database and / or algorithm. In some embodiments, the position of the AGV 123 is tracked, and the AGV controller 154 may send continuous navigation commands to the AGV 123 to direct the movement of the AGV 123 towards the intended destination (e.g., the manufacturing tool 122 or the waypoint).

[0064] In some embodiments, there is one or more waypoint stations in the clean room 121 (e.g., in each section 121A to 121J of the clean room 121). The waypoint station may include a broadcast system for guiding the AGV 123 to the waypoint station. In some embodiments, the broadcast system includes an audio system, a visual system, an electronic system, or other suitable broadcast systems. In some embodiments, the visual system may include an infrared transmitter. The AGV 123 may receive a code corresponding to the waypoint station and may automatically move towards the waypoint station based on signals from the broadcast system. For example, the AGV 123 may move towards the waypoint station by detecting an increase in the intensity of a signal (e.g., an infrared signal) and moving in the direction where the signal is stronger than before.

[0065] In some embodiments, the waypoint station includes a visual marker (e.g., a Quick Response (QR) code or other visual pattern), and the sensor system 123E of the AGV 123 (e.g., one or more cameras) may be used to detect the visual pattern and guide the AGV 123 towards the visual pattern. In some embodiments, the broadcast system may be used in combination with the visual marker to provide faster and more robust guidance of the AGV 123 towards the waypoint station.

[0066] Once at the destination, the control system 123F can stop the movement of the AGV 123 (e.g., by stopping the drive system 123A), and can further control the drive system 123A to rotate the AGV 123 to orient the fan 123C1 and the filter 123C2 in an appropriate direction for cleaning the air in the surrounding area of the location of the AGV 123. In some embodiments, the appropriate direction is determined based on the distribution map 240 (e.g., by determining the profile / gradient of the contaminant flow). Based on the profile / gradient of the flow, the control system 123F can orient the filter 123C2 in the upstream direction of the flow and the fan 123C1 in the downstream direction of the flow. Once at the destination and optionally once the orientation of the AGV 123 is set, the AGV 123 can be considered to be in a standby mode with the fan 123C1 not turned on. In the second action 220B, generally the fan 123C1 is not turned on because the peak concentration level has not reached the third threshold or the fourth threshold. If the peak concentration level increases to exceed the third threshold, the fan 123C1 can be briefly turned on during standby to verify the function of the fan 123C1 in preparation for cleaning in the third action 220C.

[0067] The third threshold can be a control (e.g., "out of control" or "OOC") threshold. In some embodiments, the control threshold is a fixed concentration level in the range of from about 0.1 ppb to about 1 ppm, but other ranges may be suitable depending on, for example, the process node or safety considerations. In some embodiments, the control threshold is dynamic and can be adjusted based on one or more process parameters (e.g., ambient temperature, production load (e.g., the number of fabrication tools 122 operating simultaneously), process node (e.g., certain nodes that are more sensitive to contaminants than other nodes), or other suitable process parameters). Generally, the third threshold is higher than the second threshold, whether fixed or dynamic. For example, the third threshold can be higher than the second threshold by a fixed amount, such as 10 ppb, 1 ppb, 0.2 ppb, or another suitable fixed amount. In some embodiments, the third threshold is higher than the second threshold by a certain percentage (e.g., about 100%, about 50%, about 20%, or other suitable percentage).

[0068] When the peak concentration level of the distribution map 240 exceeds the third threshold, a third action 220C is taken. In some embodiments, the third action 220C includes activating (e.g., powering on) the fan 123C1. In some embodiments, the fan 123C1 is configured to rotate at one or more speeds (e.g., revolutions per minute or "RPM"). In some embodiments, the speed of the fan 123C1 is set simultaneously with or shortly after activating the fan 123C1. In some embodiments, the speed of the fan 123C1 varies dynamically based on the peak concentration level. In some embodiments, the speed of the fan 123C1 is initially set at the maximum level to reduce the peak concentration level as quickly as possible. When it drops below the third threshold, the speed of the fan 123C1 can be set at an intermediate level below the maximum level to keep the peak concentration level below the third threshold. In some embodiments, the fan 123C1 can continue to operate until the peak concentration level drops to the baseline threshold.

[0069] Figure 2G Illustrates a third action 220C according to various embodiments. Curve 270 shows the peak concentration level, and curve 280 shows the fan speed waveform. Before time t1, the peak concentration level is below the first threshold and / or the second threshold, and the first threshold and / or the second threshold can be uniformly labeled as T Figure 2G in which B / W . At time t1, the peak concentration level suddenly increases above the third threshold T C while remaining below the fourth threshold T S . In this case, once the third threshold T C is exceeded, the AGV controller 154 can command one or more of the AGVs 123 to move to the fabrication tool 122 or the waypoint, and turn on the fan 123C1 immediately when arriving or shortly before arriving. When the fan 123C1 is turned on, the speed is set to the maximum speed S MAX at time t2. There may be a delay between the time t1 when the AGV 123 is dispatched and the time t2 when the AGV 123 arrives and the fan 123C1 is turned on, as Figure 2G shown in. In some embodiments, for less severe spikes, the AGV 123 can standby at the fabrication tool 122 or the waypoint before the peak concentration level exceeds the third threshold T C , such that there is a much shorter delay between time t1 and time t2, e.g., the time for detecting the peak concentration level plus the time from the AGV controller 154 transmitting a signal to the fan 123C1 being activated by the control system 123F.

[0070] At time t3, the peak concentration level drops below the third threshold T CBelow. In some embodiments, in response to the peak concentration level dropping to a third threshold T C below, the fan speed is reduced to an upper intermediate speed S2, where the upper intermediate speed S2 is lower than the maximum speed S MAX . Between time t3 and time t4, the peak concentration level continues to drop until it reaches or falls below the first threshold T B / W . In some embodiments, in response to reaching or falling below the first threshold T B / W , the fan speed is reduced to a lower intermediate speed S1, where the lower intermediate speed S1 is lower than the upper intermediate speed S2. In some embodiments, the fan 123C1 can be turned off such that the fan speed is substantially zero, e.g., 0 RPM. By maintaining the fan speed at the lower intermediate speed S1, the peak concentration level can be controlled to be less than or substantially equal to the first threshold T B / W .

[0071] In the foregoing description of Figure 2G , the fan speeds are illustrated using discrete speeds S1, S2, S MAX . In some embodiments, the fan speed can be finely controlled, e.g., according to a gradient set by pulse width modulation (PWM), analog control, or another suitable technique. In such embodiments, the fan speed can track the peak concentration level and establish a stable concentration level through, e.g., a negative feedback control loop.

[0072] In some embodiments, with reference to Figure 2H , during the third action 220C, the system 100's measurement of the peak concentration level can be focused on a single sampling unit 1325 or a group of sampling units 1325 (e.g., in the high-density sampling area 1327), while temporarily excluding sampling by the remaining sampling units 1325 of the system 100. For example, as shown in Figure 2H , the location 241 of the peak concentration level can be near three sampling units 1325A to 1325C. The location 241 is also within the high-density sampling area 1327. By sampling only a single sampling unit 1325 (e.g., sampling unit 1325B) or a small group of sampling units 1325 (e.g., sampling units 1325A to 1325C, or all sampling units 1325 in the high-density sampling area 1327) located, for example, at the location associated with the peak concentration level or in the surrounding area of the fabrication tool 122, the AGV controller 154 can respond more quickly to changes in the peak concentration level without having to wait for other sampling units 1325 distal to the location 241 to sample, where each sampling unit 1325 may take 1 to 2 minutes. Such a focused sampling scheme can be advantageous for establishing the negative feedback loop described with reference to Figure 2G .

[0073] In some embodiments, the third action 220C further includes activating an alarm or notification, which may be routed to the manufacturing and / or factory operator. In response to the alarm or notification that may indicate that the fabrication tool 122 is identified as the source of the contaminant leak and the associated peak concentration level, the operator may perform on-site verification of the leak, shut down the fabrication tool 122, perform maintenance and / or repair, schedule maintenance and / or repair, or take another suitable action to prevent and / or repair the contaminant leak.

[0074] Referring again to Figure 2A , the fourth threshold may be a specification (e.g., "out of specification", "OOS") threshold. The fourth threshold may be set to a level at which the fabrication tool 122 can be shut down to prevent safety and / or yield incidents. For example, the TVOC concentration or the concentration of individual contaminants may be toxic at a particular concentration level, or may affect the yield at other concentration levels. The fourth threshold may be set to a level far below the toxicity level or the yield impairment level, e.g., less than 50%, less than 20%, less than 10%, or less than 1% of the toxicity level or the yield impairment level.

[0075] When the fourth threshold is reached and / or exceeded, a fourth action 220D is taken. In some embodiments, the fourth action 220D includes taking measures to prevent the leak that causes the peak contaminant level to be higher than the fourth threshold. In some embodiments, the measures include preventing the FOUP from reaching the fabrication tool 122, stopping the generation through the fabrication tool 122, or one or more of other suitable measures for preventing the leak. Stopping the generation may include placing the fabrication tool 122 in a standby / idle mode such that the fabrication tool 122 remains powered on but does not perform a generation process on the semiconductor wafer. Stopping the generation may include powering off the fabrication tool 122. Stopping the generation may be automatically performed by SCADA (e.g., factory-side SCADA). In some embodiments, the fourth action 220D includes generating an alarm and / or sending a notification to the operator identifying the fabrication tool 122 and the peak concentration level, and the operator manually stops the generation through the fabrication tool 122.

[0076] Due to the delay between sampling events at the same sampling unit 1325, the measured peak concentration level may appear to surge rapidly from below the first threshold directly to above the fourth threshold, such that the first measurement of the location 241 by the nearby sampling unit 1325 is below the first threshold, and the immediately following second measurement by the nearby sampling unit 1325 is above the fourth threshold. In this way, if the fourth threshold is exceeded before the second action 220B or the third action 220C is taken, the AGV 123 may not have been dispatched to the location 241 before the fourth threshold is reached. In this case, when the generation by the production tool 122 is stopped in the fourth action 220D, the AGV controller 154 may not dispatch the AGV 123 in response to the peak concentration level exceeding the fourth threshold. Similarly, if the AGV 123 has cleaned the air near the location 241 under the third action 220C and exceeded the fourth threshold, then in the fourth action 220D, the AGV 123 may be recalled, for example, to a charging station.

[0077] Figure 3A A flowchart associated with a process 30 for fabricating an IC device according to various embodiments is shown. The process 30 may be compared to the process 30 of FIG. Figures 1A to 2H The described system 100 is used in conjunction with the process 20. The process 30 will be further described according to one or more embodiments. It should be noted that the operations of the process 30 may be rearranged or otherwise modified within the scope of various aspects. It should also be noted that additional processes may be provided before, during, and after the process 30, and some other processes may only be briefly described herein.

[0078] In operation 300, a wafer is positioned in a clean room (e.g., clean room 121, Figure 1A , Figure 1B ) in the first production tool (e.g., production tool 222A1, Figure 2B ). In some embodiments, positioning the wafer includes delivering the wafer in the FOUP to a port of the first fabrication tool. The door of the FOUP may be opened, and the robotic arm may pick up the wafer and transfer the wafer to a chamber of the first fabrication tool. The robotic arm or a second robotic arm may position the wafer on a wafer stage in the first fabrication tool for processing, such as deposition, etching, cleaning, annealing, or the like.

[0079] In operation 310, a peak contaminant level in the clean room that is above a first threshold may be detected. In some embodiments, the peak contaminant level is detected by AMC SCADA (e.g., AMC SCADA 151) based on measurements by an analysis system (e.g., analysis system 140 including TOF-MS 142). In some embodiments, the peak contaminant level is a peak contaminant level of profile 240 at location 241. The first threshold in operation 310 may correspond to a reference value.Figure 2A The warning threshold or control threshold described.

[0080] In operation 320, the second fabrication tool (e.g., fabrication tool 222B, Figure 2B ) is identified as the source of the contaminant expected to be at peak contaminant levels. The prediction for the second fabrication tool can be performed using operations 200A to 200F of process 20 described with reference to Figures 2A to 2H .

[0081] In operation 330, the AGV cleaner (e.g., AGV 123) is dispatched to the location corresponding to the second fabrication tool. The dispatch can be performed as described for the second action 220B shown in Figure 2A . In some embodiments, the AGV cleaner is dispatched by an AGV controller (e.g., the AGV controller 154 shown in Figure 1A ). In some embodiments, the AGV cleaner is dispatched to a location close to location 241 of peak contaminant levels. For example, in some embodiments, the AGV cleaner can be dispatched to a location within less than 10 meters, less than 5 meters, or less than 1 meter of location 241. In this way, the location to which the AGV cleaner is dispatched can be determined independently of the location of the second fabrication tool.

[0082] In operation 340, the first fabrication tool finishes processing the wafer. In some embodiments, the first fabrication tool processes the wafer while the AGV cleaner cleans the air near location 241 of peak contaminant levels. In some embodiments, the peak contaminant level is reduced by the action of the AGV cleaner during part or all of the processing of the wafer by the first fabrication tool. In some embodiments, the AGV cleaner reduces the peak contaminant level below the baseline threshold before the first fabrication tool finishes processing the wafer.

[0083] In operation 350, after the first fabrication tool finishes processing, the wafer is removed from the first fabrication tool. In some embodiments, the wafer is transferred from inside the first fabrication tool to a FOUP by a robotic arm, and the FOUP can be the same as or different from the FOUP that transports the wafer to the first fabrication tool.

[0084] In operation 360, in some embodiments, the AGV cleaner can be recalled when the peak contaminant level drops below a second threshold, which can be the baseline threshold. For example, the peak contaminant level may drop below the second threshold due to the cleaning performed by the AGV cleaner, due to stopping the second fabrication tool, or due to repairing a leak in the second fabrication tool.

[0085] Figure 3B FIG. shows a flowchart associated with process 31 for fabricating an IC device according to various embodiments. Process 31 can be associated with the process 20 described with reference toFigures 1A to 2H The described system 100 and process 20 are used in combination. Process 31 will be further described in accordance with one or more embodiments. It should be noted that the operations of process 31 can be rearranged or otherwise modified within the scope of various aspects. It should also be noted that additional processes can be provided before, during, and after process 31, and some other processes may only be briefly described herein.

[0086] In operation 301, the wafer is positioned in a fabrication tool (e.g., fabrication tool 222A1, Figure 1A , Figure 1B ) in a cleanroom (e.g., cleanroom 121, Figure 2B ). In some embodiments, positioning the wafer includes transporting the wafer in a FOUP to the port of the fabrication tool. The door of the FOUP can be opened, and a robotic arm can pick up the wafer and transfer it to the chamber of the fabrication tool. A robotic arm or a second robotic arm can position the wafer on the wafer stage in the fabrication tool for processing, such as deposition, etching, cleaning, annealing, or similar processes.

[0087] In operation 311, a peak contaminant level above a threshold in the cleanroom can be detected. In some embodiments, the peak contaminant level is detected by an AMC SCADA (e.g., AMC SCADA 151) based on measurements of an analysis system (e.g., analysis system 140 including TOF-MS 142). In some embodiments, the peak contaminant level is the peak contaminant level of the profile 240 at location 241. The threshold in operation 311 can correspond to the specification threshold described with reference to Figure 2A .

[0088] In operation 321, the fabrication tool is predicted to be the source of the contaminants at the peak contaminant level. The prediction of the fabrication tool can be performed using operations 200A to 200F of process 20 described with reference to Figures 2A to 2H .

[0089] In operation 331, the FOUP transfer to the fabrication tool is stopped, which can correspond to the fourth action 220D shown in Figure 2A . In some embodiments, the FOUP transfer is stopped by a SCADA system (e.g., a factory-side SCADA different from AMC SCADA 151).

[0090] In operation 341, the fabrication tool completes the processing of the wafer. In some embodiments, the processing of the wafer by the fabrication tool is performed while the FOUP transfer is stopped in operation 331.

[0091] In operation 351, after the processing by the fabrication tool is completed, the wafer is removed from the fabrication tool. In some embodiments, before stopping the FOUP-style transfer in operation 331, the wafer is transferred from inside the first fabrication tool to a FOUP, which may be a FOUP located at the fabrication tool.

[0092] In operation 361, the fabrication tool is repaired. In some embodiments, the fabrication tool is repaired after the processing of all in-progress wafers located in the fabrication tool is completed. In some embodiments, before the processing is completed, the wafers being processed are removed from the fabrication tool and temporarily stored (e.g., in a FOUP) until the repair of the fabrication tool is completed.

[0093] In operation 371, after the fabrication tool is repaired, the FOUP-style transfer to the fabrication tool is resumed.

[0094] Figure 4A 、 Figure 4B is a projected view of the fabrication tool 122 that causes peak contaminant levels according to various embodiments. Figure 4A is a block diagram of a system 3224, which may be a control system for respectively implementing operations 200C, 200D of process 20 and / or operations 320, 321 of processes 30, 31. The control system 3224 can predict the environmental quality / safety parameters of the cleanroom 121 and can implement cleanroom environment enhancement based on the predicted parameters. In some embodiments, the control system 3224 uses machine learning to predict which fabrication tool 122 is the main suspect causing contaminant leakage in the cleanroom 121.

[0095] In one embodiment, the control system 3224 includes an analysis model 3302 and a training module 3304. The training module 3304 uses machine learning techniques to train the analysis model 3302. In some embodiments, the machine learning techniques train the analysis model 3302 to select the fabrication tool 122. Although the training module 3304 is shown as separate from the analysis model 3302, in fact, the training module 3304 may be part of the analysis model 3302.

[0096] The control system 3224 includes or stores training set data 3306. The training set data 3306 includes simulated CFD data 3308, historical contaminant condition data 3310, and historical process result data 3318. The simulated CFD data 3308 includes data related to CFD of contaminants. The historical contaminant condition data 3310 includes data related to the environment in which contaminants are measured (e.g., environmental data associated with the clean room 121). The historical process result data 3318 includes data related to the wafer quality after the fabrication process performed by the fabrication tool 122 in the clean room 121. As will be described in more detail below, the training module 3304 uses the simulated CFD data 3308, the historical contaminant condition data 3310, and the historical process result data 3318 to train the analysis model 3302 using machine learning techniques.

[0097] In one embodiment, the simulated CFD data 3308 includes data related to the computational fluid dynamics of contaminants generated by the fabrication tool 122. For example, thousands or millions of calculations of the flow dynamics of contaminants can be generated by simulating the contaminant leakage of the fabrication tool 122 under the influence of the environmental conditions in the clean room 121. The simulated CFD data 3308 can include contamination distribution images 210A to 210N associated with the fabrication tool 122, which can be simulated individually for each contaminant, each environmental condition (e.g., environmental temperature, clean room layout, production schedule, chemicals / materials used, and the like), each process step, or other suitable parameters. Thus, the simulated CFD data 3308 can include CFD data for a large number of contaminants and operating conditions of the fabrication tool 122.

[0098] In one embodiment, the historical contaminant condition data 3310 includes various environmental conditions or parameters during the processing of wafers in the clean room 121. Thus, for each fabrication tool 122 that has data in the simulated CFD data 3308, the historical contaminant condition data 3310 can include the environmental conditions or parameters present during the processing of the wafers. For example, the historical contaminant condition data 3310 can include data related to temperature, potential of hydrogen (PH), humidity, light, accommodation time, vibration, electrostatic discharge (ESD), cleanliness, production schedule, and / or other suitable environmental condition parameters. The historical contaminant condition data 3310 also includes a distribution map 240 of each contaminant or TVOC measured by the system 100 during the processing of the wafers.

[0099] In one embodiment, the historical process result data 3318 includes various wafer quality parameters directly or indirectly generated by semiconductor manufacturing processes implemented by the fabrication tools 122 in the cleanroom 121. For example, the semiconductor manufacturing process may include one or more of a photoresist coating process, a planarization process, a cleaning process, a deposition process, an etching process, an annealing process, or other suitable manufacturing processes. In some embodiments, the historical process result data 3318 may include measurements of various wafer parameters. The measurements may include layer thickness, layer uniformity, roughness, cleanliness, or other suitable measurements. In some embodiments, the measurements include the results of electrical tests, wafer acceptance tests, optical tests, or other suitable tests, and the measurements may include pass / fail measurements, reliability measurements, data retention measurements, or similar measurements. In some embodiments, the historical process result data 3318 is associated with a plurality of previously processed semiconductor wafers. In some embodiments, the historical process result data 3318 is associated with individual semiconductor wafers, individual series of semiconductor wafers, and / or individual lots of semiconductor wafers.

[0100] In one embodiment, the training set data 3306 links the simulated CFD data 3308 and / or the historical contaminant condition data 3310 to the historical process result data 3318. In other words, the CFD calculations in the simulated CFD data 3308 and / or the environmental parameters in the environmental condition data 3310 are linked (e.g., by tagging) to the measurements in the historical process result data 3318. As will be described in more detail below, the tagged training set data can be used in a machine learning process to train the analysis model 3302 to generate the various predictions mentioned previously.

[0101] In one embodiment, the control system 3324 includes processing resources 3312, memory resources 3314, and communication resources 3316. The processing resources 3312 may include one or more controllers or processors. The processing resources 3312 are configured to execute software instructions, process data, make thin film etch control decisions, perform signal processing, read data from memory, write data to memory, and perform other processing operations. The processing resources 3312 may include physical processing resources 3312 and / or virtual processing resources 3312. The processing resources 3312 may include cloud-based processing resources, including processors and servers accessed via one or more cloud computing platforms.

[0102] In one embodiment, the memory resource 3314 may include one or more computer-readable memories. The memory resource 3314 is configured to store software instructions associated with the functions of the control system and its components, including but not limited to the analysis model 3302. The memory resource 3314 may store data associated with the functions of the control system 3224 and its components. The data may include training set data 3306, current process condition data, and any other data associated with the operation of the control system 3224 or any of its components. The memory resource 3314 may include physical memory resources and / or virtual memory resources. The memory resource 3314 may include cloud-based memory resources accessible via one or more cloud computing platforms. In some embodiments, the memory resource 3314 includes the database 152.

[0103] In one embodiment, the communication resource 3316 may include wired communication resources and wireless communication resources, which may facilitate communication via one or more networks such as a wired network, a wireless network, the Internet, or an intranet. The communication resource 3316 may enable the components of the control system 3224 to communicate with each other.

[0104] Figure 4B is a block diagram showing the Figure 4A operational and training aspects of the illustrated analysis model 3302 according to one embodiment. As previously mentioned, the training set data 3306 includes data related to a plurality of previously processed semiconductor wafers. Each previously processed semiconductor wafer was processed under specific environmental conditions and produced specific processing results. For example, the cleanroom environment sampling data, cleanroom raw material data sheets, cleanroom raw material lifetimes, cleanroom pipe layouts, production schedules, equipment by-product data sheets, equipment recipe schedules, equipment pipe aging data, factory layouts outside the cleanroom, and daily / weekly / monthly climate predictions are formatted into respective condition matrices 3352. The condition matrix 3352 includes a plurality of data vectors 3354. Each data vector 3354 corresponds to a specific parameter.

[0105] Figure 4B The illustrated example shows a single condition matrix 3352 that will be passed to the analysis model 3302 during the training process. In Figure 4B the illustrated example, the condition matrix 3352 includes nine data vectors 3354, each of which corresponds to a parameter of the semiconductor manufacturing process. For condition types that are not naturally represented numerically (such as raw material names), a number may be assigned to each possible material.

[0106] The analysis model 3302 includes multiple neural layers 3356a to 3356e. Each neural layer includes multiple nodes 3358. Each node 3358 can also be referred to as a neuron. Each node 3358 from the first neural layer 3356a receives the data value of each data field from the condition matrix 3352. Thus, in Figure 4B the illustrated example, each node 3358 from the first neural layer 3356a receives 36 data values because the condition matrix 3352 has 36 data scalars (9 * 4 = 36). Each neuron 3358 includes a corresponding internal mathematical function labeled F(x) in Figure 3B . Each node 3358 of the first neural layer 3356a generates a scalar value by applying the internal mathematical function F(x) to the data value of the data field 3354 from the condition matrix 3352. More details about the internal mathematical function F(x) are provided below.

[0107] Each node 3358 of the second neural layer 3356b receives the scalar value generated by each node 3358 of the first neural layer 3356a. Thus, in Figure 3B the illustrated example, each node of the second neural layer 3356b receives four scalar values because there are four nodes 3358 in the first neural layer 3356a. Each node 3358 of the second neural layer 3356b generates a scalar value by applying the corresponding internal mathematical function F(x) to the scalar value from the first neural layer 3356a.

[0108] Each node 3358 of the third neural layer 3356c receives the scalar value generated by each node 3358 of the second neural layer 3356b. Thus, in Figure 3B the illustrated example, each node of the third neural layer 3356c receives five scalar values because there are five nodes 3358 in the second neural layer 3356b. Each node 3358 of the third neural layer 3356c generates a scalar value by applying the corresponding internal mathematical function F(x) to the scalar value of the node 3358 from the second neural layer 3356b.

[0109] Each node 3358 of the neural layer 3356d receives the scalar value generated by each node 3358 of the previous neural layer (not shown). Each node 3358 of the neural layer 3356d generates a scalar value by applying the corresponding internal mathematical function F(x) to the scalar value of the node 3358 from the second neural layer 3356b.

[0110] The final neural layer includes only a single node 3358. The final neural layer receives scalar values produced by each node 3358 of the previous neural layer 3356d. The node 3358 of the final neural layer 3356e produces a data value 3368 by applying a mathematical function F(x) to the scalar values received from the nodes 3358 of the neural layer 3356d.

[0111] In Figure 4B the example shown, the data value 3368 corresponds to a predicted fabrication tool 122, and the predicted fabrication tool 122 corresponds to a value included in the condition matrix 3352. The predicted fabrication tool 122 may be represented as a confidence level (e.g., a percentage) in the data value 3368. In some embodiments, the final neural layer 3356e may produce data values corresponding to the various predictions described above. The final neural layer 3356e will include a corresponding node 3358 for each output data value to be produced.

[0112] During the machine learning process, the analysis model compares the predicted fabrication tool 122 in the data value 3368 with the actual fabrication tool 122 that is the source of the contaminant leak (as shown by the data value 3370). As previously described, for each set of historical environmental condition data, the training set data 3306 includes historical process result data indicating the characteristics of the semiconductor wafers produced by the fabrication process. Thus, the data field 3370 includes the actual contaminant levels present during the fabrication process as reflected in the condition matrix 3352. The analysis model 3302 compares the predicted fabrication tool 122 from the data value 3368 with the actual fabrication tool 122 from the data value 3370. The analysis model 3302 produces an error value 3372 that indicates the error or difference between the predicted fabrication tool 122 from the data value 3368 and the actual fabrication tool 122 from the data value 3370. The error value 3372 is used to train the analysis model 3302. In some embodiments, the error value 3372 is a difference in confidence levels (e.g., a percentage).

[0113] The training of the analysis model 3302 can be more comprehensively understood by discussing the internal mathematical function F(x). Although all nodes 3358 are labeled with the internal mathematical function F(x), the mathematical function F(x) for each node is unique. In one example, each internal mathematical function has the following form:

[0114] F(x) = x1*w1 + x2*w2 + … xn*w1 + b.

[0115] In the above equation, each value x1 to xn corresponds to a data value received from a node 3358 in the previous neural layer, or, in the case of the first neural layer 3356a, each value x1 to xn corresponds to a respective data value from a data field 3354 of the reflector condition matrix 3352. Thus, for a given node, n is equal to the number of nodes in the previous neural layer. The values w1 to wn are scalar weighting values associated with the corresponding nodes from the previous layer. The analysis model 3302 selects the values of the weighting values w1 to wn. The constant b is a scalar bias value and may also be multiplied by the weighting values. The value produced by the node 3358 is based on the weighting values w1 to wn. Thus, each node 3358 has n weighting values w1 to wn. Although not shown above, each function F(x) may also include an activation function. Multiply the sum given in the above equation by the activation function. Examples of activation functions may include a rectified linear unit (ReLU) function, a sigmoid function, a hyperbolic tangent function, or other types of activation functions.

[0116] After the error value 3372 has been calculated, the analysis model 3302 adjusts the weighting values w1 to wn of the respective nodes 3358 of the respective neural layers 3356A to 3356e. After the analysis model 3302 adjusts the weighting values w1 to wn, the analysis model 3302 again provides the reflector condition matrix 3352 to the input neural layer 3356a. Since the weighting values of the respective nodes 3358 of the analysis model 3302 are different, it is expected that the reflectivity 3368 will be different from the previous iteration. The analysis model 3302 again produces an error value 3372 by comparing the actual fabrication tool 3370 with the predicted fabrication tool 3368.

[0117] The analysis model 3302 again adjusts the weighting values w1 to wn associated with the respective nodes 3358. The analysis model 3302 again processes the condition matrix 3352 and produces a predicted expiration time 3368 and an associated error value 3372. The training process includes adjusting the weighting values w1 to wn in an iteration until the error value 3372 is minimized.

[0118] Figure 4BShows a single condition matrix 3352 passed to the analysis model 3302. In fact, the training process includes passing a large number of condition matrices 3352 through the analysis model 3302, generating a predicted fabrication tool 3368 for each condition matrix 3352, and generating an associated error value 3372 for each predicted fabrication tool. The training process may also include generating an aggregate error value indicating the average error of all predicted fabrication tools 122 for a batch of condition matrices 3352. The analysis model 3302 adjusts the weighting values w1 to wn after processing each batch of condition matrices 3352. The training process continues until the average error of all condition matrices 3352 is less than a selected threshold tolerance. When the average error is less than the selected threshold tolerance, the analysis model 3302 is trained, and the analysis model is trained to accurately predict the fabrication tool 122 as the source of a contaminant leak based on environmental conditions and / or process conditions. Then, the analysis model 3302 can be used to predict a leak and select environmental and / or process conditions that will cause a reduction in the leak or an early containment of the leak (e.g., by dispatching and activating the AGV 123). During the use of the trained model 3302, an environmental condition vector or matrix is provided to the trained analysis model 3302, the environmental condition vector or matrix represents the current environmental conditions of the clean room 121 and / or the fabrication tool 122, and has a format similar to the condition matrix 3352. The trained analysis model 3302 can then predict the leak that will be caused by those environmental conditions.

[0119] Reference has been made to Figure 4B A specific example of the neural network-based analysis model 3302 has been described. However, other types of neural network-based analysis models, or analysis models other than neural networks, may be utilized without departing from the scope of the present disclosure. Additionally, without departing from the scope of the present disclosure, the neural network may have a different number of neural layers, and the different number of neural layers may have a different number of nodes.

[0120] Embodiments can provide advantages. By utilizing the local TOF-MS 142, a faster response time (within 1 minute) is possible, enabling earlier and faster response when an electronic fluorinated liquid (Fluorinert Electronic Liquids, CxF), IPA / acetone, and / or TVOC leak event manifest is anticipated or detected. The impact of AMC on manufacturing can be minimized. Monitoring of the AMC situation is carried out in a near-real-time manner, including the global cleanroom environment and the fabrication tools or attachments in each process section, where areas deviating from the baseline, OOC, or OOS are highlighted. Alarms can be sent to manufacturing operators and plant operators for further actions such as maintenance or repair. The AI-based graphical recognition program compares the measurement data with the image database, and advanced AMC data comparison validates the selection of the fabrication tool identified as the source of the contaminant leak. The AI-based management program automatically determines and dispatches the AMC cleaner with a fan and filter towards the AMC event area for disposal.

[0121] According to at least one embodiment, a method for airborne contaminant management includes: generating a contaminant distribution map by sampling the environment of a cleanroom; selecting a first fabrication tool of the cleanroom by comparing the contaminant distribution map with at least one diffusion image in a first database; comparing parameters of the first fabrication tool with process utility information in a second database; and taking at least one measure when the parameters are consistent with the process utility information. The at least one measure may include: moving a cleaning tool to a position associated with the contaminant concentration of the contaminant distribution map; turning on a fan of the cleaning tool; stopping cassette conveyance to the first fabrication tool; and terminating generation through the first fabrication tool.

[0122] In a related embodiment, selecting the first fabrication tool includes: selecting a first diffusion image from the at least one diffusion image; and selecting the first fabrication tool when the confidence level of the first fabrication tool is the highest among all fabrication tools associated with the first diffusion image.

[0123] In a related embodiment, the confidence level is determined based on at least one computational fluid dynamics simulation.

[0124] In a related embodiment, the confidence level is predicted by a trained machine learning analysis model.

[0125] In related embodiments, when the peak concentration level of the contaminant distribution map is higher than a first threshold, moving the cleaning tool to the location is implemented; and when the peak concentration level is higher than a second threshold, turning on the fan is implemented, the second threshold being higher than the first threshold.

[0126] In related embodiments, when the peak concentration level is higher than a third threshold, stopping the pod conveyance and terminating the generation is implemented, the third threshold being higher than the second threshold.

[0127] In related embodiments, selecting the first fabrication tool includes selecting a tool attachment located beneath the raised floor of the cleanroom.

[0128] According to at least one embodiment, a method for airborne contaminant management includes: generating cleanroom contaminant data by sampling cleanroom contaminants using a sampling system; generating a first image based on the cleanroom contaminant data by a time-of-flight mass spectrometer (TOF-MS); selecting a first fabrication tool based on a prediction made using the first image and at least one other cleanroom diffusion image; and reducing the contaminant concentration near the first fabrication tool by an automatic guided vehicle (AGV) dispatched by an AGV controller.

[0129] In related embodiments, sampling the cleanroom contaminants includes sampling at least one of chlorofluorocarbons, hydrofluorocarbons, perfluorocarbons, isopropyl alcohol, acetone, and total volatile organic compounds.

[0130] In related embodiments, sampling the cleanroom contaminants includes sampling by the sampling system, the sampling system including a sampling unit having an areal density in the range of 1 per square meter to 50 per square meter.

[0131] In related embodiments, the areal density of the sampling unit is higher in a first section of the cleanroom than in a second section of the cleanroom.

[0132] In related embodiments, the first section includes an etching apparatus or an electroless copper plating apparatus.

[0133] In related embodiments, the method further includes verifying the first fabrication tool by confirming any of: an acid or solvent supply delivery from the first fabrication tool; an operating state of the first fabrication tool; utility information of the first fabrication tool; or chemical usage information of the first fabrication tool.

[0134] According to at least one embodiment, a method for airborne contaminant management includes: positioning a wafer in a fabrication tool; detecting a peak concentration level of a contaminant above a first threshold; predicting that the fabrication tool is a source of the contaminant; stopping the conveyance of additional wafers to the fabrication tool; completing the processing of the wafer through the fabrication tool; removing the wafer from the fabrication tool; reducing the peak concentration level by repairing the fabrication tool; and resuming the conveyance of the additional wafers to the fabrication tool when the fabrication tool is repaired and the peak concentration level is below a baseline threshold, the baseline threshold being lower than the first threshold.

[0135] In a related embodiment, the method further includes dispatching a cleaning tool to the fabrication tool when the peak concentration level exceeds a second threshold between the first threshold and the baseline threshold.

[0136] In a related embodiment, the method further includes filtering the air near the fabrication tool when the peak concentration level exceeds a third threshold between the second threshold and the first threshold.

[0137] In a related embodiment, the method further includes recalling the cleaning tool when the peak concentration level exceeds the first threshold.

[0138] In a related embodiment, filtering the air includes turning on a fan of the cleaning tool that is in fluid communication with a filter of the cleaning tool.

[0139] In a related embodiment, filtering the air includes orienting a filtration system towards a location associated with the peak concentration level by a drive system of the cleaning tool.

[0140] In a related embodiment, dispatching the cleaning tool includes wirelessly transmitting a dispatch command from an automated guided vehicle controller to an automated guided vehicle having a filtration system.

[0141] The foregoing outlines features of several embodiments so that those skilled in the art may better understand aspects of the present disclosure. Those skilled in the art should appreciate that they may readily use the present disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and / or achieving the same advantages as the embodiments introduced herein. Those skilled in the art should also realize that such equivalent constructs do not depart from the spirit and scope of the present disclosure, and that they may make various changes, substitutions, and alterations herein without departing from the spirit and scope of the present disclosure.

Claims

1. A method for managing airborne pollutants, comprising: By sampling the environment of a clean room, a contaminant distribution map is generated; By comparing the contaminant distribution map with at least one diffusion image in a first database, a first fabrication tool of the clean room is selected, wherein the at least one diffusion image is an airborne molecular contamination diffusion image generated by computational fluid dynamics techniques assuming that the first fabrication tool is the source of contaminant leakage; Compare the parameters of the first fabrication tool with the process utility information in a second database; And When the parameters are consistent with the process utility information, at least one of the following measures is taken: Move a cleaning tool to a position associated with the contaminant concentration of the contaminant distribution map; Turn on the fan of the cleaning tool; Stop the cassette conveyance to the first fabrication tool; And Terminate the generation through the first fabrication tool.

2. The method according to claim 1, wherein selecting the first production tool comprises: Select a first diffusion image from the at least one diffusion image; And When the confidence level of the first fabrication tool is the highest confidence level among all fabrication tools associated with the first diffusion image, select the first fabrication tool.

3. The method according to claim 2, wherein the confidence level is determined based on at least one computational fluid dynamics simulation.

4. The method according to claim 3, wherein the confidence level is predicted by a trained machine learning analysis model.

5. The method according to claim 1, wherein: When the peak concentration level of the contaminant distribution map is higher than a first threshold, move the cleaning tool to the position; and When the peak concentration level is higher than a second threshold, turn on the fan, the second threshold being higher than the first threshold.

6. The method according to claim 5, wherein: When the peak concentration level is higher than a third threshold, stop the cassette conveyance and terminate the generation, the third threshold being higher than the second threshold.

7. The method according to claim 1, wherein selecting the first production tool comprises: Select tool accessories located under the raised floor of the clean room.

8. A method for managing airborne pollutants, comprising: By using a sampling system to sample clean room contaminants, clean room contaminant data is generated; Based on the clean room contaminant data, a first image is generated by a time-of-flight mass spectrometer; Based on predictions using the first image and at least one other clean room diffusion image, a first fabrication tool is selected, wherein the at least one other clean room diffusion image is an airborne molecular contamination diffusion image generated by computational fluid dynamics techniques assuming that the first fabrication tool is the source of contaminant leakage; And By an automated guided vehicle dispatched by an automated guided vehicle controller, reduce the contaminant concentration near the first fabrication tool.

9. The method according to claim 8, wherein sampling the cleanroom pollutants comprises sampling at least one of chlorofluorocarbons, hydrofluorocarbons, perfluorocarbons, isopropanol, acetone, and total volatile organic compounds.

10. The method according to claim 8, wherein sampling the cleanroom contaminants includes sampling through the sampling system, the sampling system including a sampling unit having a areal density in the range of from 1 per square meter to 50 per square meter.

11. The method according to claim 10, wherein the areal density of the sampling unit is higher in a first section of the cleanroom than in a second section of the cleanroom.

12. The method according to claim 11, wherein the first section includes an etching apparatus or an electroless copper plating apparatus.

13. The method according to claim 8, wherein the method further includes validating the first fabrication tool by confirming: an acid or solvent supply transmission from the first fabrication tool; an operating state of the first fabrication tool; utility information of the first fabrication tool; or chemical usage information of the first fabrication tool.

14. A method for airborne contaminant management, comprising: Position a wafer in a fabrication tool; Detect a peak concentration level of contaminants higher than a first threshold; By computational fluid dynamics techniques, assuming that the fabrication tool is the source of the contaminant leakage, generate at least one diffusion image; Compare the position of the peak concentration level with the position of the peak concentration level in the at least one diffusion image, thereby predicting that the fabrication tool is the source of the contaminant; Stop conveying additional wafers to the fabrication tool; Through the fabrication tool, complete the processing of the wafer; Remove the wafer from the fabrication tool; By repairing the fabrication tool, reduce the peak concentration level; And When the fabrication tool is repaired and the peak concentration level is lower than a baseline threshold, resume conveying the additional wafers to the fabrication tool, the baseline threshold being lower than the first threshold.

15. The method according to claim 14, further comprising dispatching a cleaning tool to the fabrication tool when the peak concentration level exceeds a second threshold between the first threshold and the baseline threshold.

16. The method according to claim 15, further comprising filtering air near the fabrication tool when the peak concentration level exceeds a third threshold between the second threshold and the first threshold.

17. The method according to claim 16, wherein the method further comprises: When the peak concentration level exceeds the first threshold, recall the cleaning tool.

18. The method according to claim 16, wherein filtering the air includes: Turn on a fan of the cleaning tool that is in fluid communication with a filter of the cleaning tool.

19. The method according to claim 18, wherein filtering the air comprises: Orient a filtration system towards a location associated with the peak concentration level via a drive system of the cleaning tool.

20. The method according to claim 15, wherein dispatching the cleaning tool comprises: Transmit a dispatch command wirelessly from an automated guided vehicle controller to an automated guided vehicle having a filtration system.

Images