Multi-substrate inspection systems and techniques for semiconductor device inspection

Abstract

A system includes a multi-substrate platform having a plurality of substrate holders, each substrate holder holding a respective substrate of a plurality of substrates. The system further includes an inspection system having a plurality of inspection heads to perform, concurrently, an inspection of the plurality of substrates, each inspection head collecting inspection data for the respective substrate of the plurality of substrates. The system further includes a processing device to identify, using the collected inspection data, at least one of (i.) presence of one or more defects in at least one substrate of the plurality of substrates or (ii.) one or more measurements of one or more properties of at least one substrate of the plurality of substrates.

Description

Attorney Docket No. :39361.320 (L0150PCT)MULTI-SUBSTRATE INSPECTION SYSTEMS AND TECHNIQUES FOR SEMICONDUCTOR DEVICE INSPECTIONTECHNICAL FIELD

[0001] This instant specification generally relates to ensuring quality control of materials manufactured in substrate processing systems. More specifically, the instant specification relates to systems and techniques of efficient inspections of semiconductor manufacturing products during various stages of the manufacturing process, including initial, intermediate, and / or final manufacturing of products.BACKGROUND

[0002] Manufacturing of modern materials often involves various deposition techniques, such as chemical vapor deposition (CVD) or physical vapor deposition (PVD), in which atoms of one or more selected types are deposited on a substrate (wafer) held in low or high vacuum environments that are provided by vacuum deposition chambers. Manufacturing further includes various other techniques, such as etching, patterning (e.g., photolithography), polishing, cleaning, stress mitigation, and / or the like. Materials manufactured in this manner include monocrystals, semiconductor structures, fine coatings, and numerous other substances used in practical applications, e.g., electronic device manufacturing.

[0003] During various stages of manufacturing of electronic devices (e.g., semiconductor devices) on substrates, metrology or inspection tools may inspect the substrates for defects. Inspection of substrates can be time consuming, and increases the cost of production of electronic devices.SUMMARY

[0004] In one embodiment, a system includes an inspection system (e.g., one or more inspection tools (also referred to a metrology devices and / or inspection devices)) that is capable of performing inspection on multiple substrates in parallel. In some embodiments, the system includes a multi-substrate platform including a plurality of substrate holders. Each substrate holder of the plurality of substrate holders may hold a respective substrate of a plurality of substrates. The inspection system may include a plurality of inspection heads to perform, concurrently, an inspection of two or more of the plurality of substrates. Each inspection head of the plurality of inspection heads is configured to collect data for a respective substrate of the plurality of substrates. The system further includes a processing device to identify, using the collected data, at least one of (i.) presence of one or more defectsAttorney Docket No. :39361.320 (L0150PCT) in at least one substrate of the plurality of substrates or (ii.) one or more measurements of one or more properties of at least one substrate of the plurality of substrates.

[0005] In one embodiment, the system includes a platform including a plurality of chucks. Each chuck is configured to hold a substrate of a plurality of substrates. The system further includes one or more aligner devices to identify reference features of each of the plurality of substrates. The system further includes a robot including a plurality of arms. Each arm is configured to place a respective substrate of the plurality of substrates on a respective chuck of the plurality of chucks. The system further includes a plurality of inspection heads. An individual inspection head is configured to perform an inspection of a corresponding substrate of the plurality of substrates. The inspections of two or more of the plurality of substrates are performed concurrently in embodiments (e.g., each by a different inspection head).

[0006] In one embodiment, a method includes supporting, using a multi-substrate platform, a plurality of substrates. Each substrate of the plurality of substrates is supported by a respective substrate holder of a plurality of substrate holders. The method further includes collecting, in parallel, inspection data for the plurality of substrates. The inspection data for each substrate of the plurality of substrates is collected using a respective inspection head of a plurality of inspection heads. The method further includes identifying, by a processing device and using the collected inspection data for the plurality of substrates, at least one of (i.) presence of one or more defects in at least one substrate of the plurality of substrates or (ii.) one or more measurements of one or more properties of at least one substrate of the plurality of substrates.

[0007] In one embodiment, a substrate inspection system includes a chuck configured to secure multiple substrates for inspection. The system further includes an inspection tool. The inspection tool includes multiple imaging devices configured to capture image data of the multiple substrates simultaneously. The system further includes a processing device. The processing device is configured to receive the image data from the multiple imaging devices and to determine one or more defects of the multiple substrates based on the image data.

[0008] In one embodiment, a substrate inspection tool includes a support structure. The substrate inspection tool further includes multiple image sensors supported by the support structure above an inspection zone. Each of the multiple image sensors are configured to simultaneously (or in parallel) capture image data of a respective substrate. The tool further includes a processing device configured to determine one or more defects of one or more of the respective substrates based on corresponding image data.Attorney Docket No. :39361.320 (L0150PCT)

[0009] In one embodiment, a method includes receiving multiple substrates on a chuck. The method further includes capturing, by multiple imaging devices, image data of the multiple substrates simultaneously or in parallel. The method further includes determining, by a processing device, one or more defects of the multiple substrates based on the image data.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Aspects and embodiments of the present disclosure will be understood more fully from the detailed description given below and from the accompanying drawings of various aspects and embodiments of the disclosure, which, however, should not be taken to limit the disclosure to the specific aspects or embodiments, but are for explanation and understanding only. The drawings, described below, are for illustrative purposes and are not necessarily drawn to scale.

[0011] FIG. 1 illustrates an example system capable of performing parallel multisubstrate inspections, in accordance with some embodiments of the present disclosure.

[0012] FIG. 2A illustrates schematically example operations that facilitate performance of parallel multi-substrate inspections by the system of FIG. 1, in accordance with some embodiments of the present disclosure.

[0013] FIG. 2B illustrates a spiral pattern of parallel inspection of multiple substrates using multiple inspection heads, in accordance with some embodiments of the present disclosure.

[0014] FIG. 2C illustrates inspection of a large-size rectangular substrate (“panel”) having a substantially rectangular form and supported by a single chuck, in accordance with some embodiments of the present disclosure.

[0015] FIG. 2D illustrates another example architecture of a multi-substrate inspection system deploying multiple inspection heads per substrate, in accordance with some embodiments of the present disclosure.

[0016] FIG. 3 illustrates an example multi-head optical inspection system that can be used for parallel multi-substrate inspections, operating in accordance with some embodiments of the present disclosure.

[0017] FIG. 4 is a flow diagram of an example method of parallel inspection of multiple substrates using a multi-head inspection system, in accordance with some embodiments of the present disclosure.Attorney Docket No. :39361.320 (L0150PCT)

[0018] FIG. 5 is a block diagram illustrating an exemplary system architecture, in accordance with some embodiments of the present disclosure.

[0019] FIG. 6 is a block diagram illustrating a system for generating output data, in accordance with some embodiments of the present disclosure.

[0020] FIG. 7A-7B are simplified process flow diagrams for inspecting substrates using an example multi-substrate inspection system, in accordance with some embodiments of the present disclosure.

[0021] FIGS. 7C-7E are flow diagrams of methods for training and using a machine learning model to select a wavelength of light for inspecting substrates, in accordance with some embodiments of the present disclosure.

[0022] FIG. 8 is a flow diagram of an example method for inspecting dies placed on substrates, in accordance with some embodiments of the present disclosure.

[0023] FIG. 9 is a simplified block diagram illustrating a computer system, in accordance with some embodiments of the present disclosure.DETAILED DESCRIPTION

[0024] Semiconductor device manufacturing and other device manufacturing (e.g., such as for displays, photovoltaic devices, etc.) often involves tens and even hundreds of complex operations to implement raw substrate (e.g., wafer) preparation, polishing, material deposition, etching, and the like. Substrates that are delivered for processing in manufacturing chambers can include bare substrates (e.g., silicon substrates, quartz substrates, Gallium Arsenide substrates, corundum substrates), substrates that have been preprocessed (e.g., covered with one or more films, such as carbon films), or substrates that have already undergone one or more processing operations (e.g., deposition, patterning, etching, and so on). Operations with substrates (including bare wafer manufacturing) and transportation of substrates can leave or cause various defects and impurities in wafers, including but not limited to chipping near substrate edges, pitting, staining, film peeling, particles, pattern defects, (e.g., such as bridges and protrusions) and / or various other substrate imperfections. Undiscovered defects can result in expensive wasteful processing, sub-optimal and unusable manufacturing products, and damage processing tools.

[0025] Since even a small number of impurities or other defects introduced into processing environments can render the manufacturing products (masks, wafers, chips, etc.) unusable fortheir intended purposes, various manufacturing operations are often interspersed with quality control (QC) evaluations to verify adherence of intermediate and final productsAttorney Docket No. :39361.320 (L0150PCT) to specifications of the technological processes being performed. QC evaluations can include metrology evaluations (also customarily referred to as simply “metrology”) that measure a specific sample property, such as dimensions or any physical / mechanical / chemical material property, e.g., composition, using optical probes, electron probes, ion probes, and / or the like. Metrology evaluations can include measuring dimensions such as substrate thickness, roughness, and / or any other measurable property of the substrate. QC evaluations can also include defect inspections that determine the presence of particle and pattern defects on the substrates. Defect inspections can, for example, deploy optical systems that illuminate substrates with light (which can include visible light, infrared light, ultraviolet light, etc.) while detecting light reflected and / or scattered back from the sample. Defect inspections can deploy various defect detection algorithms, such as image processing-based heuristic algorithms, machine learning techniques (e.g., decision-tree algorithms, deep learning neural networks, etc.), and / or the like. Such inspections, however, may introduce an additional step into the manufacturing process, increase the total processing time, and adversely affect the manufacturing throughput.

[0026] For example, inspection of a single substrate can take minutes to tens of minutes, depending on the nature of the characterization process. Additionally, overhead may be introduced due to substrate handling. Substrate handling may include using a robot to move the substrate from a storage or temporary location (e.g., a front opening unified pod (FOUP) that holds bare or partially processed substrates) or from a process (deposition, etching, lithography, etc.) chamber, placing the substrate on a chuck (e.g., electrostatic chuck), moving the chuck to an inspection station for inspection of the substrate (e.g., metrology, defect inspection, and / or the like), retrieving the chuck with the substrate after completion of the inspection, removing the substrate from the chuck and moving the substrate to a storage destination or to the next processing operation. As a result, a few up to 10s of substrates can be handled and analyzed within one hour. However, it is advantageous to increase the number of substrates that are inspected per hour.

[0027] Aspects and embodiments of the present disclosure improve on the throughput of substrate inspections by enabling systems and techniques for concurrent multi-substrate inspections. In some embodiments, two or more substrates can be inspected at the same time, e.g., in parallel or simultaneously. For example, a robot can fetch four substrates from a FOUP at the same time and place the substrates on a multi-substrate platform having multiple substrate-holding chucks. The platform canbe moved to a multi-substrate inspection station where each substrate can be inspected — in parallel with other substrates — by one or moreAttorney Docket No. :39361.320 (L0150PCT) inspection devices, e.g., inspection heads. Subsequent to the inspection, the four substrates can be retrieved from the inspection station and delivered to their next destination, such as a process chamber, a FOUP, and / or the like. As a result, the throughput of the substrate inspection can be increased by more than four times. In addition to the increased throughput, the advantages of the disclosed embodiments include, but are not limited to, reduced inspection costs per substrate, and / or the like.

[0028] FIG. 1 illustrates an example system 200 capable of performing parallel multisubstrate inspections, in accordance with some embodiments of the present disclosure. In some embodiments, example system 200 can be a free-standing substrate inspection system.

[0029] The system 200 includes a factory interface (FI) 101 and load ports 128x (e.g., load ports 128A-D). In some embodiments, the load ports 128A-D are directly mounted to (e.g., sealed against) FI 101. Enclosure systems 130x (e.g., cassette, FOUP, process kit enclosure system, or the like) are configured to removably couple (e.g., dock) to the load ports 128A-D. In some embodiments, enclosure system 130A is coupled to load port 128A, enclosure system 130B is coupled to load port 128B, enclosure system 130C is coupled to load port 128C, and enclosure system 130D is coupled to load port 128D. In some embodiments, one or more enclosure systems 13 Ox are coupled to the load ports 128x for transferring substrates and / or other items into and out of the system 200. Each of the enclosure systems 130xmay seal against a respective load port 128x. In some embodiments, a first enclosure system 130A is docked to a load port 128A. Once such operation or operations are performed, the first enclosure system 130A is undocked from the load port 128A, and then a second enclosure system 130x (e.g., a FOUP containing substrate(s)) is docked to the same load port 128A. In some embodiments, an enclosure system 13 Ox (e.g., enclosure system 130A) is a system for performing a calibration operation or a diagnostic operation.

[0030] In some embodiments, a load port 128x includes a front interface that forms an opening. The load port 128x additionally includes a horizontal surface for supporting an enclosure system 13 Ox. Each enclosure system 13 Ox has a front interface that forms a vertical opening. The front interface of the enclosure system 130x is sized to interface with (e.g., seal to) the front interface of the load port 128x (e.g., the vertical opening of the enclosure system 130x is approximately the same size as the vertical opening of the load port 128x). The enclosure system 130x is placed on the horizontal surface of the load port 128x and the vertical opening of the enclosure system 13 Ox aligns with the vertical opening of the load port 128x. The front interface of the enclosure system 130x interconnects with (e.g., clampAttorney Docket No. :39361.320 (L0150PCT) to, be secured to, be sealed to) the front interface of the load port 128x. A bottom plate (e.g., base plate) of the enclosure system 130xhas features (e.g., load features, such as recesses or receptacles, that engage with load port kinematic pin features, a load port feature for pin clearance, and / or an enclosure system docking tray latch clamping feature) that engage with the horizontal surface of the load port 128x. The same load ports 128x that are used for different types of enclosure systems 130x.

[0031] The system 200 includes chambers, such as FI 101 (e.g., equipment front end module, EFEM) and adjacent chambers (e.g., load port 128x, enclosure system 130x, or the like) that are adjacent to FI 101. Some or all of the chambers can be sealed. In some embodiments, inert gas (e.g., one or more of nitrogen, argon, neon, helium, krypton, or xenon) is provided into one or more of the chambers (e.g., FI 101 and / or adjacent chambers) to provide one or more inert environments. In some examples, FI 101 is an inert EFEM that maintains the inert environment (e.g., inert EFEM minienvironment) within FI 101 so that users do not need to enter FI 101 (e.g., the system 200 is configured for no manual access within FI 101).

[0032] In some embodiments, gas flow (e.g., inert gas, nitrogen) is provided into one or more chambers (e.g., FI 101) of the system 200. In some embodiments, the gas flow is greater than leakage through the one or more chambers to maintain a positive pressure within the one or more chambers. In some embodiments, the inert gas within FI 101 is recirculated. In some embodiments, a portion of the inert gas is exhausted. In some embodiments, the gas flow of non-recirculated gas into FI 101 is greater than the exhausted gas flow and the gas leakage to maintain a positive pressure of inert gas within FI 101. In some embodiments, FI 101 is coupled to one or more valves and / or pumps to provide the gas flow into and out of FI 101. A processing device (e.g., of controller 109) controls the gas flow into and out of FI 101. In some embodiments, the processing device receives sensor data from one or more sensors (e.g., oxygen sensor, moisture sensor, motion sensor, door actuation sensor, temperature sensor, pressure sensor, etc.) and determines, based on the sensor data, the flow rate of inert gas flowing into and / or out of FI 101 .

[0033] The enclosure system 130x seals to the load port 128xresponsiveto being docked on the load port 128x. The enclosure system 130x provides purge port access so that the interior of the enclosure system 130x can be purged prior to opening the enclosure system 13 Ox to minimize disturbance of the inert environment within FI 101.

[0034] System 200 can include an aligner station 150 to determine orientation of content (e.g., substrates). For example, orientations of substrates inside enclosure systems 130A-DAttorney Docket No. :39361.320 (L0150PCT) are typically not controlled to a sufficient degree that would enable FI robot 111 to pick up an automatically aligned substrate in a way that would enable immediate processing or inspection. Orientation of crystallographic axes (and / or directionality of various features that can be patterned on the wafer) of a substrate fetched from a FOUP can thus be arbitrary. Aligner station 150 is deployed to align substrates relative to some reference direction, e.g., a specific direction associated with FI robot 111 . In some embodiments, aligner station 150 can spin the substrate and locate, e.g., using various techniques of machine vision, a reference feature on the substrate that communicates to FI robot 111 (and / or other wafer manufacturing tools) orientation of the substrate. Such reference features can include a notch that is cut into an edge of the substrate, a flat (cut-out) portion of the substrate’s edge, or any other reference feature thatbreaks the circular symmetry of the substrate and is detectable by mechanical or optical techniques. Aligner station 150 typically locates such reference features over one-to- several seconds of the substrate’s spinning, e.g., at about 50-200 rpm.

[0035] After aligner station 150 determines the substrate’s orientation, FI robot 111 can place the substrate together with one or more other substrates on a multi-substrate platform 230 that is transported to an inspection area 240 of multi-substrate inspection station 160. Having detected the arrival of the multi-substrate platform, multi-substrate inspection station 160 can perform parallel inspection of multiple substrates supported by the multi- chuck platform 230 to collect inspection data. For example, multi-substrate inspection station 160 may perform parallel inspection of multiple substrates supported by the multi-chuck platform 230 to collect metrology data and / or defect data. A data processing server can process the collected inspection data and determine one or more characteristics of the substrate(s). In some embodiments, inspection data includes data indicative of one or more defects and / or data indicative of one or more measurable properties such as substrate thickness, substrate material composition, and / or substrate surface roughness, etc. The data processing server may identify the presence of one or more defects of a substrate and / or identify one or more measurements of one or more properties of a substrate based on the inspection data.

[0036] FI 101 can deploy a system of one or more FI robots 11 lx capable of handling multiple substrates 220A-D in parallel. For example, as shown in FIG. 1, the system of FI robots can include FI robot 111 A capable of handling substrates 220A-B and FI robot 11 IB capable of handling substrates 220C-D. In other embodiments, a single FI robot can be used to handle more than two substrates (e.g., all four substrates 220A-D) at the same time. In yet other embodiments, each substrate 220x can be handled by a separate FI robot 11 lx.Attorney Docket No. :39361.320 (L0150PCT)

[0037] Example system 200 can also include one or more buffer stations 21 Ox. Four buffer stations 21 OA-D are shown in FIG. 1 but any other number of buffer stations 21 Ox can be deployed. Buffer stations 210 A-D can be used to reduce substrate handling time and ensure more efficient use of the FI robots. Individual buffer stations 210A-D can include multiple load slots and / or unload slots, which can be used to hold individual substrates before and / or after an inspection operation is performed. Buffer stations 210A-D can be positioned a short distance from FI 101 and multi-substrate inspection station 160. During inspection of a current set of substrates by multi-substrate inspection station 160, FI robots 111 A-B can deliver the next set of uninspected substrates from enclosure systems 130A-D and place the uninspected substrates into load slots of the corresponding buffer stations 210A-D. After the inspection of the current set of substrates is completed, FI robots 111 A-B can fetch the inspected substrates from multi-substrate inspection station 160 and place the current set of substrates into unload slots of buffer stations 210A-D, fetch the next set of uninspected substrates from the load slots and deliver these substrates to multi-substrate inspection station 160. Once the inspection of the delivered substrates begins, FI robots 111 A-B can return to buffer stations 210A-D to pick up the most recently inspected substrates from the load ports and carry the inspected substrates to the respective enclosure systems 130A-D. FI robots 111 A-B can then acquire a new set of the uninspected substrates from enclosure systems 130A-D and take these substrates to buffer stations 210A-D to place the new set of uninspected substrates into load slots of buffer stations 210A-D while awaiting the completion of the ongoing inspection operations. The number of slots in each buffer station can be different in various embodiments. In some embodiments, a buffer station can have one load slot and one unload slot. In some embodiments, a buffer station can have multiple load slots and / or multiple unload slots.

[0038] In some embodiments, prior to delivering new sets of uninspected substrates from enclosure systems 130A-D to load slots of buffer stations 210A-D, FI robots 111 A-B can bring the substrates to (one or more) aligner station(s) 150. Aligner station(s) 150 can use optical and / or mechanical techniques to identify positions of reference features, such as a notch, a flat cut-out region, and / or the like, on the substrates, e.g., by imparting rotational motion to the substrates until the reference features are located. The substrate’s rotational motion can then be stopped with a substrate held by a robot arm with the reference feature aligned in a predetermined fashion relative to the robot arm (e.g., a matching feature on the robot arm). The substrate alignment operation performed by aligner station(s) 150 is typically fast, taking substantially less time (e.g., 2-3 seconds) than the inspection. Accordingly, aAttorney Docket No. :39361.320 (L0150PCT) single aligner station 150 can align multiple substrates sequentially and the FI robots 11 lx can deliver multiple (e.g. all) aligned substrates to the intended buffer stations 21 OA-D while a previous set of buffers is being inspected. In some embodiments, multiple aligner stations 150 can be deployed to perform alignment of at least some of the substrates in parallel. For example, if two aligner stations are deployed, a first aligner station can align substrate 220A in parallel to the second aligner station aligning substrate 220C. FI robot 111 A can then deliver substrate 220 A to buffer station 210 A while FI robot 11 IB can deliver substrate 220D to buffer station 210D. Subsequently, the first aligner station can align substrate 220B in parallel to the second aligner station aligning substrate 220C. Next, FI robots 111A and 11 IB can deliver, respectively, the aligned substrates 220B and 220C to buffer stations 21 OB and 210C.

[0039] In some embodiments, when multi-substrate inspection station 160 is ready to accept substrates 220A-D (e.g., when inspection of the previous set of substrates has finished and the substrates have been removed from multi-substrate inspection station 160), FI robots 111 A and 11 IB can fetch the aligned substrates 220A-D from buffer stations 210A-D and place the substrates on a multi-substrate platform 230. The fetching and placement of the substrates can be performed in parallel (e.g., concurrently), where each substrate is fetched by a separate arm of one of the FI robots.

[0040] Multi-substrate platform 230 can support multiple individual chucks capable of accepting and supporting individual substrates 220 A-D. The individual chucks can include one or more electrostatic chucks that hold substrates using forces of electrostatic attraction, one or more vacuum chucks (e.g., vacuum suction chucks, etc.) that hold substrates using vacuum-induced suction forces, and / or chucks deploying any other suitable holding mechanisms. In some embodiments, forces applied by individual chucks can be sufficiently strong to (at least partially) flatten the substrates (e.g., reduce a bow or saddle deformation of the substrates) while the substrates are positioned on the chucks. In some embodiments, the system includes a substrate flattening unit configured to apply a downward force on a top surface of the substrates to flatten the substrates on the chuck. In some embodiments, multisubstrate platform 230 can support substrates of multiple sizes. For example, multi-substrate platform 230 can support a large rectangular (or square) “panel” substrate.

[0041] In some embodiments, multi-substrate platform 230, with substrates 220A-D positioned and secured thereon, may be immersed in a liquid medium, such as water, etc. Immersion of the substrates 220A-D may be for performance of acoustic substrate inspection where a transducer (e.g., a piezoelectric transducer) generates acoustic waves for inspectingAttorney Docket No. :39361.320 (L0150PCT) the substrates (e.g., instead of an optical head, etc.). The liquid medium may be used as a coupling medium in which a tested substrate is immersed.

[0042] Multi-substrate platform 230, with substrates 220A-D positioned and secured thereon, can be transported into an inspection area 240 where multiple inspection heads, e.g., inspection heads 250 A-D, are located. A support structure may be configured to support the inspection heads 250 A-D. Transportation of multi-substrate platform 230 can be facilitated by any suitable transportation mechanism, e.g., a rail 232. In some embodiments, one or more actuators are to move the platform 230 between a loading position and an inspection position (e.g., an inspection position corresponding to inspection area 240). When the platform 230 is in the loading position, multiple substrates can be loaded onto the platform 230. When the platform 230 is in the inspection position, an inspection tool is operatable to inspect the substrates.

[0043] In some embodiments, the inspection heads 250A-D can perform multiple functions. Each of the inspection heads 250A-D may be capable of searching for and / or locating a target area. Image processing may be used to identify the center of a target area (on a substrate), such as for performing a digital lithography operation. After performance of the digital lithography operation, the target area may be inspected.

[0044] Inspection, e.g., metrology inspection, defect inspection, etc., of substrates 220 A- D can be performed by inspection heads 250A-D responsive to instructions generated by an inspection control module 260. Inspection control module 260 can also control various operations of other FI tools and components, such as opening one of the load ports 128 A-D, transferring (using FI robots 111 A-B ) substrates from the respective enclosure systems 130A-D to buffer stations 210A-D, aligning the substrates (using aligner station 150), closingthe load ports, unsealing one ofvacuum ports 103a, 103b, transferring the substrates to and from buffer stations 210A-D, placing the substrates onto multi-substrate platform 230, moving multi-substrate platform 230 to and from inspection area 240, returning the inspected substrates to buffer stations 210 A-D, and then further to enclosure systems 130A-D, and / or any other associated operations. The inspection heads 250A-D and the inspection control module 250 may together form an inspection tool (e.g., a substrate inspection tool, etc.). The inspection tool may be capable of performing both metrology inspection and defect inspection. Metrology inspection may include measuring one or more properties of a substrate, such as dimensions (e.g., thickness, etc.), roughness (e.g., surface roughness, etc.), and / or material composition of a substrate. Defect inspection may include identifying any kind of defect of a substrate.Attorney Docket No. :39361.320 (L0150PCT)

[0045] In some embodiments, once multi-substrate platform 230 has moved to inspection area 240, a suitable sensor (e.g., a photo-sensor, an electro-mechanical sensor, etc.) can detect arrival of the platform and output a signal informing inspection control module 260 that the substrates are ready for inspection. Inspection control module 260 can generate a signal to inspection heads 250A-D to begin the inspection process. Inspection heads 250A-D may include photo-sensors and / or imaging devices, etc. Signals collected by inspection heads 250A-D can be digitized, denoised, authenticated, compressed, and / or otherwise preprocessed, and converted to a data format recognizable by inspection control module 260. Conversion of the received data can include decompressing the data, rescaling the data, reformatting the data, tokenizing the data, and / or the like.

[0046] Inspection control module 260 can receive (e.g., from an outside processing device) metadata about types of substrates undergoing inspection, e.g., as a bare substrate, a substrate with one or more deposited films, a patterned substrate (substrate with placed dies, etc.), and / or the like, and apply a suitable set of heuristics and / or a model trained to detect defects in substrates of the identified type. The selected tools (e.g., heuristics / models / etc.) can process the collected data and measure profiles of the substrate, identify classes of defects present in the substrates (e.g., in the edge regions of the substrates, the full area of the substrates, and / or the like), determine the number (or density) of such defects, and / or generate any other quality characteristics representative of the quality of the substrates.

[0047] Inspection control module 260 can use the generated characteristics to determine suitability of the substrates for one or more processing operations. For example, a quality score can be computed for a substrate that is based on a number and classes of detected defects. Defects can include cracks, chipped areas, pits / holes, particle defects, pattern defects, contaminated areas, deformations, flaking / peeling, impurities, and / or any other types of imperfections and / or deviations from prescribed specifications. Quality scores can be computed in any suitable way, e.g., with weights being assigned to different classes of defects and to different numbers / densities of those defects. If the quality score is above (or at) a certain empirically determined threshold (which can be dependent on the specific wafer type), the substrate can be determined to be suitable for subsequent downstream processing. If the computed quality score is below (or at) the empirically determined threshold, the substrate can be prevented from undergoing further processing. In some instances, such substrates can be directed for remedial processing (e.g., removal of deposited films, reapplication of the films, edge and / or area polishing, etc.). In other instances, substrates with quality scores below a minimum acceptable threshold may be discarded. In some instances,Attorney Docket No. :39361.320 (L0150PCT) the manufacturing line may be stopped for troubleshooting, identifying a problem with the manufacturing line, changing manufacturing conditions, and / or the like.

[0048] In some embodiments, inspection control module 260 can perform periodic calibration of inspection heads 250A-D. For example, one or more known calibration substrates can be inspected using inspection heads 250A-D. Calibration substrates can be used to verify and / or adjust direction, focus, polarization, illumination intensity, pulse dose, pulse rate, image quality, and / or other characteristics of inspection heads 250A-D.

[0049] In some embodiments, system 200 has a digital-lithography function. System 200 may be capable of performing digital lithography, such as on the multiple substrates supported by the multi-chuck platform. In some embodiments, system 200 may perform a digital lithography operation on multiple substrates supported by the multi-chuck platform after collecting inspection / metrology data. The digital lithography operation may be performed based on the inspection data and / or metrology data, etc. Inspection data may include metrology data. For example, inspection station 160 may perform parallel inspection of multiple substrates supported by the multi-chuck platform to collect metrology data.

[0050] FIG. 2A illustrates schematically example operations 201 that facilitate performance of the multi-substrate inspections by the processing system of FIG. 1, in accordance with some embodiments of the present disclosure. As illustrated, after multisubstrate platform 230 supporting the substrates slides over rail 232 into inspection area 240, inspection heads 250A-D begin scanning the substrates. In some embodiments, inspection heads 250A-D remain stationary while (and / or a dedicated moving stage operating in the inspection area 240) multi-substrate platform 230 is moved along the X-direction and / or F- direction to reposition the substrates relative to inspection heads 250A-D and expose a different region of each substrate to inspection at different times. In some embodiments, platform 230 is moved over rail 232 by one or more actuators.

[0051] In some embodiments, individual chucks can be moved independently relative to inspection heads 250 A-D. For example, individual chucks can be movable relative to multisubstrate platform 230, e.g., by individual moving stages or effectors mounted on multisubstrate platform 230.

[0052] In some embodiments, one or more of inspection heads 250A-D can move during inspection of substrates 220A-D. In some embodiments, the motion of inspection heads 250A-D can additionally be along the vertical (Z) direction. Such motion can be used to change the size of the illuminated spot, resolution, focus, depth of inspection, and / or other inspection parameters. In some embodiments, individual inspection heads 250A-D can moveAttorney Docket No. :39361.320 (L0150PCT) independently along the f-di recti on and / or T-direction to implement individual inspection of the substrates, e.g., using different substrate-specific patterns inspection, speed of inspection, resolution of inspection, types of inspection used (e.g., bright-field inspection, dark -field inspection, and / or the like).

[0053] In some embodiments, inspection heads 250A-D each include one or more imaging devices configured to capture an image of a respective substrate. The imaging devices may include a light emitter to emit light toward a respective substrate. The light emitter may emit an infra-red light (e.g., having a wavelength between approximately 700 nanometers and approximately 1 millimeter), a violet light (e.g., having a wavelength between approximately 380 nanometers and approximately 450 nanometers), an ultraviolet light (having a wavelength between approximately 10 nanometers and approximately 400 nanometers), an extreme ultraviolet (EUV) light (e.g., having a wavelength between approximately 10 nanometers and approximately 120 nanometers), a visible light (having a wavelength between approximately 380 nanometers and approximately 700 nanometers), and / or an x-ray radiation (having a wavelength between approximately 0.01 nanometers and approximately 10 nanometers). Infra-red light may be for detecting cracks in a die and / or voids under a die. Violet and / or ultraviolet light may be for detecting organic material residue. Visible light may be for detecting other defects, such as for determining basic pattern correctness (e.g., the correctness of a printed pattern, deposition pattern, or etched pattern, etc.). In some embodiments, the imaging devices may include an acoustic emitter to emit an acoustic wave and / or an acoustic excitation. In some embodiments, the imaging devices may include an electron beam emitter to emit an electron beam excitation.

[0054] In some embodiments, different substrates 220A-D can be of the same type. In some embodiments, any, some or all substrates 220A-D can be different from other substrates 220A-D being concurrently inspected. “Type” of a substrate may refer to material type, e.g., silicon substrate, glass substrate, corundum substrate, plastic substrate, and / or the like. “Type” of a substrate may further or alternatively include a stage of processing of the substrate, e.g., a bare substrate, a substrate with one or more deposited films, a substrate subjected to plasma deposition, a substrate covered with a mask or a photoresist, a substrate exposed to light, a substrate subjected to etching, a substrate subjected to chemical mechanical polishing (CMP), and / or a substrate that has undergone any number of such operations. “Type” of a substrate also may refer to a kind and / or size of features formed on the substrate, e.g., a die size, a number of dies on the substrate, a height (thickness) of a die, and / or the like.Attorney Docket No. :39361.320 (L0150PCT)

[0055] In some embodiments, different inspection heads 250A-D can be the same or configured to perform the same kind of inspection, e.g., a bright-field inspection (e.g., inspection performed with reflected light), a dark-field inspection (e.g., inspection performed with scattered light), a combination of the bright-field inspection and the dark-field inspection, and / or the like. In some embodiments, any, some or all inspection heads 250 A-D can be configured differently from other inspection heads 250A-D. In such embodiments, after a first stage of inspection, multi-substrate platform 230 can be rotated to 90 degrees, 180 degrees, and / or to any other suitable angle to perform a second (third, fourth, etc.) stage of inspection to expose the substrates to inspection by heads having a different configuration. For example, inspection heads 250A and 250D can be configured to perform a bright-field inspection and inspection heads 250B and 250C can be configured to perform a dark -field inspection. During the first stage, a first pair of substrates may undergo the bright-field inspection while a second pair of substrates undergoes the dark-field inspection. After the first stage, multi-substrate platform 230 can be rotated to 180 degrees so that during the second stage, the first pair of substrates may undergo the dark-field inspection while the second pair of substrates undergoes the bright-field inspection.

[0056] In some embodiments, different inspection heads 250A-D can be configured to perform inspection such as scanning electron microscopy (SEM) imaging inspection, atomic force microscopy (AFM) inspection, and / or scanning acoustic microscopy (SAM) imaging inspection.

[0057] In some embodiments, one or more inspection heads 250A-D can operate in a time delay and integration (TDI) mode. For example, a light source of an inspection head 250x can be a pulsed light source. Each pulse can correspond to a different sensing frame. In the TDI mode, each sensing pixel may aggregate electrical signals (e.g., charge signals, voltage signals, etc.) generated during multiple sensing frames. As a result, multiple low- intensity pulses can be used to achieve high optical sensitivity and resolution without exposing substrates to high-intensity inspection beams capable of causing damage to the substrates. In those instances where inspection is performed while the substrates are moving (e.g., by moving multi-substrate platform 230 in the A-direction and / or F-direction), the signal aggregation in the TDI mode can be performed for pixels that are sequentially exposed to the light reflected or scattered from the same region of the moving substrates.

[0058] In some embodiments, each substrate canbe inspected in a spiral pattern. In other embodiments, one or more substrates may be inspected using a different inspection pattern, such as a raster pattern. FIG. 2B illustrates a spiral pattern of parallel inspection of multipleAttorney Docket No. :39361.320 (L0150PCT) substrates 220A-D, in accordance with some embodiments of the present disclosure. Motion imparted to multi-substrate platform 230 causes inspection spots 252A-D illuminated by the corresponding inspection heads 250A-D to move, relative to substrates 220A-D, along a spiral so that the full area of substrates 220A-D is inspected. Although inspection in FIG. 2B is performed from starting from the center of the wafer, in other embodiments, inspection can be performed in the opposite direction, starting from the edge of the substrates. In some embodiments, inspection can be performed in a zigzag pattern, e.g., with substrates 220A-D scanned along horizontal (or vertical) lines interspaced with shifts along the vertical (horizontal) direction.

[0059] In some embodiments, the placement of the substrates on the multi-substrate platform 230 can differ from a square arrangement, e.g., individual chucks can be arranged in a linear array on multi-substrate platform 230. Similarly, inspection heads 250A-D can be arranged in a matching linear array, each inspection head 250x positioned above the respective substrate.

[0060] In some embodiments, the number of substrates supported by multi-substrate platform 230 canbe less than four (e.g., one, two or three) or more than four (e.g., five, six, eight, and so on). FIG. 2C illustrates inspection of a large-size rectangular substrate (e.g., a “panel”) having a substantially rectangular form and supported by a single chuck of multisubstrate platform 230, in accordance with some embodiments of the present disclosure. In some embodiments, inspection of a panel substrate 270 can be performed in parallel by multiple (e.g., four, as shown) inspection heads. Motion imparted to multi-substrate platform 230 causes inspection spots 252A-D (illuminated by respective inspection heads 250A-D) to move, relative to panel substrate 270, along zigzag paths so that the full area of panel substrate 270 is inspected in a shortened (by a factor 4, compared with a single-head inspection) time. Although FIG. 2C illustrates the zigzag patter of inspection in application to a panel substrate, a similar zigzag pattern can be used for inspection of individual substrates, e.g., instead of the spiral pattern of FIG. 2B.

[0061] FIG. 2D illustrates another example architecture 202 of a multi-substrate inspection system deploying multiple inspection heads per substrate, in accordance with some embodiments of the present disclosure. As illustrated, multiple inspection heads can be used to inspect individual substrates. For example, inspection heads 251 A and 252A inspect a first substrate, inspection heads 25 IB and252B inspect a second substrate, inspection heads 251C and 252C inspect a third substrate, and inspection heads 25 ID and 252D inspect a fourth substrate. Although for brevity and ease of viewing, two inspection heads are shownAttorney Docket No. :39361.320 (L0150PCT) inspecting an individual substrate, the number of inspection heads per substrate need not be limited and can include three inspection heads, four inspection heads, and / or the like. In some embodiments, multiple inspection heads of a same type (e.g., including a same sensor, using a same light wavelength, etc.) may be used for inspection of a single substrate. This may reduce an amount of time that it takes to fully scan the substrate. In some embodiments, different types of inspection heads are used for a single substrate. This may enable multiple different wavelengths of light and / or types of measurements / scans to be performed on a substrate in parallel and / or without moving the substrate to a different station.

[0062] Although for the sake of concreteness, the embodiments of FIGS. 1-2D are illustrated with reference to inspection (e.g., inspection that involves electromagnetic radiation, such as visible light, infrared light, ultraviolet light), similar techniques can be used to perform parallel inspection of multiple substrates (e.g., wafers) using other inspection techniques. In some embodiments, such techniques can include single or multi-head scanning electron microscopy (SEM) imaging to produce SEM imaging data, etc. In some embodiments, SEM imaging can be performed in conjunction with X-ray spectroscopy (e.g., energy dispersive X-ray spectroscopy, EDX). In some embodiments, inspection techniques can include focused Ion beam (FIB) imaging or sample preparation / modification to produce FIB imaging data, etc., single or multi-head atomic force microscopy (AFM) imaging to produce AFM imaging data, etc., multi-head tunneling electron microscopy (TEM) imaging, to produce TEM imaging data, etc., and / or any other applicable inspection technique to produce corresponding inspection data, etc. In such embodiments, multi-head inspection can be performed in a vacuum environment. For example, the substrates being inspected can be delivered, e.g., by an FI robot, from enclosure system(s) to a transfer chamber, a degassing chamber, or any other suitable chamber capable of transitioning from atmospheric conditions to vacuum conditions. In such embodiments, buffer stations 21 Ox, which temporarily store the substrate(s), can be located in the transfer chamber. After the load ports 128x have been sealed and the transfer chamber environment has transitioned from the atmospheric pressure to vacuum conditions (e.g., using a pump), one or more ports interfacing the transfer chamber and a multi-station inspection station can be opened and a second robot (e.g., a transfer chamber robot) can move the substrates from the transfer chamber (e.g., buffer stations) to a multi-substrate platform230 located in the multi-station inspection station, e.g., as illustrated in FIGS. 1-2A. The multi-substrate platform 230 then slides into the inspection area. The reverse process of removal of the substrates through the transfer chamber and the FI (e.g., to enclosure system(s)) can be performed in the order that is opposite to the order of operationsAttorney Docket No. :39361.320 (L0150PCT) illustrated above and includes bringing the environment of the transfer chamber from the vacuum to the atmospheric conditions.

[0063] FIG. 3 illustrates an example multi-head optical inspection system 300 that can be used for parallel multi-substrate inspections, in accordance with at least one embodiment. Multi-head optical inspection system 300 can correspond to multi-substrate inspection system 160 of the previous figures in embodiments. For conciseness and ease of viewing, FIG. 3 illustrates two inspection heads 302 A and 302B inspecting respective substrates 220 A and 220B, but any other number, e.g., three, four, etc., of inspection heads 302x can be used to inspect an individual substrate, in various embodiments. Inspection heads 302A and 302B can be enclosed in separate housings that are depicted schematically with the dashed lines. In some embodiments, inspection heads 302 A and 302B can be enclosed in a single common housing. Substrates 220A and 220B inspected by optical inspection system 300 can be supported by multi-substrate platform 230 that can be moved (e.g., in one or more horizontal and / or vertical direction) by a movable stage 304 (which can include rail 232, with reference to FIG. 2A, a robot blade, and / or any other suitable movable stage). Each inspection head 302x can include an illumination subsystem configured to generate light that is normally and / or obliquely incident on substrates 220A and 220B.

[0064] As depicted in FIG. 3, the illumination subsystem can include a first light source 306 configured to generate a beam of light that is used for normal illumination of substrate 220A (and, possibly, substrate 220B) and that facilitates a bright-field inspection mode. The light emitted by the first light source 306 can be redirected by semi-transparent mirrors 308A and 308B alongthe optical axes ofthe respective inspection heads 302A and 302B. Although in FIG. 3, a single first light source 306 serves both inspection heads 302A and 302B, in other embodiments, inspection heads 302A and 302B can use separate light sources for normal substrate illumination. In some embodiments, the illumination subsystem can include expander optics 310 that transforms the light beam output by first light source 306 into a flood light beam. Although (for brevity) expander optics 310 is depicted in FIG. 3 as a single lens, it should be understood that more than one lens (e.g., including focusing lens(es), defocusing lens(es), collimating lens(es), etc.) can be part of expander optics 310. In some embodiments, expander optics 310 can have one or more reflective optical elements, e.g., curved mirror(s). Parameters of expander optics 310 can be adjustable to enable control over the size (e.g., diameter) of the illuminated spot on substrate(s) 220A and / or 220B.

[0065] The illumination subsystem can further include second light sources 312A and 312B configured to generate light that is used for oblique illumination of substrates 220AAttorney Docket No. :39361.320 (L0150PCT) and / or 220B. As illustrated in FIG. 3, oblique incidence light can be directed to substrates 220 A and / or 220B using reflective optical elements 314A and / or 314B. In some embodiments, oblique incidence light can be a flood light processed by expander optics 316A and / or 316B. Expander optics 316x can be configured to illuminate a portion of substrates 220A and / or 220B that are adjustable in size (e.g., semi-axes of elliptical regions of substrates 220A and / or 220B illuminated by the obliquely -incident light). The illumination subsystem of each inspection head 302A and / or 302B can further include one or more polarizing elements (not shown in FIG. 3 for conciseness) configured to control polarization of the light incident on substrates 220A and 220B (e.g., a polarizing element placed in the optical path of the light beam produced by the first light source 306). The illumination subsystem of each inspection head 302A and 302B can be capable of delivering light of adjustable intensity, e.g., by controlling the size of the illuminated spot and / or by controlling the intensity of light output by the first light source 306 (or multiple first light sources 306).

[0066] Each inspection head 302x can include a collection subsystem configured to collect light reflected from the respective substrate 220x. Collection subsystem of inspection head 302x can include corresponding objective 318x. Objective(s) 318x can include one or more lenses (three lenses are shown in FIG. 3, as an example) configured to collect bright- field light reflected from respective substrate 220x and / or dark-field light scattered from substrate 220x. The number and types of lenses of the objectives 318x can be selected, e.g., using any known techniques, to reduce light aberration in the objectives 318x, including chromatic aberration. In some embodiments, each objective 318x can have an outer diameter of 75 mm or less. Collection subsystem of inspection heads 302 A and / or 302B can include one or more polarization elements (e.g., filters) 320x configured to pass light of a specific target polarization, e.g. s-polarization, p-polarization, right-handed circular (or elliptic) polarization, left-handed circular (or elliptic) polarization, and so on. Although objectives 318x in FIG. 3 are shown to include lenses, in some embodiments, one or more parabolic or spherical mirrors can be use as part of (catadioptric) objectives 318x.

[0067] Collection subsystems of inspection head 302 A and / or inspection head 302B can further include one or more directional filters 322x configured to pass light collected from a particular interval of angles of reflection (or scattering) from substrates 220 A, 220B. Directional filters 322x can be implemented via a light absorbing plate in which suitable apertures are cut out for the passage of light. For example, a central aperture 324 admits the normally reflected light (and allows passage of incident light from first light source 306) whereas a side aperture 326 admits scattered light (an example scattered light beam 328 isAttorney Docket No. :39361.320 (L0150PCT) depicted schematically). In some embodiments, directional filters 322x can be positioned at the Fourier plane of the respective objective 318x. In some embodiments, directional filters 322x can be positioned at some distance from the Fourier plane. Although FIG. 3 depicts directional filters 322x positioned farther away from objectives 318x than the polarizing elements 320x, in some embodiments, directional filters 322x can be positioned closer to the corresponding objectives than the polarizing elements 320x. In some embodiments, directional filters 322x are characterized by adjustable numerical apertures controlled by the size of the physical apertures and / or positioning of the physical apertures relative to objectives 318x.

[0068] Detection subsystem of each inspection head 302 A and / or 302B can include relay optics 330A and / or 33 OB. Each relay optics 33 Ox can include one or more optical elements (e.g., lenses, mirrors, waveguides, arrays of waveguides, etc.) to deliver (e.g., focus) the reflected and scattered light on a corresponding array of light detectors 332A and 332B. Light detectors 332A and / or 332B can use complementary metal-oxide-semiconductor (CMOS) image sensors, charge-coupled devices (CCDs), hybrid CMOS-CCD image sensors, photomultiplier tubes (e.g., an array of photocathode-based pixels), photodiodes, phototransistors, or any other suitable photon detectors. Light detectors 332A and / or 332B can image a separate spot on the respective substrates 220 A and 220B illuminated by the corresponding inspection head 302x. The light intensity (e.g., reflectivity) data collected by light detectors 332x can be provided to inspection control module 260 that determines the sizes / types / concentrations / locations of various defects and imperfections of substrates 220A and 220B. Inspection control module 260 can be in communication with inspection control module 260 capable of changing settings of inspection heads 302 A and / or 302B based on instructions from inspection control module 260. For example, an initial inspection can be performed with a certain set resolution. When the presence of a defect is identified by inspection control module 260, e.g., based on light reflectivity data collected by light detectors 332A and / or 332B, etc., wafer quality control module 260 can output an instruction to change resolution of inspection by zooming the corresponding inspection head 302 A and / or 302B to a specific region on substrates 220A and 220B where the defect is located. More specifically, inspection control module 260 can change focal distance of objective(s) 318 A and / or 318B, change the distance from the respective objective 318 A and / or 318B to substrates 220A and 220B, e.g., using Z-motion of movable stage 304 and / or inspection heads 302A and / or 302B, for individualized defect detection in separate substrates.Inspection control module 260 can additionally change numerical apertures of directionalAttorney Docket No. :39361.320 (L0150PCT) filters 322A and / or 322B to change resolution of the inspection. The inspection of the current locations of substrates 220 A and / or 220B can be completed by the inspection heads 302A and / or 302B after a target amount of light is collected by each of light detectors 332 A and / or 332B from substrate(s) 220A and / or 220B. Subsequently, the movable stage 304 can reposition substrates 220A and / or 220B. Stage control module 340 can determine the distance and direction of repositioning of substrate(s) 220A and / or 220B so that previously uninspected spots are exposed to inspection heads 302A and / or 302B. Coordination between the motion of the movable stage 304 and the collection of inspection data can be facilitated by a synchronization module 350.

[0069] In some embodiments, CMOS image sensors, CCD image sensors, and / or any other sensing elements of light detectors 332x can operate in a time delay and integration (TDI) mode. For example, if the first light source 306 and / or second light sources 312x are pulsed light sources, each pulse can correspond to a sensing frame. In the TDI mode, each sensing pixel may aggregate electrical signals (e.g., charge signals, voltage signals, etc.) generated during multiple sensing frames. As a result, a number of low-intensity pulses can be used to achieve high inspection sensitivity and resolution without exposing substrate(s) 220A and / or 220B to high-intensity beams capable of causing damage to the wafer. In those instances, where inspection is performed on a moving substrate(s) 220A and / or 220B (e.g., transported by movable stage 304) the signal aggregation in the TDI mode can be performed for pixels that are sequentially exposed to the light reflected or scattered from the same region of the moving substrate(s) 220A and / or 220B.

[0070] In some embodiments, CMOS image sensors used in light detectors 332x can be high-speed and low-noise sensors. For example, CMOS image sensors can have speed at or above 1 Gigapixel per second and noise at lOe or less, e.g., in the range of 2e-5e or even less, in some embodiments.

[0071] In some embodiments, inspection heads 302A and / or 302B can be independently configurable into one of a plurality of spectral configurations, each configuration characterized by a different wavelength of the normally -incident beam of light and / or the obliquely -incident beam of light. In some embodiments, inspection heads 302A and / or 302B are independently configurable into one of a plurality of configurations, wherein in each of the plurality of configurations is characterized by a different numerical aperture for collection of the generated light.

[0072] In some embodiments, inspection heads 302A and / or 302B can be independently configurable into one of a plurality of field-of-view configurations, in which the respectiveAttorney Docket No. :39361.320 (L0150PCT) collection subsystem is characterized by a different numerical aperture for collection of the generated light.

[0073] In some embodiments, inspection heads 302A and / or 302B can be independently configurable into one of a plurality of detection configurations, in which the respective detection subsystem is characterized by at least one of a different gain, a different data rate, or a different dynamic range.

[0074] In some embodiments, inspection heads 302A and / or 302B can include a phase contrast function. For example, a normally -incident beam generated by one or multiple inspection heads can include polarizers that cause the normally-incident beam to be split (e.g., using Wollaston prisms) into two beams with different polarizations (e.g., an s- polarized beam and a p-polarized beam). The reflected polarized beams can then pass through the polarizers to obtain a combined beam having an interference pattern that is detected by light detectors 332x. In some embodiments, inspection heads 302A and / or 302B can include a differential interference contrast (DIC) functionality where the normally-incident beam is split into two beams of different polarizations that follow close but different optical paths and probe two closely spaced locales of substrate(s) 220A and / or 220B.

[0075] In some embodiments, incident light (e.g., normally-incident light) can be in a polarized state, e.g., an s-polarized state or a p-polarized state, in a right-handed circularly (or elliptically) polarized state or in a left-handed circularly (or elliptically) polarized state, or in any combination thereof. In some embodiments, the collection subsystem of inspection system 300 can separately collect reflected light with different polarizations. For example, the reflected beam may pass through a polarizing optical element (e.g., a prism) so that the components of the reflected light with different polarization can be directed to different optical paths and can be detected independently. In some embodiments, polarizing elements 320x can let one of the polarization states (e.g., s-polarized or right-handed light) of the reflected light to pass through and reject the other polarization state (e.g., p-polarized or light- handed light) of the reflected light, or vice versa. In some embodiments, polarizing elements 320x can let through one of the polarization states of the reflected light reflected off a first region of a substrate 220x and let through the other polarization state of the reflected light reflected off a second region of the substrate 220x.

[0076] In some embodiments, each (or some) region of substrate(s) 220A and / or 220B can be inspected using two or more angles of incidence of the normally-incident and / or obliquely incident beams. The directions of the incident beams can be controlled, in some embodiments, by tilting inspection head(s) 302A and / or 302B and / or some elements ofAttorney Docket No. :39361.320 (L0150PCT) inspection head(s) 302A and / or 302B, e.g., semi-transparent mirror(s) 308x (e.g., to change directions of the normally-incident beams) and / or reflective optical element(s) 314x (e.g., to change directions of the obliquely -incident beams). Two or more sets of inspection data obtained by light detectors 332A and / or 332B for different tilt angles can be aggregated (e.g., averaged) to reduce effects of reflection signal loss (or distortion) due to speckle artifacts.

[0077] In some embodiments, multiple sets of inspection data for specific regions of substrate(s) 220A and / or 220B generated using inspection head(s) 302A and / or 302B can be fused to obtain a combined data for the region. The combined data may include multiple data sets obtained using one or more inspection head(s) 302 A and / or 302B one or more angles of incidence of normally-incident light, one or more angles of incidence of obliquely-incident light, one or more resolutions, one or more sensing beam powers, and the like. In some embodiments, different data sets obtained for the same region can be used to eliminate or reduce noise in the combined data for that region. Different data sets obtained for the same region may provide complementary information about defects and imperfections located in those regions. In some embodiments, various defects can be classified among a plurality of classes (bins), e.g., a particle defect, a narrow contaminated area, a wide contaminated area, a hump, a groove, a wafer crack, a wafer deformation, a flaking of a film deposited on the wafer, and the like. Classification of defects among the classes can be based on the multiple sets of data obtained by various modes (channels) of inspection head(s) 302 A and / or 302B with different inspection modes detecting different optical features of the respective defects / imperfections.

[0078] Although configuration and operations of the multi-head optical inspection system are illustrated above for two inspection heads 302A and 302B, which inspect two substrates 220A and 220B, the number of inspection heads and substrates need not be limited. In some embodiments, N inspection heads 302x can inspect (in parallel) N substrates 220x. In some embodiments, k X N inspection heads 302x can inspect (in parallel) N substrates 220x, with k =2, 3, or more inspection heads 302x inspecting an individual substrate.

[0079] FIG. 4 is a flow diagram of an example method 400 of parallel inspection of multiple substrates using a multi-head inspection system, in accordance with some embodiments of the present disclosure. In some embodiments, method 400 is performed using systems and components shown in FIGS. 1-3 and / or any combination thereof. In some embodiments, method 400 is performed responsive to instructions from various processing devices, e.g., a processing device of data processing server 162 of FIG. 1 or inspectionAttorney Docket No. :39361.320 (L0150PCT) control module 260 of FIGS. 1-2A. In one embodiment, the processing device(s) used to perform method 400 deploys one or more central processing units (CPUs), microprocessors, DSPs, ASICs, finite-state machines, FPGAs, and so on, coupled to one or more memory devices (e.g., a random-access memory, a read-only memory, a flash memory, a static memory, and so on). In some embodiments, the processing devices execute software or firmware instructions of method 400 stored in a memory device(s). In some embodiments, some of the blocks of method 400 are optional.

[0080] Method 400 can be performed to perform concurrent inspection of a plurality of substrates, e.g., two substrates, three substrates, four substrates, or any other number of substrates. The inspection can include collecting optical or other inspection data for each of the substrates. For example, inspection data can include images of substrates (or any portion of substrates), spectral characteristics of substrates, transmissivity of substrates, reflectivity of substrates, absorptivity of substrates, and / or any other data representative of physical and / or chemical state of substrates. At block 410, method 400 can include identifying, using one or more aligner devices, a reference feature for each of the plurality of substrates. In some embodiments, identifying the reference features can be performed while the inspection data is being collected for a preceding plurality of substrates.

[0081] At block 420, method 400 can continue with aligning, using the identified reference features, the plurality of substrates. For example, aligning a reference feature of a given substrate can include positioning the reference feature in a predetermined arrangement relative to a robot (or an arm of the robot). A reference feature can include a notch, a cut-out portion, and / or any other suitable marker that indicates orientation of a suitable set of axes (e.g., directions of patterning, grid of dies, crystallographic directions, and / or the like).

[0082] At block 430, method 400 can include supporting, using a multi-substrate platform, the plurality of substrates. Each substrate of the plurality of substrates can be supported by a respective substrate holder of a plurality of substrate holders in or on the multi-substrate platform. A substrate holder can include any suitable device capable of securely holding a substrate, including but not limited to an electrostatic chuck, a vacuum suction chuck, and / or the like. In some embodiments, operations of block 430 can include operations performed at different times before, during, and / or after collection of inspection data, as illustrated with the top callout portion of FIG. 4.

[0083] More specifically, at block 432, operations include placing, using a robot, the plurality of substrates on the multi-substrate platform, e.g., after retrieving the substrates from substrate buffers where the substrates are stored temporarily. Placing two or more of theAttorney Docket No. :39361.320 (L0150PCT) substrates on the multi-substrate platform can be performed concurrently, e.g., with a separate robot arm handling each substrate. The multi-substrate platform may be loaded with multiple substrates while in a loading position. Atblock 434, method400 can include transporting the multi-substrate platform to an inspection area (e.g., an inspection position). Placing the substrates on the multi-substrate platform and transporting the multi-substrate platform to an inspection area can be performed prior to collecting the inspection data. At block 436, method 400 can include removing, using the robot, the plurality of substrates from the multisubstrate platform. In some embodiments, the removed substrates can be placed temporarily into the substrate buffers. In some embodiments, removal of the substrates can be performed concurrently, e.g., with separate robot arms handling individual substrates in parallel. The removal of the substrates can be performed after collecting the inspection data.

[0084] At block 440, method 400 can continue with collecting inspection data for the plurality of substrates. In some embodiments, collecting the inspection data for multiple substrates can be performed in parallel. In some embodiments, the inspection data for each substrate of the plurality of substrates is collected using a respective inspection head of a plurality of inspection heads. In some embodiments, the inspection data can include a bright- field inspection data, e.g., data collected using specularly reflected light, a dark -field inspection data, e.g., data collected using diffusively scattered light. In some embodiments, the inspection data can be performed using a visible light, an infrared (IR) light, an ultraviolet (UV) light, and / or any combination thereof. In some embodiments, inspection data can include SEM imaging data for the respective substrate, AFM imaging data for the respective substrate, and / or SAM imaging data for the respective substrate.

[0085] In some embodiments, block 440 can include operations illustrated with the bottom callout portion of FIG. 4. Collecting the inspection data at block 440 can be performed in parallel for each of the plurality of substrates. More specifically, at block 442, method 400 can include directing an incident light to a corresponding substrate. At block 444, method 400 can include collecting light generated, in response to the incident light, by the corresponding substrate. At block 446, method 400 can include detecting, using one more light detectors, intensity of the generated light for a plurality of angles of propagation of the generated light. For example the intensity can be detected as a function of angle 6 of scattering or reflection.

[0086] At block 450, method 400 can include identifying, by a processing device and using the collected inspection data for the plurality of substrates, at least one of (i) presence of one or more defects in at least one substrate of the plurality of substrates or (ii.) one orAttorney Docket No. :39361.320 (L0150PCT) more measurements of one or more properties of at least one substrate of the plurality of substrates. In some embodiments, method 400 can include determining (e.g., evaluating), in view of the identified presence of the one or more defects in a particular substrate, a quality of the substrate. Depending on the determined quality, method 400 can select one of the following actions. If the quality of the substrate is at or above a threshold quality, the substrate can be determined to be suitable for further processing and directed the substrate for processing in a processing chamber. If the quality of the substrate is below a threshold quality, the substrate can be determined to be unsuitable for further processing and the substrate can be prevented from entering the processing chamber. In some instances, e.g., when the substrate is not suitable for further processing but is not irreversibly impaired, the substrate can be directed for a defect-mitigation processing. Such defect-mitigation processing can include additional polishing, drying, cleaning, application of solvents, removal of improperly deposited films, and / or the like. In some embodiments, method 400 can include identifying one or more measurements of a corresponding substrate, such as a thickness measurement, a roughness measurement, or any other dimensional measurement, etc. In some embodiments, method 400 can include identifying a material composition of a corresponding substrate based on the inspection data.

[0087] FIG. 5 is a block diagram illustrating an exemplary system 500 (exemplary system architecture), according to some embodiments. In some embodiments, system 500 can be used to inspect substrates and / or update pattern designs that are to be formed on substrates. The system 500 includes a client device 520, manufacturing equipment 524, inspection equipment 528, predictive server 512, and data store 540. The predictive server 512 may be part of predictive system 510. Predictive system 510 may further include server machines 570 and 580.

[0088] Inspection equipment 528 may inspect substrates for defects. In some embodiments, inspection equipment 528 includes a multi-substrate inspection platform as described herein above. In some embodiments, inspection equipment 528 includes multiple inspection heads for concurrent inspection of multiple substrates. The multiple substrates may be secured to a multi-substrate platform such as by one or more vacuum chucks or electrostatic chucks, etc.

[0089] Imaging devices of the inspection equipment 528 may capture images of inspected substrates. The captured images may be stored as substrate image data 560, for example, in data store 540. Substrate image data 560 may include historical image data 564 and / or current image data 566. Historical image data 564 may include data indicative of historical images ofAttomey Docket No. :39361.320 (L0150PCT) historical substrates. Current image data 566 may include data indicative of current images of current substrates, such as substrates currently under inspection, such as by inspection equipment 528. In some embodiments, current image data 566 may be provided to one or more processing devices for performance of defect detection and / or defect determination, etc. In some embodiments, current image data 566 may be provided to one or more trained machine learning models for generation of a reference image, and / or for generation of a printing pattern, etc. In some embodiments, substrate image data 560 is indicative of images captured using visible light, infra-red light, violet light, and / or ultraviolet light.

[0090] Predictive system 510 may be utilized to generate current design correction data 576. Different current design correction data 576 may be determined for different types of input substrate image data and / or other input data related to defects (e.g., engineer observations, etc.). Some current design correction data 576 may be generated by performing image processing operations and / or machine learning operations on image data of one or more substrates. Current design correction data 576 may be indicative of one or more changes to a pattern design, such as a design (e.g., a die design, etc.) printed on a substrate. Operations for determining current design correction data 576 may be performed by predictive server 512, predictive component 514, etc.

[0091] In some embodiments, predictive system 510 may be utilized to generate current design correction data 576 using one or more trained machine learning models. In some embodiments, predictive system 510 may generate current design correction data 576 using supervised machine learning. Supervised machine learning refers to operations associated with a machine learning model that was provided with labeled training data, such as image data labeled by counts of defects. In some embodiments, predictive system 510 may generate current design correction data 576 using unsupervised machine learning. Unsupervised machine learning refers to operations associated with a machine learning model that was trained using unlabeled input. Unsupervised machine learning operations may include clustering results, principle component analysis, anomaly detection, etc. In some embodiments, predictive system 510 may generate current design correction data 576 using semi-supervised learning (e.g., training data may include a mix of labeled and unlabeled data, etc.).

[0092] Client device 520, manufacturing equipment 524, inspection equipment 528, predictive server 512, data store 540, server machine 570, and server machine 580 may be coupled to each other via network 530 for generating current design correction data 576 to perform corrective actions. In some embodiments, network 530 may provide access to cloud-Attorney Docket No. :39361.320 (L0150PCT) based services. Operations performed by client device 520, predictive system 510, data store 540, etc., may be performed by virtual cloud-based devices.

[0093] In some embodiments, network 530 is a public network that provides client device 520 with access to the predictive server 512, data store 540, and other publicly available computing devices. In some embodiments, network 530 is a private network that provides client device 520 access to manufacturing equipment 524, inspection equipment 528, data store 540, and other privately available computing devices. Network 530 may include one or more Wide Area Networks (WANs), Local Area Networks (LANs), wired networks (e.g., Ethernet network), wireless networks (e.g., an 802.11 network or a Wi-Fi network), cellular networks (e.g., a Long Term Evolution (LTE) network), routers, hubs, switches, server computers, cloud computing networks, and / or a combination thereof.

[0094] Client device 520 may include computing devices such as Personal Computers (PCs), laptops, mobile phones, smartphones, tablet computers, netbook computers, network connected televisions (“smart TV”), network-connected media players (e.g., Blu-ray player), a set-top-box, Over-the-Top (OTT) streaming devices, operator boxes, etc. Client device 520 may include a pattern update component 522. Pattern update component 522 may receive user input (e.g., via a Graphical User Interface (GUI) displayed via the client device 520) of an indication associated with manufacturing equipment 524. In some embodiments, pattern update component 522 transmits the indicationto the predictive system 510, receives output (e.g., current design correction data 576) from the predictive system 510, determines a corrective action based on the output, and causes the corrective action to be implemented. In some embodiments, pattern update component 522 obtains substrate image data 560 and provides substrate image data 560 to predictive system 510.

[0095] In some embodiments, pattern update component 522 may retrieve current image data 566 (e.g., one or more images of a substrate of interest, etc., possibly including indications of defects). Image data 566 my be indicative of one or more images of one or more dies, etc. Pattern update component 522 may provide the current image data 566 to predictive system 510 for determining updates to a pattern design on the current substrate (e.g., for determining a modified die pattern design, etc.). Updates to the pattern may include updating substrate design parameters 550. Predictive system 510 may utilize image processing, feature detection operations, image filtering, frequency domain transformations, and / or other techniques for making determinations in association with the image data and defects.Attorney Docket No. :39361.320 (L0150PCT)

[0096] Predictive system 510 may provide one or more images of a substrate and / or one or more images of one or more dies placed on a substrate to one or more trained machine learning models (e.g., model(s) 590) for making determinations in association with patterns of the substrate. In some embodiments, different ML models are trained to determine pattern update information (e.g., modified die pattern design information, etc.) for different types of input data. In some embodiments, model(s) 590 may represent one or more physics-based models, image processing algorithms, and / or machine learning models. In some embodiments, model(s) 590 may instead or additionally be configured to make recommendations (e.g., generate current design correction data 576) indicative of one or more updates to a pattern design (e.g., a die pattern design, etc.) that is to be formed on a substrate (e.g., a modified pattern design that is to be formed on the substrate, a modified die pattern design, etc.). In some embodiments, model(s) 590 may be configured to make recommendations about an optical wavelength that is to be used for illuminating a substrate under inspection as described herein with respect to FIGS. 7C-7E.

[0097] In some embodiments, substrate design parameters 550 are indicative of a pattern that is to be formed on a substrate (e.g., the wire connection patterns between pad to pad on different die, etc.), such as by manufacturing equipment 524. The pattern may be formed on the substrate (e.g., on the die, etc.) by printing, deposition, and / or by etching. In some embodiments, the pattern is formed based on the position / location of contact pads and / or traces on the substrate which are indicated in the substrate image data 560.

[0098] In some embodiments, pattern update component 522 receives an indication of a corrective action (e.g., such as an update to the substrate design parameters 550 indicated in the design correction data 571, predicted substrate design parameters 550, etc.) and causes the corrective action to be implemented. Each client device 520 may include an operating system that allows users to one or more of generate, view, or edit data (e.g., indication associated with manufacturing equipment 524, corrective actions associated with manufacturing equipment 524, etc.).

[0099] Historical image data 564 may correspond to products processed using manufacturing parameters and / or recipes associated with historical sensor data and historical manufacturing parameters. Historical image data 564 may correspond to image data generated for substrates that experienced one or more historical defects. Historical image data 564 may correspond to image data generated for manufactured dies that experienced one or more historical defects. Such historical image data 564 may be associated with historical useAttorney Docket No. :39361.320 (L0150PCT) cases. Historical design correction data 574 for the historical defects may include design corrections to the substrate parameters 550 associated with the historical defects.

[0100] Manufacturing parameters maybe suboptimal for producing products. This may have costly results of increased resource (e.g., energy, coolant, gases, etc.) consumption, increased amount of time to produce the products, increased component failure, increased amounts of defective products, increased environmental impact, etc. By inputting substrate image data 560 into predictive system 510, receiving output of design correction data 571, and performing a corrective action of updating substrate design parameters 550 (e.g., updatingthe design of a pattern formed on substrates, predicting substrate design parameters 550, etc.), system 500 can have the technical advantage of using optimal pattern design to avoid costly results of suboptimal patterns formed on substrates.

[0101] Substrate design parameters 550 may include hardware parameters (e.g., information indicative of which components are installed in manufacturing equipment 524, indicative of component replacements, indicative of component age, indicative of software version or updates, etc.) and / or process parameters (e.g., temperature, pressure, flow, rate, electrical current, voltage, gas flow, lift speed, etc.). In some embodiments, the corrective action includes causing design optimization (e.g., updating manufacturing parameters, manufacturing processes, manufacturing equipment 524, etc. for an optimized product). In some embodiments, the corrective action includes updating a recipe (e.g., altering the timing of manufacturing subsystems entering an idle or active mode, altering set points of various property values, etc.). In some embodiments, substrate design parameters 550 may include parameters indicating a pattern to be formed (e.g., by manufacturing equipment 524) on a substrate, such as a die pattern, etc. In some embodiments, the corrective action is an update to the pattern to avoid future manufacturing errors, etc.

[0102] Predictive server 512, server machine 570, and server machine 580 may each include one or more computing devices such as a rackmount server, a router computer, a server computer, a personal computer, a mainframe computer, a laptop computer, a tablet computer, a desktop computer, Graphics Processing Unit (GPU), accelerator Application- Specific Integrated Circuit (ASIC) (e.g., Tensor Processing Unit (TPU)), etc. Operations of predictive server 512, server machine 570, server machine 580, data store 540, etc., may be performed by a cloud computing service, cloud data storage service, etc.

[0103] Predictive server 512 may include a predictive component 514. In some embodiments, the predictive component 514 may receive current image data 566 and generate output (e.g., current design correction data 576) that provides information aboutAttorney Docket No. :39361.320 (L0150PCT) updates to the substrate design parameters 550 (e.g., provides information about predicted substrate design parameters 550, etc.). In some embodiments, current design correction data 576 may include an adjustment of a size and / or location of a die pattern feature such as a contact pad, electrical trace, or a ground pattern layer, etc. of a die placed on a substrate. Receiving current image data 566 may include obtaining the data from client device 520. Receiving current image data 566 may include retrieving the data from data store 540.

[0104] In some embodiments, predictive component 514 may determine information about defects represented in image data (e.g., substrate image data 560, indicative of one or more manufactured dies, etc.) based on image processing operations. For example, predictive component 514 may receive substrate image data (e.g., indicative of one or more manufactured dies). Predictive component 514 may perform pre-processing to improve usability of the image, such as gaussian blur operations, sharpening operations, etc. Predictive component 514 may perform transform operations (e.g., Fourier transform operations) to transform the image data from a spatial domain to a frequency domain. Predictive component 514 may apply filteringto the frequency domain image data. The filters may be designed to accentuate, highlight, or the like one or more target defects. In some embodiments, an image of a substrate may have several filters applied, to generate a number of filtered image data for determining different types of defects. Predictive component 514 may perform additional image processing techniques, such as transforming the image data back to a spatial domain, performing image thresholding, etc. Image thresholding may include increasing contrast of an image, e.g., to improve feature detection, defect detection, defect classification, etc. Image thresholding may include adjusting each pixel with brightness above a threshold value to maximum brightness (e.g., to a brightness of 1) and adjusting each pixel with brightness below the threshold value to a minimum brightness (e.g., to a brightness of 0). Predictive component 514 may perform feature detection operations such as contour detection, circle detection, or the like, to determine presence evidence of defects (e.g., die defects, etc.) in the image data.

[0105] Manufacturing equipment 524 may be associated with one or more machine leaning models, e.g., model(s) 590. Machine learning models associated with manufacturing equipment 524 may perform many tasks, including process control, classification, performance predictions, etc. Model(s) 590 may be trained using data associated with manufacturing equipment 524 or products processed by manufacturing equipment 524, substrate design parameters 550 (e.g., associated with substrate design), substrate image data 560 (e.g., generated by inspection equipment 528), design correction data 571, etc.Attorney Docket No. :39361.320 (L0150PCT)

[0106] In some embodiments, predictive system 510 may include a machine learning model that outputs updates to substrate design parameters 550 (e.g., predicted substrate design parameters 550). For example, imaging techniques may be performed that generate images of at least a subset of a surface of the substrate. Images of one or more dies forms on a surface of the substrate may be captured. A machine learning model (e.g., executed by predictive component 514) may be utilized to determine an update to a pattern (e.g., a die pattern) formed on the substrate depicted in the image(s). The machine learning model may be utilized to determine a modified pattern (e.g., a modified die pattern, etc.) to be formed on a substrate. In some embodiments, predictive system 510 includes a trained machine learning model that receives current image data 566. The trained machine learning model may determine updates to a pattern design or a modified pattern design (e.g., indicated in current design correction data 576).

[0107] One type of machine learning model that may be used to perform some or all of the above tasks is an artificial neural network, such as a deep neural network. Artificial neural networks generally include a feature representation component with a classifier or regression layers that map features to a desired output space. A convolutional neural network (CNN), for example, hosts multiple layers of convolutional filters. Pooling is performed, and nonlinearities may be addressed, at lower layers, on top of which a multi-layer perceptron is commonly appended, mapping top layer features extracted by the convolutional layers to decisions (e.g. classification outputs). CNNs have found particular applicability in the area of image processing, e.g., processing substrate image data for epitaxial defect determination.

[0108] A recurrent neural network (RNN) is another type of machine learning model. A recurrent neural network model is designed to interpret a series of inputs where inputs are intrinsically related to one another, e.g., time trace data, sequential data, etc. Output of a perceptron of an RNN is fed back into the perceptron as input, to generate the next output.

[0109] Deep learning is a class of machine learning algorithms that use a cascade of multiple layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. Deep neural networks may learn in a supervised (e.g., classification) and / or unsupervised (e.g., pattern analysis) manner. Deep neural networks include a hierarchy of layers, where the different layers learn different levels of representations that correspond to different levels of abstraction. In deep learning, each level learns to transform its input data into a slightly more abstract and composite representation. In an image recognition application, for example, the raw input may be a matrix of pixels; the first representational layer may abstract the pixels and encode edges; theAttorney Docket No. :39361.320 (L0150PCT) second layer may compose and encode arrangements of edges; the third layer may encode higher level shapes (e.g., teeth, lips, gums, etc.); and the fourth layer may recognize a scanning role. Notably, a deep learning process can learn which features to optimally place in which level on its own. The "deep" in "deep learning" refers to the number of layers through which the data is transformed. More precisely, deep learning systems have a substantial credit assignment path (CAP) depth. The CAP is the chain of transformations from input to output. CAPs describe potentially causal connections between input and output. For a feedforward neural network, the depth of the CAPs may be that of the network and may be the number of hidden layers plus one. For recurrent neural networks, in which a signal may propagate through a layer more than once, the CAP depth is potentially unlimited.

[0110] In some embodiments, predictive component 514 receives current sensor data, current image data 566 and / or current manufacturing parameters, performs signal processing to break down the current data into sets of current data, provides the sets of current data as input to a trained model(s) 590, and obtains outputs indicative of current design correction data 576 from the trained model(s) 590. In some embodiments, predictive component 514 receives image data of a substrate and provides the image data to trained model(s) 590.Model(s) 590 may be configured to accept data indicative of substrate images and generate as output updates to substrate pattern design (e.g., output a predicted substrate pattern design).

[0111] In some embodiments, the various models discussed in connection with model(s) 590 (e.g., supervised machine learning model, unsupervised machine learning model, etc.) may be combined in one model (e.g., an ensemble model), or may be separate models.

[0112] Data may be passed back and forth between several distinct models included in model(s) 590 and predictive component 514. In some embodiments, some or all of these operations may instead be performed by a different device, e.g., client device 520, server machine 570, server machine 580, etc. It will be understoodby one of ordinary skill in the art that variations in data flow, which components perform which processes, which models are provided with which data, and the like are within the scope of this disclosure.

[0113] Data store 540 may be a memory (e.g., random access memory), a drive (e.g., a hard drive, a flash drive), a database system, a cloud-accessible memory system, or another type of component or device capable of storing data. Data store 540 may include multiple storage components (e.g., multiple drives or multiple databases) that may span multiple computing devices (e.g., multiple server computers). The data store 540 may store manufacturing parameters 550, substrate image data 560, design correction data 571, and substrate design parameters 550.Attorney Docket No. :39361.320 (L0150PCT)

[0114] In some embodiments, predictive system 510 further includes server machine 570 and server machine 580. Server machine 570 includes a data set generator 572 that is capable of generating data sets (e.g., a set of data inputs and a set of target outputs) to train, validate, and / or test model(s) 590, including one or more machine learning models. In some embodiments, data set generator 572 may partition the historical data (e.g., historical image data 564 and / or historical design correction data 574) into a training set (e.g., sixty percent of the historical data), a validating set (e.g., twenty percent of the historical data), and a testing set (e.g., twenty percent of the historical data).

[0115] In some embodiments, predictive system 510 (e.g., via predictive component 514) generates multiple sets of features. For example a first set of features may correspond to a first set of types of image data (e.g., from a first set of imaging equipment, first type of substrate imaging technique, etc.) that correspond to each of the data sets (e.g., training set, validation set, and testing set) and a second set of features may correspond to a second set of types of image data (e.g., from a second type of imaging device) that correspond to each of the data sets.

[0116] Server machine 580 includes a training engine 582, a validation engine 584, selection engine 585, and / or a testing engine 586. An engine (e.g., training engine 582, a validation engine 584, selection engine 585, and a testing engine 586) may refer to hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (such as instructions run on a processing device, a general purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof. The training engine 582 may be capable of training a model(s) 590 using one or more sets of features associated with the training set from data set generator 572. The training engine 582 may generate multiple trained models 590, where each trained model(s) 590 corresponds to a distinct set of features of the training set (e.g., metrology data from distinct metrology devices). Data set generator 572 may receive the output of a trained model, collect that data into training, validation, and testing data sets, and use the data sets to train a second model (e.g., a machine learning model configured to output predictive data, corrective actions, etc.).

[0117] Validation engine 584 may be capable of validating a trained model(s) 590 using a corresponding set of features of the validation set from data set generator 572. For example, a first trained machine learning model(s) 590 that was trained using a first set of features of the training set may be validated using the first set of features of the validation set. The validation engine 584 may determine an accuracy of each of the trained models 590 based on the corresponding sets of features of the validation set. Validation engine 584 may discardAttorney Docket No. :39361.320 (L0150PCT) trained models 590 that have an accuracy that does not meet a threshold accuracy. In some embodiments, selection engine 585 maybe capable of selecting one or more trained models 590 that have an accuracy that meets a threshold accuracy. In some embodiments, selection engine 585 may be capable of selecting the trained model(s) 590 that has the highest accuracy of the trained models 590.

[0118] Testing engine 586 may be capable of testing a trained model(s) 590 using a corresponding set of features of a testing set from data set generator 572. For example, a first trained machine learning model(s) 590 that was trained using a first set of features of the training set may be tested using the first set of features of the testing set. Testing engine 586 may determine a trained model(s) 590 that has the highest accuracy of all of the trained models based on the testing sets.

[0119] In the case of a machine learning model, model(s) 590 may refer to the model artifact that is created by training engine 582 using a training set that includes data inputs and corresponding target outputs (correct answers for respective training inputs. Patterns in the data sets can be found that map the data input to the target output (the correct answer), and machine learning model(s) 590 is provided mappings that capture these patterns. The machine learning model(s) 590 may use one or more of Support Vector Machine (SVM), Radial Basis Function (RBF), clustering, supervised machine learning, semi-supervised machine learning, unsupervised machine learning, k -Nearest Neighbor algorithm (k-NN), linear regression, random forest, neural network (e.g., artificial neural network, recurrent neural network), etc.

[0120] Predictive component 514 may provide current data to model(s) 590 and may run model(s) 590 on the input to obtain one or more outputs. For example, predictive component 514 may provide current image data 566 to model(s) 590 and may run model(s) 590 on the input to obtain one or more outputs (e.g., current design correction data 576). Predictive component 514 may be capable of determining (e.g., extracting) current design correction data 576 from the output of model(s) 590. Predictive component 514 may determine (e.g., extract) confidence data from the output that indicates a level of confidence that current design correction data 576 is an accurate predictor of an update to substrate design parameters 550 (e.g., predicted substrate design parameters 550) for a substrate produced using the manufacturing equipment 524. Predictive component 514 or pattern update component 522 may use the confidence data to decide whether to cause a corrective action associated with the manufacturing equipment 524 based on current design correction data 576. The corrective action may include causing a pattern design to be updated, etc.Attorney Docket No. :39361.320 (L0150PCT)

[0121] The confidence data may include or indicate a level of confidence that the current design correction data 576 is an accurate prediction for products or components associated with at least a portion of the input data. In one example, the level of confidence is a real number between 0 and 1 inclusive, where 0 indicates no confidence that the current design correction data 576 is an accurate prediction for products processed according to input data or component health of components of manufacturing equipment 524 and 1 indicates absolute confidence that the current design correction data 576 accurately predicts properties of products processed according to input data or component health of components of manufacturing equipment 524. Responsive to the confidence data indicating a level of confidence below a threshold level for a predetermined number of instances (e.g., percentage of instances, frequency of instances, total number of instances, etc.) predictive component 514 may cause trained model(s) 590 to be re-trained (e.g., based on current image data 564, current manufacturing parameters, etc.). In some embodiments, retraining may include generating one or more data sets (e.g., via data set generator 572) utilizing historical data.

[0122] For purpose of illustration, rather than limitation, aspects of the disclosure describe the training of one or more machine learning models 590 using historical data (e.g., historical sensor data, historical manufacturing parameters) and inputting current data (e.g., current sensor data, current manufacturing parameters, and current image data) into the one or more trained machine learning models to determine current design correction data 576. In other embodiments, a heuristic model, physics-based model, or rule-based model is used to determine current design correction data 576. In some embodiments, such models may be trained using historical data. Predictive component 514 may monitor historical image data 564, historical manufacturing parameters, historical design correction data 574, and substrate image data 560.

[0123] In some embodiments, the functions of client device 520, predictive server 512, server machine 570, and server machine 580 may be provided by a fewer number of machines. For example, in some embodiments server machines 570 and 580 may be integrated into a single machine, while in some other embodiments, server machine 570, server machine 580, and predictive server 512 may be integrated into a single machine. In some embodiments, client device 520 and predictive server 512 may be integrated into a single machine. In some embodiments, functions of client device 520, predictive server 512, server machine 570, server machine 580, and data store 540 may be performed by a cloudbased service.Attorney Docket No. :39361.320 (L0150PCT)

[0124] In general, functions described in one embodiment as being performed by client device 520, predictive server 512, server machine 570, and server machine 580 can also be performed on predictive server 512 in other embodiments, if appropriate. In addition, the functionality attributed to a particular component can be performed by different or multiple components operating together. For example, in some embodiments, the predictive server 512 may determine the corrective action based on the defect classification data 568. In another example, client device 520 may determine the current design correction data 576 based on output from the trained machine learning model.

[0125] In addition, the functions of a particular component can be performed by different or multiple components operating together. One or more of the predictive server 512, server machine 570, or server machine 580 may be accessed as a service provided to other systems or devices through appropriate application programming interfaces (API).

[0126] In embodiments, a “user” may be represented as a single individual. However, other embodiments of the disclosure encompass a “user” being an entity controlled by a plurality of users and / or an automated source. For example, a set of individual users federated as a group of administrators may be considered a “user.”

[0127] FIG. 6 is a block diagram illustrating system 600 for generating output data (e.g., current design correction data 576 of FIG. 5), according to some embodiments. In some embodiments, system 600 may be used in conjunction with a machine learning model configured to output information about updates to a substrate design parameters and / or predicted substrate design parameters (e.g., model(s) 590 of FIG. 5). In some embodiments, system 600 may be used in conjunction with a machine learning model to determine an update to a substrate pattern design (e.g., an update to a die pattern design) and / or to determine a predicted substrate pattern design (e.g., a predicted die pattern design).

[0128] At block 610, system 600 (e.g., components of predictive system 510 of FIG. 5) performs data partitioning (e.g., via data set generator 572 of server machine 570 of FIG. 5) of data to be used in training, validating, and / or testing a machine learning model. In some embodiments, training design correction data 664 includes historical data, such as historical image data, historical design correction data, etc. In some embodiments, design correction data may be provided by a predictive system, e.g., predictive system 510 of FIG. 5. Training design correction data 664 may undergo data partitioning at block 610 to generate training set 602, validation set 604, and testing set 606. For example, the training set may be 60% of the training data, the validation set may be 20% of the training data, and the testing set may be 20% of the training data.Attorney Docket No. :39361.320 (L0150PCT)

[0129] The generation of training set 602, validation set 604, and testing set 606 may be tailored for a particular application. For example, the training set may be 60% of the training data, the validation set may be 20% of the training data, and the testing set may be 20% of the training data. System 600 may generate a plurality of sets of features for each of the training set, the validation set, and the testing set. Different models may be trained on different sets of data.

[0130] At block 612, system 600 performs model training (e.g., via training engine 582 of FIG. 5) using training set 602. Training of a machine learning model and / or of a physicsbased model (e.g., a digital twin) may be achieved in a supervised learning manner, which involves providing a training dataset including labeled inputs through the model, observing its outputs, defining an error (by measuring the difference between the outputs and the label values), and using techniques such as deep gradient descent and backpropagation to tune the weights of the model such that the error is minimized. In many applications, repeating this process across the many labeled inputs in the training dataset yields a model that can produce correct output when presented with inputs that are different than the ones present in the training dataset. In some embodiments, training of a machine learning model may be achieved in an unsupervised manner, e.g., labels or classifications may not be supplied during training. An unsupervised model may be configured to perform anomaly detection, result clustering, etc.

[0131] For each training data item in the training dataset, the training data item may be input into the model (e.g., into the machine learning model). The model may then process the input training data item (e.g., an image of a substrate etc.) to generate an output. The output may include, for example, information defects of the substrate (e.g., a characterization of the substrate defects, one or more matches to historical defects, etc.). The output may be compared to a label of the training data item (e.g., information generated by another reliable method).

[0132] Processing logic may then compare the generated output (e.g., substrate defect information, die defect information, etc.) to the label (e.g., labeled substrate information) that was included in the training data item. Processing logic determines an error (i.e., a classification error) based on the differences between the output and the label(s). Processing logic adjusts one or more weights and / or values of the model based on the error.

[0133] In the case of training a neural network, an error term or delta may be determined for each node in the artificial neural network. Based on this error, the artificial neural network adjusts one or more of its parameters for one or more of its nodes (the weights for one orAttorney Docket No. :39361.320 (L0150PCT) more inputs of a node). Parameters may be updated in a back propagation manner, such that nodes at a highest layer are updated first, followed by nodes at a next layer, and so on. An artificial neural network contains multiple layers of “neurons”, where each layer receives as input values from neurons at a previous layer. The parameters for each neuron include weights associated with the values that are received from each of the neurons at a previous layer. Accordingly, adjusting the parameters may include adjusting the weights assigned to each of the inputs for one or more neurons at one or more layers in the artificial neural network.

[0134] At block 614, system 600 performs model validation (e.g., via validation engine584 of FIG. 5) using the validation set 604. The system 600 may validate each of the trained models using a corresponding set of features of the validation set 604. Responsive to determining that one or more of the trained models has an accuracy that meets a threshold accuracy, flow continues to block 616.

[0135] At block 616, system 600 may perform model selection (e.g., via selection engine585 of FIG. 5) to determine which of the one or more trained models that meet the threshold accuracy has the highest accuracy (e.g., the selected model 608, based on the validating of block 614). Responsive to determining that two or more of the trained models that meet the threshold accuracy have the same accuracy, flow may return to block 612 where the system 600 performs model training using further refined training sets corresponding to further refined sets of features for determining a trained model that has the highest accuracy.

[0136] At block 618, system 600 performs model testing (e.g., via testing engine 586 of FIG. 5) using testing set 606 to test selected model 608. System 600 may test the first trained model to determine the first trained model meets a threshold accuracy. Determining whether the first trained model meets a threshold accuracy may be based on the first set of features of testing set 606. Responsive to accuracy of the selected model 608 not meeting the threshold accuracy, flow continues to block 612 where system 600 performs model training (e.g., retraining) using different training sets corresponding to different sets of features. Accuracy of selected model 608 may not meet threshold accuracy if selected model 608 is overly fit to the training set 602 and / or validation set 604. Responsive to determining that selected model 608 has an accuracy that meets a threshold accuracy based on testing set 606, flow continues to block 620. In at least block 612, the model may learn patterns in the training data to make classifications. In block 618, the system 600 may apply the model on the remaining data (e.g, testing set 606) to test the classifications.Attorney Docket No. :39361.320 (L0150PCT)

[0137] At block 620, system 600 uses the trained model (e.g., selected model 608) to receive current data 622 and determines (e.g., extracts), from the output of the trained model, design correction data 624. Current data 622 may be data related to a substrate having a defect, in some embodiments. Current data 622 may be image data of at least a portion of a substrate of interest, in some embodiments. Current data 622 may be textual data associated with a substrate of interest. A corrective action associated with the manufacturing equipment 524 of FIG. 5 may be performed in view of design correction data 624. In some embodiments, current data 622 may correspond to the same types of features in the historical data used to train the machine learning model such as metrology data, image data, and / or textual data, etc.

[0138] FIG. 7A-7B are simplified process flow diagrams for inspecting substrates using an example multi-substrate inspection system, in accordance with some embodiments of the present disclosure. Referring to FIG. 7A, a simplified process flow 700A is shown for inspecting substrates. In some embodiments, multiple substrates 702 are affixed to a chuck 715. Chuck 715 may be a vacuum chuck or an electrostatic chuck of a multi-substrate platform (e.g., multi-substrate platform 230). In some embodiments, chuck 715 supports up to four substrates 702. The substrates 702 may be supported beneath imaging devices 722. Each of the imaging devices 722 may be configured to capture an image of a respective substrate 799. For example, imaging device 722A may capture an image of substrate 799A, imaging device 722B may capture an image of substrate 799B, imaging device 722C may capture an image of substrate 799C, and imaging device 722D may capture an image of substrate 799D. In some embodiments, each of the imaging devices 722 may be configured to capture images of dies formed on a respective substrate 799.

[0139] In some embodiments, each of the imaging devices 722 include an emitter to emit one or more of infrared light, violet light, ultraviolet light, and / or visible light. The wavelength of light may be selected to illuminate certain features and / or defects. Infra-red light may be for detecting cracks in a die and / or voids under a die. Violet and / or ultraviolet light may be for detecting organic material residue. Visible light may be for detecting other defects, such as for determining basic pattern correctness (e.g., the correctness of a printed pattern, deposited pattern, or etched pattern, etc.).

[0140] A data collection module 730 may collect data from the imaging devices 722. In some embodiments, data collection module 730 collects image data indicative of the captured image(s) of the substrates and / or data indictive of the captured image(s) of the dies formed on the substrates. The data collection module 730 may organize the collected data, e.g., such asAttorney Docket No. :39361.320 (L0150PCT) for preprocessing by data preprocessing module 740. In some embodiments, data preprocessing module 740 receives image data from the data collection module 730 and performs one or more preprocessing operations. The one or more preprocessing operations can include image processing operations such as grayscale conversion, image resizing, cropping, thresholding, contrast adjustment, brightness adjustment, sharpening, smoothing / blurring, filtering, scaling, segmenting, corner detection, shape detection, etc.

[0141] In some embodiments, the preprocessed image data is provided to the image comparison module 750. The image comparison module 750 may compare the received preprocessed image data (e.g., indicative of one or more images of substrates 702, etc.) to reference image data. For example, and in some embodiments, the image comparison module 750 compares an image of a die placed on a substrate 799to a reference image corresponding to the die. The image comparison module 750 may determine whether the image of the die placed on the substrate 799 is sufficiently similar to the reference image (e.g., whether the die meets a similarity threshold criterion, etc.). The reference image may be generated by one or more systems and / or methods described herein. In some embodiments, the threshold criterion corresponds to whether each pad of the die has a good electrical signal connection.

[0142] Referring to FIG. 7B, a flow chart of an example method of process flow 700B is shown. Atblock 790, four substrates are loaded onto a chuck (e.g., chuck 715). The chuck may be disposed at a loading position (e.g., a loading area) while loaded with substrates. At block 792, the chuck is moved to an inspection position (e.g., inspection area, etc.). The chuck may be moved between the loading position and the inspection position by one or more actuators (e.g., linear actuators, pneumatic actuators, electro-mechanical actuators, etc.). At block 793, the optical wavelength is selected for illuminating the substrates on the chuck. The wavelength of light may be selected to illuminate certain features and / or defects. Infrared light may be for detecting cracks in a die and / or voids under a die. Violet and / or ultraviolet light may be for detecting organic material residue. Visible light may be for detecting other defects, such as for determining basic pattern correctness (e.g., the correctness of a printed, deposited, or etched pattern, etc.). In some embodiments, the wavelength may be selected by a trained machine learning model. More details regarding use and / or training of a machine learning model for selection of the optical wavelength are described herein with respect to FIGS. 7C-7E.

[0143] At block 794, image data is captured for each of the four substrates on the chuck. Image data may be captured for one or more dies formed on each of the substrates. The substrates may be illuminated using the selected optical wavelength (e.g., selected at blockAttorney Docket No. :39361.320 (L0150PCT)793) and an image captured (e.g., by imaging devices 722) concurrently for each of the substrates. The captured images may be collected and prepared for preprocessing. At block 796, image data (e.g., indicative of the captured images) is preprocessed. Various preprocessing operations may be performed. For example, preprocessing operations can include image processing operations such as grayscale conversion, image resizing, cropping, thresholding, contrast adjustment, brightness adjustment, sharpening, smoothing / blurring, filtering, scaling, segmenting, corner detection, shape detection, etc. At block 798, the images are uploaded (e.g., to an image comparison module 750) for image comparison. The images may be compared to one or more reference images generated according to method(s) described herein.

[0144] FIG. 7C is a flow diagram of a method 700C for generating a data set for training a machine learning model to select an optical wavelength for illuminating one or more substrates for imaging, in accordance with some embodiments of the present disclosure.

[0145] In some embodiments, at block 701 the processing logic implementing method 700C initializes a training set T to an empty set.

[0146] At block 702, processing logic generates first data input (e.g., first training input, first validating input) that may include wavelength range data (e.g., indicative of one or more ranges of wavelengths, etc.) and / or substrate data, etc. In some embodiments, the first data input may include a first set of features for types of data and a second data input may include a second set of features for types of data.

[0147] At block 703, processing logic generates a first target output for one or more of the data inputs (e.g., first data input). In some embodiments, the first target output is a selected optical wavelength (e.g., corresponding to infra-red light, violet light, ultraviolet light, or visible light, etc.).

[0148] At block 704, processing logic optionally generates mapping data that is indicative of an input / output mapping. The input / output mapping (or mapping data) may refer to the data input (e.g., one or more of the data inputs described herein), the target output for the data input, and an association between the data input(s) and the target output.

[0149] At block 705, processing logic adds the mapping data generated at block 704 to data set T, in some embodiments.

[0150] At block 706, processing logic branches based on whether data set T is sufficient for at least one of training, validating, and / or testing machine learning model 590 of FIG. 5. If so, execution proceeds to block 707, otherwise, execution returns to block 702. It should be noted that in some embodiments, the sufficiency of data set T may be determined basedAttorney Docket No. :39361.320 (L0150PCT) simply on the number of inputs, mapped in some embodiments to outputs, in the data set, while in some other implementations, the sufficiency of data set T may be determined based on one or more other criteria (e.g., a measure of diversity of the data examples, accuracy, etc.) in addition to, or instead of, the number of inputs.

[0151] At block 707, processing logic provides data set T (e.g., to server machine 580 of FIG. 5) to train, validate, and / or test machine learning model 590. In some embodiments, data set T is a training set and is provided to training engine 582 of server machine 580 to perform the training. In some embodiments, data set T is a validation set and is provided to validation engine 584 of server machine 580 to perform the validating. In some embodiments, data set T is a testing set and is provided to testing engine 588 of server machine 580 to perform the testing.

[0152] Over many training sets, the machine learning model chooses an appropriate value of a parameter controlling the number of non-zero coefficients in the reduced order model. After block 707, machine learning model (e.g., machine learning model 590) can be at least one of trained using training engine 582 of server machine 580, validated using validating engine 584 of server machine 580, ortested using testing engine 588 of server machine 580. The trained machine learning model may be implemented by predictive component 514 (of predictive server 512) to generate selected optical wavelength data for inspecting one or more substrates as described herein.

[0153] FIG. 7D is a method 700D for training a machine learning model (e.g., model 590 of FIG. 5) to select an optical wavelength for inspecting one or more substrates, in accordance with some embodiments of the present disclosure.

[0154] Referringto FIG. 7D, atblock 710 of method 700D, one or more emitters of an inspection tool emit at least one of infra-red light, violet light, ultraviolet light, or visible light having a selected wavelength toward multiple substrates. Multiple emitters may emit the light at a corresponding one of the multiple substrates.

[0155] At block 712, image data is captured indicative of the multiple substrates. The image data may be indicative of one or more dies placed on the multiple substrates. Multiple imaging devices may be used to capture the image data. In some embodiments, each of the multiple imaging devices may capture an image of a corresponding substrate.

[0156] At block 714, processing logic trains a machine learning model with training input data including historical wavelength range data and historical data indicative of historical substrates. The historical data may be indicative of substrate type and / or substrate defects, such as cracks, voids, organic material residue, and / or basic pattern correctness, etc. TheAttorney Docket No. :39361.320 (L0150PCT) machine learning model may be trained with training output data including selected historical wavelength data. The selected historical wavelength data may be indicative of selected historical light wavelengths used to illuminate historical substrates for inspection, such as to detect cracks, voids, organic material residue, and / or basic pattern correctness, etc. In some embodiments, different types of products / substrates may be subject to different types of defects. Each wavelength may be better at detecting different types of defects. The training data may indicate, for each historical data entry, a product / substrate type, wavelengths of light used to perform defect detection, and / or defects detected. Once trained, the machine learning model may be used to select a wavelength of inspection light used for substrate inspection.

[0157] In some embodiments, more than one machine learning model can be parts of a single compound machine learning model. In this case, training one component of this compound model may involve receiving output from another component of the model as training input to the component of the model to be trained.

[0158] At block 716, the machine learning model can be retrained, using additional data. The machine learning model can be retrained with training input data including wavelength range data and data indicative of the multiple substrates and / or training output data including selected wavelength data. In some embodiments, the machine learning model can be continually trained to account for drift in the manufacturing equipment, sensors, metrology equipment, etc., or to reflect changes to procedures or recipes, etc.

[0159] FIG. 7E is a method 700E for using a trained machine learning model (e.g., model 590 of FIG. 5) to select an optical wavelength for inspecting one or more substrates, in accordance with some embodiments of the present disclosure.

[0160] At block 770 of method 700E, processing logic receives data indicative of a wavelength range of optical sensors of an inspection tool. The wavelength range may correspond to the wavelength at which the optical sensors can operate (e.g., one or more of the infra-red range, the violet range, the ultraviolet range, the visible range, etc.).

[0161] At block 771, processing logic receives data indicative of multiple substrates on a chuck of the inspection tool. The data may be indicative of the type of substrate and / or the type of one or more expected defects, etc. (e.g., cracks, voids, organic matter, pattern correctness, etc.). The data may be indicative of one or more dies formed on the multiple substrates, such as die type, die pattern design, expected die defects, etc.

[0162] At block 772, processing logic inputs, into a trained machine learning model, the data indicative of the wavelength range and the data indicative of the multiple substrates.Attorney Docket No. :39361.320 (L0150PCT)

[0163] At block 773, processing logic obtains a selected wavelength as output from the trained machine learning model. The selected wavelength is within the wavelength range of the optical sensors.

[0164] At block 774, emitters of multiple imaging devices (e.g., of the inspection tool) emit light having the selected wavelength toward the multiple substrates. One or more defects may be illuminated by the selected wavelength of light.

[0165] At block 775, image data of the multiple substrates is captured simultaneously (e.g., substantially simultaneously, concurrently, etc.) by the optical sensors. The image data may be indicative of one or more defects of the substrates. In some embodiments, image data of one or more dies formed on the multiple substrates is captured simultaneously.

[0166] FIG. 8 is a flow diagram of an example method 800 for inspecting dies placed on substrates, in accordance with some embodiments of the present disclosure. Method 800 may be performed by processing logic that may include hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (such as instructions run on a processing device, a general purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof.

[0167] At least a portion of method 800 may be performed by predictive server 512 (e.g., predictive component 514) and / or server machine 580 (e.g., training, validating, and testing operations may be performed by server machine 580). In some embodiments, a non-transitory machine-readable storage medium stores instructions that when executed by a processing device (e.g., of predictive system 510, of server machine 580, of predictive server 512, etc.) cause the processing device to perform method 800.

[0168] For simplicity of explanation, method 800 is depicted and described as a series of operations. However, operations in accordance with this disclosure can occur in various orders and / or concurrently and with other operations not presented and described herein. Furthermore, not all illustrated operations may be performed to implement method 800 in accordance with the disclosed subject matter. In addition, those skilled in the art will understand and appreciate that method 800 could alternatively be represented as a series of interrelated states via a state diagram or events.

[0169] At block 802 of method 800, a system receives, by a load port of the system, one or more substrate carriers. The one or more substrate carriers may be docked at the load port. At block 804, multiple substrate are transferred, by a substrate-handling robot, from the one or more substrate carriers to a chuck. The substrate-handling robot may be operable to place the multiple substrates on the chuckwhen the chuckis in a loading position. The chuck mayAttorney Docket No. :39361.320 (L0150PCT) be a multi-substrate chuck. For example, the chuckmay be capable of handling two or more substrates. In some examples, the chuck is configured to concurrently support four or more substrates. The chuck may be a vacuum chuck, an electrostatic chuck, or another kind of chuck capable of securing substrates. At block 806, the multiple substrates are flattened on the chuck by a substrate flattening unit. In some embodiments, the substrate flattening unit exerts a downward force on the substrates to flatten the substrates on the chuck. At block 808, the chuck is moved by one or more actuators from a loading position (where the substrates are loaded onto the chuck) to an inspection position. While in the loading position, the chuck may be loaded with one or more substrates. While in the inspection position, the one or more substrates on the chuck may be inspected, such as by one or more imaging devices of an inspection tool (e.g., a multi-substrate inspection tool), etc.

[0170] At block 810, an inspection tool receivesthe multiple substrates on the chuck. The inspection tool may be a multi-substrate inspection tool. The inspection tool may be operable to inspect the multiple substrates (e.g., dies formed on the multiple substrates) on the chuck when the chuck is in the inspection position. The chuck may be positioned so that each of the substrates is beneath a corresponding imaging device of the inspection tool. In some embodiments, the inspection tool includes two or more image sensors (e.g., imaging devices). In some embodiments, the inspection tool includes four or more image sensors (e.g., four or more imaging devices). At block 812, image data of the multiple substrates is simultaneously captured by multiple imaging devices. In some embodiments, image data of dies formed on the multiple substrates is simultaneously captured by the multiple imaging devices. In some embodiments, the inspection tool includes an imaging device for each of the substrates on the chuck. Each of the imaging devices may simultaneously capture an image of the corresponding substrate. Each of the imaging devices may simultaneously capture an image of one or more dies formed on the corresponding substrate. In some embodiments, images of the substrate (e.g., images of one or more dies formed on the substrate, etc.) may be captured concurrently, such as during the same period of time that the chuck is in the inspection position. In some embodiments, the substrates are illuminated on the chuck by one or more emitters of the imaging devices. In some embodiments, an emitter of each of the imaging devices emits at least one of infra-red light, violet light, ultraviolet light, or visible light to illuminate the substrates. The wavelength of emitted light may be selected using a trained machine learning model as described herein above. A different wavelength of light may be emitted depending on the purpose of the inspection. For example, infra-red light may be for detecting cracks in a die and / or voids under a die. Violet and / or ultraviolet light may be forAttorney Docket No. :39361.320 (L0150PCT) detecting organic material residue. Visible light may be for detecting other defects, such as for determining basic pattern correctness (e.g., the correctness of a printed pattern, deposition pattern, or etched pattern, etc.).

[0171] At block 816, a processing device determines one or more defects of the multiple substrates based on the captured image data. The processing logic may determine one or more defects of one or more dies formed on the multiple substrates based on the captured image data. The captured image data may be analyzed for die defects and / or compared to one or more reference images. A defect may include a mismatch between a target pattern and a formed pattern on a substrate. A corrective action may be performed based on a detected defect. In some embodiments, the captured image data may be used for image comparison.

[0172] FIG. 9 is a block diagram illustrating a computer system 900, in accordance with some embodiments of the present disclosure. In some embodiments, computer system 900 may be connected (e.g., via a network, such as a Local Area Network (LAN), an intranet, an extranet, or the Internet) to other computer systems. Computer system 900 may operate in the capacity of a server or a client computer in a client-server environment, or as a peer computer in a peer-to-peer or distributed network environment. Computer system 900 may be provided by a personal computer (PC), a tablet PC, a Set-Top Box (STB), a Personal Digital Assistant (PDA), a cellular telephone, a web appliance, a server, a network router, switch or bridge, or any device capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by that device. Further, the term "computer" shall include any collection of computers that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods described herein.

[0173] In a further aspect, the computer system 900 may include a processing device 902, a volatile memory 904 (e.g., Random Access Memory (RAM)), a non-volatile memory 906 (e.g., Read-Only Memory (ROM) or Electrically -Erasable Programmable ROM (EEPROM)), and a data storage device 918, which may communicate with each other via a bus 908.

[0174] Processing device 902 may be provided by one or more processors such as a general purpose processor (such as, for example, a Complex Instruction Set Computing (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, a Very Long Instruction Word (VLIW) microprocessor, a microprocessor implementing other types of instruction sets, or a microprocessor implementing a combination of types of instruction sets) or a specialized processor (such as, for example, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), or a network processor).Attorney Docket No. :39361.320 (L0150PCT)

[0175] Computer system 900 may further include a network interface device 922 (e.g., coupled to network 974). Computer system 900 also may include a video display unit 910 (e.g., an LCD), an alphanumeric input device 912 (e.g., a keyboard), a cursor control device 914 (e.g., a mouse), and a signal generation device 920.

[0176] In some embodiments, data storage device 918 may include a non-transitory computer-readable 924 (e.g., non-transitory machine-readable storage medium storing instructions) on which may store instructions 926 encoding any one or more of the methods or functions described herein, including instructions encoding components of FIG. 5 (e.g., predictive component 514, pattern update component 522, model(s) 590, etc.) and for implementing methods described herein.

[0177] Instructions 926 may also reside, completely or partially, within volatile memory 904 and / or within processing device 902 during execution thereof by computer system 900, hence, volatile memory 904 and processing device 902 may also constitute machine-readable storage media.

[0178] While non-transitory computer-readable medium 924 is shown in the illustrative examples as a single medium, the term "computer-readable storage medium" shall include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store the one or more sets of executable instructions. The term "computer-readable storage medium" shall also include any tangible medium that is capable of storing or encoding a set of instructions for execution by a computer that cause the computer to perform any one or more of the methods described herein. The term "computer- readable storage medium" shall include, but not be limited to, solid-state memories, optical media, and magnetic media.

[0179] It should be understood that the above description is intended to be illustrative, and not restrictive. Many other embodiment examples will be apparent to those of skill in the art upon reading and understanding the above description. Although the present disclosure describes specific examples, it will be recognized that the systems and methods of the present disclosure are not limited to the examples described herein, but may be practiced with modifications within the scope of the appended claims. Accordingly, the specification and drawings are to be regarded in an illustrative sense rather than a restrictive sense. The scope of the present disclosure should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

[0180] The embodiments of methods, hardware, software, firmware or code set forth above may be implemented via instructions or code stored on a machine-accessible, machineAttorney Docket No. :39361.320 (L0150PCT) readable, computer accessible, or computer readable medium which are executable by a processing element. “Memory” includes any mechanism that provides (i.e., stores and / or transmits) information in a form readable by a machine, such as a computer or electronic system. For example, “memory” includes random-access memory (RAM), such as static RAM (SRAM) or dynamic RAM (DRAM); ROM; magnetic or optical storage medium; flash memory devices; electrical storage devices; optical storage devices; acoustical storage devices, and any type of tangible machine-readable medium suitable for storing or transmitting electronic instructions or information in a form readable by a machine (e.g., a computer).

[0181] Reference throughout this specification to “one embodiment” or “an embodiment’ means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosure. Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0182] In the foregoing specification, a detailed description has been given with reference to specific exemplary embodiments. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the disclosure as set forth in the appended claims. The specification and drawings are, accordingly, to be regarded in an illustrative sense rather than a restrictive sense.Furthermore, the foregoing use of embodiment, embodiment, and / or other exemplary language does not necessarily refer to the same embodiment or the same example, but may refer to different and distinct embodiments, as well as potentially the same embodiment.

[0183] The words “example” or “exemplary” are used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “example’ or “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects or designs. Rather, use of the words “example” or “exemplary” is intended to present concepts in a concrete fashion. As usedin this application, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless specified otherwise, or clear from context, “X includes A or B” is intended to mean any of the natural inclusive permutations. That is, if X includes A; X includes B; or X includes both A and B, then “X includes A orB” is satisfied under any of the foregoing instances. In addition, the articles “a” and “an” as used in this application and the appended claims should generally be construed toAttorney Docket No. :39361.320 (L0150PCT) mean “one or more” unless specified otherwise or clear from context to be directed to a singular form. Moreover, use of the term “an embodiment” or “one embodiment” or “an embodiment” or “one embodiment” throughout is not intended to mean the same embodiment or embodiment unless described as such. Also, the terms “first,” “second,” “third,” “fourth,” etc. as used herein are meant as labels to distinguish among different elements and may not necessarily have an ordinal meaning according to their numerical designation.

Claims

Attorney Docket No. :39361.320 (L0150PCT)CLAIMSWhat is claimed is:

1. A system comprising: a multi-substrate platform comprising a plurality of substrate holders, each substrate holder of the plurality of substrate holders to hold a respective substrate of a plurality of substrates; an inspection system comprising a plurality of inspection heads to perform, concurrently, an inspection of the plurality of substrates, wherein each inspection head of the plurality of inspection heads is to collect data for the respective substrate of the plurality of substrates; and a processing device to identify, using the collected data, at least one of (i.) presence of one or more defects in at least one substrate of the plurality of substrates or (ii.) one or more measurements of one or more properties of at least one substrate of the plurality of substrates.

2. The system of claim 1, further comprising a robot to: prior to the inspection, place the plurality of substrates on the multi-substrate platform; and after the inspection, remove the plurality of substrates from the multi-substrate platform.

3. The system of claim 2, wherein the robot concurrently places the plurality of substrates on the multi-substrate platform.

4. The system of claim 1, further comprising a plurality of buffer stations to: hold the plurality of substrates prior to the inspection; and receive the plurality of substrates after the inspection.

5. The system of claim 1, further comprising a rail to transport the multi-substrate platform to an inspection area.

6. The system of claim 1, wherein the plurality of substrate holders comprises at least (i) one or more electrostatic chucks, or (ii) one or more vacuum suction chucks.Attorney Docket No. :39361.320 (L0150PCT)7. The system of claim 1, further comprising one or more aligner devices to: identify, while the inspection system performs the inspection of a preceding plurality of substrates, a reference feature for each of the plurality of substrates; and align, using the identified reference features, the plurality of substrates for the inspection.

8. The system of claim 1, wherein an individual inspection head of the plurality of inspection heads collects at least one of: a bright-field inspection data for the respective substrate, a dark-field inspection data for the respective substrate, a scanning electron microscopy (SEM) imaging data for the respective substrate, an atomic force microscopy (AFM) imaging data for the respective substrate, or a scanning acoustic microscopy (SAM) imaging data for the respective substrate.

9. The system of claim 1, wherein at least one inspection head of the plurality of inspection heads collects both: a bright-field inspection data for the respective substrate, and a dark-field inspection data for the respective substrate.

10. The system of claim 1, wherein the inspection system deploys at least one of: a visible light, an infrared (IR) light, an ultraviolet (UV) light, an extreme ultraviolet (EUV) light, an x-ray radiation, an acoustic wave, an acoustic excitation, or an electron beam excitation.

11. A system comprising: a platform comprising a plurality of chucks, each chuck to hold a substrate of a plurality of substrates; one or more aligner devices to identify reference features of each of the plurality of substrates;Attorney Docket No. :39361.320 (L0150PCT) a robot comprising a plurality of arms, each arm to place a respective substrate of the plurality of substrates on a respective chuck of the plurality of chucks; and a plurality of inspection heads, an individual inspection head to perform an inspection of a corresponding substrate of the plurality of substrates, wherein the inspections of the plurality of substrates are performed concurrently.

12. The system of claim 11, further comprising: a plurality of substrate buffers to: temporarily store the plurality of substrates prior to the inspection; and receive the plurality of substrates after the inspection.

13. The system of claim 11, system of claim 1, wherein to perform the inspection, the individual inspection head is to: direct an incident light to the corresponding substrate; collect light generated, in response to the incident light, by the corresponding substrate; and detect, using one more light detectors, intensity of the generated light for a plurality of angles of propagation of the generated light.

14. The system of claim 11, wherein the individual inspection head is further to generate an inspection data for the corresponding substrate, the system further comprising: a processing device to determine, using the inspection data for the corresponding substrate, at least one of (i.) presence of one or more defects of the corresponding substrate or (ii.) one or more measurements of one or more properties of the corresponding substrate.

15. A method comprising: supporting, using a multi-substrate platform, a plurality of substrates, wherein each substrate of the plurality of substrates is supported by a respective substrate holder of a plurality of substrate holders; collecting, in parallel, inspection data for the plurality of substrates, wherein the inspection data for each substrate of the plurality of substrates is collected using a respective inspection head of a plurality of inspection heads; and identifying, by a processing device and using the collected inspection data for the plurality of substrates, at least one of (i.) presence of one or more defects in at least oneAttorney Docket No. :39361.320 (L0150PCT) substrate of the plurality of substrates or (ii.) one or more measurements of one or more properties of at least one substrate of the plurality of substrates.

16. The method of claim 15, wherein supporting the plurality of substrates comprises: prior to collecting the inspection data, concurrently placing, using a robot, the plurality of substrates on the multi-substrate platform; and after collecting the inspection data, concurrently removing, using the robot, the plurality of substrates from the multi-substrate platform.

17. The method of claim 15, wherein supporting the plurality of substrates comprises: transporting, prior to collecting the inspection data, the multi-substrate platform to an inspection area.

18. The method of claim 15, further comprising: identifying, using one or more aligner devices and while the inspection data is being collected for a preceding plurality of substrates, a reference feature for each of the plurality of substrates; and aligning, using the identified reference features, the plurality of substrates.

19. The method of claim 15, wherein the inspection data comprises at least one of: b right-field inspection data, dark -field inspection data, scanning electron microscopy (SEM) imaging data, atomic force microscopy (AFM) imaging data, or scanning acoustic microscopy (SAM) imaging data.

20. The method of claim 15, wherein collecting the inspection data comprises, for each of the plurality of substrates: directing an incident light to a corresponding substrate; collecting light generated, in response to the incident light, by the corresponding substrate; and detecting, using one more light detectors, intensity of the generated light for a plurality of angles of propagation of the generated light.Attorney Docket No. :39361.320 (L0150PCT)21. A substrate inspection system, comprising: a chuck configured to secure multiple substrates for inspection; an inspection tool, comprising multiple imaging devices configured to capture image data of the multiple substrates simultaneously; and a processing device configured to: receive the image data from the multiple imaging devices; and determine one or more defects of the multiple substrates based on the image data.

22. The substrate inspection system of claim 21, wherein the chuck is configured to concurrently supportfour or more substrates, and wherein the inspection tool comprises four or more imaging devices.

23. The substrate inspection system of claim 21, further comprising: a load port configured to receive one or more substrate carriers; and a substrate-handling robot configured to transfer the multiple substrates from the one or more substrate carriers to the chuck.

24. The substrate inspection system of claim 23, further comprising: an aligner station configured to align each of the multiple substrates, wherein the substrate-handling robot is further configured to place each of the multiple substrates at the aligner station for alignment.

25. The substrate inspection system of claim 23, further comprising: one or more actuators configured to move the chuck between a loading position and an inspection position, wherein the substrate-handling robot is operable to place the multiple substrates on the chuck when in the loading position, and wherein the inspection tool is operable to inspect the multiple substrates on the chuck when in the inspection position.

26. The substrate inspection system of claim 23, wherein the substrate-handling robot is configured to handle two or more substrates simultaneously.Attorney Docket No. :39361.320 (L0150PCT)27. The substrate inspection system of claim 23, further comprising: one or more buffer stations configured to store a substrate for loading by the substrate-handling robot onto the chuck.

28. The substrate inspection system of claim 21, wherein the multiple imaging devices each comprises one or more emitters configured to emit at least one of infra-red light, violet light, ultraviolet light, or visible light.

29. The substrate inspection system of claim 21, further comprising: a substrate flattening unit configured to apply a downward force to a top surface of the multiple substrates to flatten the multiple substrates on the chuck.

30. The substrate inspection system of claim 21 , wherein the chuck comprises a vacuum chuck or an electrostatic chuck.

31. A substrate inspection tool, comprising: a support structure; multiple image sensors supported by the support structure above an inspection zone, wherein each of the multiple image sensors are configured to simultaneously capture image data of a respective substrate; and a processing device configured to determine one or more defects of one or more of the respective substrates based on corresponding image data.

32. The substrate inspection tool of claim 31, comprising four or more image sensors configured to simultaneously capture image data of four or more substrates.

33. The substrate inspection tool of claim 32, wherein the support structure is configured to support the multiple image sensors above a chuck disposed in the inspection zone, and wherein the chuck is configured to secure the four or more substrates for inspection.

34. The substrate inspection tool of claim 31, wherein the multiple image sensors each comprise one or more emitters configured to emit at least one of infra-red light, violet light, ultraviolet light, or visible light.Attorney Docket No. :39361.320 (L0150PCT)35. A method, comprising: receiving multiple substrates on a chuck; capturing, by multiple imaging devices, image data of the multiple substrates simultaneously; and determining, by a processing device, one or more defects of the multiple substrates based on the image data.

36. The method of claim 35, wherein four or more substrates are received on the chuck, and wherein four or more imaging devices capture the image data of the four or more substrates.

37. The method of claim 35, further comprising: receiving, by a load port, one or more substrate carriers; and transferring, by a substrate-handling robot, the multiple substrates from the one or more substrate carriers to the chuck.

38. The method of claim 37, further comprising: moving, by one or more actuators, the chuck between a loading position and an inspection position, wherein the substrate-handling robot is operable to place the multiple substrates on the chuck when in the loading position, and wherein the multiple imaging devices are operable to capture image data of the multiple substrates on the chuck when in the inspection position.

39. The method of claim 35, further comprising: emitting, by the multiple imaging devices, at least one of infra-red light, violet light, ultraviolet light, or visible light.

40. The method of claim 39, further comprising: inputting, into a trained machine learning model, data indicative of a wavelength range and data indicative of the multiple substrates, wherein the trained machine learning model is trained with training input data comprising historical wavelength range data and historical data indicative of historical substrates, and wherein the trained machine learning model is trained with training output data comprising selected historical wavelength data; andAttorney Docket No. :39361.320 (L0150PCT) receiving, from the trained machine learning model, selected wavelength data indicative of a selected wavelength within the wavelength range, wherein the wavelength range corresponds to the at least one of infra-red light, violet light, ultraviolet light, or visible light.

Images