Regional camera substrate pre-aligner

Abstract

Various examples include a substrate pre-aligner system that can align a substrate, determine an amount of arcuate curvature in the substrate, and determine other characteristics of the substrate by detecting a fiducial on the substrate. In one example, the pre-aligner of the disclosed subject matter may determine, for example, the location of a fiducial and arcuate bend in a substrate by imaging a 0-degree orientation and an orientation after a single 180-degree rotation of the substrate. In other embodiments, multiple cameras are used to capture images of the substrate substantially simultaneously, and determine, for example, the location of the fiducial and arcuate bends in the substrate. The multi-camera embodiments may also allow for higher substrate throughput compared to 0 degree to 180 degree embodiments. Other systems and methods are also disclosed.

Description

[0001] Claiming priority

[0002] This application claims priority to U.S. patent application serial number 17 / 840,154, filed on June 14, 2022, and entitled “AREA CAMERA SUBSTRATE PRE-ALIGNER,” which claims priority to U.S. provisional patent application serial number 63 / 325,924, filed on March 31, 2022, and entitled “AREA CAMERA SUBSTRATE PRE-ALIGNER,” each of which is incorporated herein by reference in its entirety. Technical Field

[0003] The disclosed subject matter generally relates to the field of substrate inspection and metrology tools used in semiconductor and related industries, such as flat panel display and solar cell production facilities. More specifically, in various embodiments, the disclosed subject matter relates to a substrate pre-aligner that can, for example, align a substrate, determine an amount of bow in a substrate, and determine other characteristics of a substrate. In various embodiments, the disclosed subject matter can align bonded wafers, such as a gallium arsenide (GaAs) wafer bonded to a transparent sapphire wafer. Background Art

[0004] Various types of substrates (such as semiconductor wafers) are typically placed in various types of production tools for processing and metrology tools for measurement within a manufacturing facility (e.g., such as an integrated circuit manufacturing facility). A robot is used to place the substrate on a substrate platform within the tool to prepare the substrate for processing within a processing chamber or measurement within a metrology tool. Typically, the substrate has one or more fiducial indicators etched or otherwise formed into the edge of the substrate (e.g., notches or flat portions on the edge of the substrate). The fiducial indicators are arranged along the substrate in a unique orientation to indicate a characteristic of the substrate, such as alignment along the crystallographic orientation of the substrate (e.g., the notches may be arranged within ±1° of a {110} orientation). Therefore, for various types of processing operations that may rely on a known crystallographic orientation, it is desirable to align the fiducial marks relative to the processing chamber in a particular orientation. Similar considerations may also exist for alignment of the substrate to the metrology tool. Summary of the invention

[0005] This document describes, inter alia, a substrate pre-aligner that can, for example, align a substrate to a fiducial, determine an amount of bow in a substrate, and determine other characteristics of the substrate.

[0006] In various embodiments, the disclosed subject matter is a device for detecting a reference on a substrate and pre-aligning it to a predetermined position on a substrate holder. The device includes at least one camera, and when the substrate is located on the substrate holder, the at least one camera will be positioned above the substrate. The at least one camera is configured to capture multiple images of at least the top surface of the substrate, and at least two of the multiple images show at least a partially unique view of the substrate. A hardware-based processor is configured to determine the position of the edge of the substrate based on the captured multiple images. The hardware-based processor is further configured to determine the characteristic size of the substrate across the top surface, determine the position of the reference on the substrate, calculate at least the x-direction offset, y-direction offset and theta-direction offset of the substrate relative to the predetermined position on the substrate holder, and determine the position of the reference based on the captured multiple images.

[0007] In various embodiments, the disclosed subject matter is a method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder. The method includes capturing at least one first image of at least a top surface of a substrate and capturing at least one second image of at least the top surface of the substrate, wherein the at least one first image and the at least one second image show at least a partially unique view of the substrate. Processing the at least one first image and the at least one second image to determine the position of the fiducial, wherein the processing also includes calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder.

[0008] In various embodiments, the disclosed subject matter is a computer-readable medium containing instructions that, when executed by a machine, cause the machine to perform operations including: receiving at least one first image of at least a top surface of a substrate and receiving at least one second image of at least a top surface of the substrate. The at least one first image and the at least one second image show at least a partially unique view of the substrate. Processing the at least one first image and the at least one second image to determine the position of the reference on the substrate. The processing also includes calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder, and determining the position of the reference based on the at least one first image and the at least one second image.

[0009] In various embodiments, the disclosed subject matter is a method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder. The method includes capturing at least one unique image of the substrate by each camera of a plurality of cameras after the substrate is placed on the substrate holder; processing each of the captured images; determining at least one edge of the substrate based on the processed images; determining the position of the fiducial on the substrate based on the processed images; and calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder based on the processed images. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The figures in the drawings depict only example implementations of the present disclosure and should not be considered limiting of its scope.

[0011] Figure 1A Bonded substrates useful for various embodiments of the disclosed subject matter are shown;

[0012] Figure 1B Shows Figure 1A a cross-sectional portion of a bonded substrate;

[0013] Figure 2A An example of a single camera pre-aligner system according to various embodiments of the disclosed subject matter is shown, which can be used to pre-align, for example, a substrate;

[0014] Figure 2B An example of a multi-camera pre-aligner system that may be used to pre-align a substrate according to various embodiments of the disclosed subject matter is shown;

[0015] Figure 3 A generalized example of a method for detecting a fiducial mark (e.g., a notch) on a substrate and detecting corrected coordinates of the substrate according to various embodiments of the disclosed subject matter is shown;

[0016] Figure 4 shows a generalized example of a method for detecting fiducial marks and other characteristics of a substrate using multiple cameras in accordance with various embodiments of the disclosed subject matter;

[0017] Figure 5 An example of a machine learning framework for detecting fiducial marks (e.g., notches) on a substrate and other desired features or characteristics of the substrate is shown in accordance with various embodiments of the disclosed subject matter; and

[0018] Figure 6 A block diagram is shown that includes an example of a machine on which any one or more of the techniques (eg, methodologies) discussed herein may be performed. DETAILED DESCRIPTION

[0019] The disclosed subject matter relates to a substrate pre-aligner that can align a substrate and determine the amount of bow in the substrate. In various embodiments, the disclosed subject matter can also align bonded wafers (e.g., a GaAs wafer bonded to a transparent sapphire wafer). Conventional pre-aligners are generally unable to perform these functions. For example, since wax is often used to bond the wafers in the case of bonded wafers, conventional pre-aligners are generally unable to distinguish between references, such as notches on a GaAs wafer, because the wax can protrude into the notches. In addition, by imaging both 0 degrees and a single 180 degree rotation of the substrate, the pre-aligner of the disclosed subject matter can determine and quantify the amount of bow in the substrate (if any).

[0020] In one example, the disclosed subject matter may use a color area camera. In an embodiment, the pre-aligner may be configured with a single camera that is placed above the surface of the substrate, viewing the entire substrate. In an embodiment, the pre-aligner may be configured with two cameras viewing opposite portions of the substrate, or four cameras viewing four quadrants of the substrate. In another example, two of the four cameras may be placed above the substrate, and the remaining two cameras may be placed below the substrate. As described below with reference to Figure 5 As discussed in more detail, the images captured by the one or more cameras can then be processed by a machine learning framework to detect fiducial marks on the substrate. The machine learning framework can also be used to determine other features or characteristics of the substrate, such as whether the substrate is bowed or warped.

[0021] As disclosed herein, image data can be processed using a machine learning framework such as a convolutional neural network (CNN or convnet). Processing image data includes, for example, finding spatial relationships within the captured image to determine features or characteristics of the substrate disclosed herein. A generalized example of a machine learning framework that can be used for the disclosed subject matter is described in more detail below.

[0022] The collected images are used as input to a machine learning framework that generates at least one output indicating, for example, the position of the fiducial, the amount of misalignment of the substrate on the substrate platform (e.g., misalignment in the x-direction, the y-direction, the z-direction, and theta-direction), and the amount of bow or warp in the substrate.

[0023] In an embodiment, one or more images of a substrate captured by the one or more cameras are processed in a machine learning framework to determine features or characteristics of the substrate (e.g., to distinguish the imaged substrate from wax and a carrier substrate (if present), to determine the location of one or more fiducials, to detect and quantify the amount of bow in the substrate (if present), and, if necessary, to detect corrected coordinates of the substrate.)

[0024] For example, now refer to Figure 1A , shows a bonded substrate 100 that can be used with various embodiments of the disclosed subject matter. Figure 1A It is shown to include a substrate 101, a carrier substrate 105, and a bonding agent 107. The bonding agent 107 is used to bond the substrate 101 to the carrier substrate 105. The substrate 101 is also shown to include an example of a reference notch 103. As described herein, the reference notch 103 can be used to indicate an orientation, for example, along a crystallographic orientation of the substrate 101.

[0025] Substrate 101 may include, for example, various types of substrates, such as wafers including elemental semiconductors (e.g., silicon or germanium), wafers including compound semiconductors (e.g., gallium arsenide (GaAs) or gallium nitride (GaN)), or various other substrate types known in the art (including conductive, semiconductive, and non-conductive substrates). Thus, substrate 101 may include, for example, any one or more of 400 mm, 300 mm, 200 mm, 150 mm, 125 mm, and 100 mm circular substrates (e.g., wafers), including opaque (e.g., at visible wavelengths), transparent, and bonded substrates. Each of the listed substrate types may include a notch and a single plane or a single plane and a sub-plane reference. In an embodiment, the material constituting carrier substrate 105 may include a material different from the material constituting substrate 101. In other embodiments, carrier substrate 105 may include the same material as substrate 101. In an embodiment, the size of carrier substrate 105 may be the same or similar to the size of substrate 101. In other embodiments, the size of carrier substrate 105 may be different from the size of substrate 101. In an embodiment, the shape of the carrier substrate 105 can be similar to or different from the shape of the substrate 101 .

[0026] The bonding agent 107 may include various types of wax to attach the substrate 101 to the carrier substrate 105. However, since wax is often used to bond various types of substrates 101, 105 to each other, the wax may protrude into the reference recess 103, as highlighted by the portion 110 of the bonded substrate 100 and will be discussed below with reference to FIG. Figure 1B Have a discussion.

[0027] Section 110 Figure 1B Shown in Figure 1A The cross-sectional portion of the bonding substrate 100. Figure 1BAs shown, the bonding agent 107 at least partially fills the reference recess 103 of the substrate 101. The bonding agent may also obscure one or more edges of the substrate 101 and / or the carrier substrate 105. Currently available pre-aligners are unable to distinguish the bonding agent 107 from the recess or other reference or edge of the substrate 101 and / or the carrier substrate 105. Therefore, conventional pre-aligners are generally unable to align at least the reference recess 103 to the processing system or metrology system into which the bonded substrate 100 is introduced. However, it is understood by those of ordinary skill in the art that many, for example, substrate processing systems require the substrate 101 to be in a predefined orientation before being placed within the processing system.

[0028] Figure 2A An example of a single camera pre-aligner system 200 is shown that can be used to pre-align a substrate according to various embodiments of the disclosed subject matter. The single camera pre-aligner system 200 can also be used to determine and quantify characteristics in the substrate, such as bow and warp (if present), and other characteristics of the substrate. The single camera pre-aligner system 200 can operate as a stand-alone pre-aligner system, a pre-aligner system that can be embedded in a processing and metrology tool, or a pre-aligner system that can be embedded in a robotic substrate handling system.

[0029] The single camera pre-aligner system 200 is shown to include an overhead camera 201 having a substantially covering substrate 203 (eg, Figure 1A The field of view 207 of the uppermost area of ​​the bonded substrate 100 or any bonded substrate or unbonded substrate). In various embodiments, the translation platform 205 is configured to move the substrate 203 in, for example, the x-direction, the y-direction, the z-direction, and the θ-direction (rotating the substrate 203 as indicated by the arrows around the translation platform 205). In various embodiments, the translation platform 205 is configured to move the substrate 203 only in, for example, the θ-direction. The translation platform 205 can be activated (rotated) using, for example, a brushless DC motor. A rotary encoder can be added to the translation platform 205 to rotate the substrate 203 from a given starting point to a known position (e.g., 180°). In a specific exemplary embodiment, the rotary encoder can have a resolution of 25,000 counts per revolution. However, larger or smaller resolution values ​​can be selected depending on a given application and the desired level of rotational accuracy.

[0030] In various embodiments, the translation stage 205 of the single camera pre-aligner system 200 may include a variety of substrate holding mechanisms known in the art.

[0031] For example, in various embodiments, the translation platform 205 may include a substrate pin-lifter to hold the substrate 203. A typical substrate pin-lifter is composed of a plurality of pins (e.g., typically three pins comprising metal, sapphire, or metal inlaid with sapphire), a pneumatic actuator to lift the substrate pin-lifter, and one or more position sensors to measure the level of the substrate pin-lifter. In various embodiments, the translation platform 205 may include an electrostatic chuck (ESC) known in the relevant art to hold the substrate 203. In various embodiments, the translation platform 205 may include a vacuum chuck known in the relevant art to hold the substrate 203.

[0032] The overhead camera 201 can include one or more lenses (e.g., there can be a single variable focal length lens or multiple single focal length lenses) and an image sensor (e.g., a CCD array, a CMOS-based sensor, an active pixel sensor, or other sensor type). The overhead camera 201 can also include a camera board with associated circuitry to facilitate image extraction. In one example, the overhead camera 201 is a color camera. A color camera may be desirable because capturing color can help distinguish the substrate 101 from the bonding agent 107 and the carrier substrate 105. In addition, the machine learning framework discussed below can be trained on color images that would otherwise pose integration challenges to grayscale images collected from a monochrome camera. However, for known substrate types using a network trained using grayscale images, a monochrome camera can also be used.

[0033] In various embodiments, as noted herein, multiple cameras may be used. Linear or linear cameras used in conventional systems may be replaced by area cameras discussed herein. In an embodiment, the overhead camera 201 may include a camera (e.g., a color area camera) having a resolution sufficient to identify the fiducial notch 103 (or other fiducial marker). In an embodiment, the overhead camera 201 may include a camera having a resolution sufficient to read an identification (ID) number (not shown) etched or otherwise formed onto the face of the substrate 101, 203. In a particular exemplary embodiment, and depending at least in part on the physical size of the ID number, the overhead camera 201 may have a resolution of 25 megapixels or more.

[0034] like Figure 2A As shown, the top camera 201 can be mounted approximately perpendicular to the uppermost portion of the substrate 203. As noted above, the field of view 207 of the top camera 201 substantially covers the uppermost region of the substrate 203.

[0035] In other embodiments not explicitly shown, the pre-aligner system may include, for example, two cameras that observe different (e.g., opposing) portions of the substrate 203. In various other embodiments, the pre-aligner system may include, for example, four cameras that observe four quadrants of the substrate 203. The portions or quadrants observed by the cameras may or may not overlap. In various other embodiments, two of the four cameras may be placed above the substrate 203, while the remaining two cameras may be placed below the substrate 203.

[0036] Reference below Figure 3 An example of a method for detecting a fiducial marker using the single camera pre-aligner system 200 is discussed. Figure 3 A similar method can also be used, for example, in a dual-camera or quad-camera pre-aligner system disclosed herein.

[0037] Figure 2B An example of a multi-camera pre-aligner system 230 that can be used to pre-align a substrate according to various embodiments of the disclosed subject matter is shown. The multi-camera pre-aligner system 230 can also be used to determine and quantify characteristics such as bow and warp (if present), as well as other characteristics of the substrate. Figure 2B It is shown as including a top camera 201 , and two cameras 231A, 231B mounted above a substrate 233 , and two cameras 231C, 231D mounted below the substrate 233 .

[0038] The two cameras 231A and 231B installed above the substrate 233 and the two cameras 231C and 231D installed below the substrate 233 can be used as described above. Figure 2A The top camera 201 described above is the same or similar. In addition, more than two cameras can be installed above or below the substrate. In addition, the number of cameras installed above the substrate 233 does not need to be the same as the number of cameras installed below the substrate 233.

[0039] exist Figure 2B In the multi-camera prealigner system 230 of FIG. 2 , the top camera 201 can be considered optional. In this embodiment, two cameras 231A, 231B mounted above the substrate 233 and two cameras 231C, 231D mounted below the substrate 233 can be used to capture all images of the substrate 233, as described below. If the top camera 201 is used in one embodiment, the top camera 201 can be mounted substantially perpendicular to the uppermost surface of the substrate 233 along the centerline 235 of the translation platform 205.

[0040] Substrate 233 is shown to include an exaggerated bow (and / or warp) to better explain various embodiments of the disclosed subject matter. The bow may be symmetrical or asymmetrical. For example, if the bow is substantially symmetrical, the amount of bow d on the left side of substrate 233 is 1 The amount of bowing on the right side of the substrate 233 may be 2 In the case of a substantially asymmetric bowing of the substrate 233, the amount of bowing d on the left side of the substrate 233 is 1 The amount of bowing d on the right side of the substrate 233 may be different from 2 In either case, the arcuate curvature is symmetrical or asymmetrical. Figure 2A Single camera pre-aligner system 200 or Figure 2B The multi-camera pre-aligner system 230 can be used to determine the amount of bowing. This determination is described below.

[0041] Regarding the substrate alignment characteristics of the single-camera pre-aligner system 200 and the multi-camera pre-aligner system 230, and in certain exemplary embodiments, the single-camera pre-aligner system 200 and the multi-camera pre-aligner system 230 are configured to align a bonded or unbonded substrate in about 4 seconds or less. The lateral accuracy of the translation stage 205 (e.g., lateral translation of the substrate 203, 233 in the x-direction and the y-direction) can be within, for example, ±50 μm. The rotational accuracy of the translation stage 205 (rotating the substrate 203, 233 to a predetermined angle to align the substrate 203, 233) can be within ±50 μm. Figure 1A and Figure 1B The reference notch 103 for alignment) can be within, for example, ±0.06°.

[0042] In one high level exemplary method for using the pre-aligner system 200, 230, various steps may be included. However, not all of the steps shown may be required for a given operation. The exemplary steps may include at least some of the following steps discussed below.

[0043] The end effector of the robot arm places a substrate (e.g., substrate 203, 233) on a pre-aligner chuck (e.g., a substrate holder such as translation stage 205). One camera (e.g., overhead camera 201) or multiple cameras (e.g., overhead camera 201 and / or one or more of cameras 231A, 231B, 231C, 231D) captures an image of the substrate 203, 233. Depending on the number of cameras used, the process may vary.

[0044] When using a single camera (e.g., overhead camera 201), the substrate is rotated at least once 180 degrees and at least one new image is captured. Thus, using a 180° rotation and two images (one in each orientation of substrate 203, 233), several characteristics about the substrate (e.g., location of the fiducial (notch), θ offset, x-direction offset, and y-direction offset, etc.) can be quickly determined. If the use of a single camera also allows compensation for bowing in the substrate, the location of one or more substrate edges in the captured images is paired at 180°. When using multiple cameras, the substrate may not need to be rotated because each camera often captures a unique or partially unique portion of the substrate. Therefore, in the case of multiple cameras, rotation may not be required.

[0045] The captured image is processed (e.g., by a machine learning framework or convolutional neural network, as described below) Figure 5 The following information can be extracted from the image: the location of an edge (e.g., on a circular substrate) or multiple edges (e.g., on a square substrate); the location of a fiducial on the substrate; determining the substrate size (e.g., diameter); the amount of misalignment of the substrate on the substrate platform (e.g., misalignment in the x-direction, y-direction, z-direction, and / or theta-direction), and the amount of bow or warp in the substrate. If at least one top-side camera has sufficient resolution, the top-side substrate identification (ID) number or code can be read from the image (e.g., in an optical character recognition (OCR) operation). If misalignment of the substrate is detected from the processed image, the system can calculate translation corrections and theta corrections. The corrections can be transferred, for example, to the translation platform 205 or the transfer robot to correct the placement of the substrates 203, 233.

[0046] The fiducials on the substrate 203, 233 may be viewed directly by at least one camera, for example, using front lighting, back lighting, and / or light reflectors, if desired.

[0047] The angular misalignment of the substrate can be derived by observing, for example, a grid on the substrate (if present, such as on a wafer containing printed and / or etched dies). Additionally or alternatively, the misalignment of the substrate can be derived from a pre-learned pattern recognition image. If desired, lateral corrections (e.g., in the x-direction and y-direction) can also be calculated from the captured images. If a single camera is used (also for compensating for bowing in the substrate), the positions of one or more substrate edges in the captured images are paired at 180°.

[0048] Continuing with the exemplary method for using the pre-aligner systems 200, 230, for each image captured, the following operations may also be performed. If desired, the radial lens correction may be reversed to correct for lens distortion. For example, radial lens distortion (e.g., curvilinear distortion, such as pincushion or barrel distortion) is often present in the captured image as an artifact of the imaging system (e.g., distortion originating from the lens). Estimating the radial lens distortion and preparing the radial lens correction therefrom may be implemented using a convnet within the machine learning framework described below to prepare the radial lens correction. One of ordinary skill in the art will recognize that a variety of architectures of CNNs may be used to perform image classification.

[0049] The pixels within the captured image can be converted into physical units (e.g., linear dimensions, such as millimeters) via an algorithm such as a direct linear transformation (DLT) transformation matrix. The DLT transformation matrix can be predetermined and embedded into a machine learning framework or other processing environment. A two-dimensional (2D) or three-dimensional (3D) transformation matrix can be calculated to determine the translation and / or angular (θ) transformation used to move the observed substrate position to the desired substrate position.

[0050] If the pre-aligner has an XY (and optionally a Z) stage, the translation corrections are transferred to the XY stage before the robot picks up and removes the substrate from the translation stage. If the pre-aligner does not have an XY (or Z) stage, the translation corrections can be directly input into the transfer robot. If the pre-aligner has a theta stage, the rotation corrections are transferred to the theta stage before the robot picks up and removes the substrate from the translation stage. If the pre-aligner does not have a theta stage, the rotation corrections can be directly transferred to the transfer robot.

[0051] Figure 3 A generalized example of a method 300 for detecting a fiducial mark (e.g., a notch) on a substrate and detecting corrected coordinates of the substrate according to various embodiments of the disclosed subject matter is shown. At operation 301, the method 300 attempts to detect a notch and one or more edges of a substrate. In operation 301, one or more cameras (e.g., Figure 2A and Figure 2B The top camera 201 and / or cameras 231A, 231B, 231C, 231D) capture an image of a substrate (e.g., substrate 203 or substrate 233) including an edge (e.g., a circular substrate) or multiple edges (e.g., a square substrate) of the substrate and one or more fiducials (e.g., notches). Figure 2A After the translation platform 205), one or more images are captured.

[0052] In the case of a single camera, after capturing the first image, at operation 303, the substrate may be rotated approximately 180°. If more than one camera is used, a smaller rotation of the substrate (e.g., approximately 90°) or no rotation is used, depending on the viewing angle of the camera. Thus, at least two of the plurality of images display different views of the substrate (e.g., at least partially unique views). Using the plurality of images captured by the single camera after rotating the substrate, or using the images captured by each of the more than one camera, the images are transmitted to a machine learning framework, as described below with reference to Figure 5 Give a description.

[0053] The machine learning framework may be proximate (e.g., in-situ) to the pre-aligner system 200, 230. Alternatively, the machine learning framework may be located remotely from the pre-aligner system 200, 230. In other embodiments, components of the machine learning framework may be distributed with some components located at or near the pre-aligner system 200, 230 and other components remotely from the pre-aligner system 200, 230.

[0054] After rotating the substrate at operation 303, if necessary, at operation 305, additional images of the substrate are captured by the one or more cameras. The additional images of the substrate allow one or more edges of the substrate to be compared with the first image. If the edges from the first image to the additional image are not aligned when the two images are stacked on each other (e.g., virtually stacked in a machine learning framework), there is a translation error in the position of the substrate. Comparison of the first image with the additional image also allows for better delineation of the fiducial markers. At operation 307, all images are transferred to the machine learning framework to calculate center offsets (e.g., at least in the x- and y-directions) and rotation corrections (if necessary).

[0055] At operation 309, a decision is made as to whether the center of the substrate and the rotational orientation of the substrate are correct. For example, based on the detection of the edges of the substrate, it can be determined whether the position of each edge is symmetrical relative to the calculated center of the substrate. If the center and rotational orientation are correct, the method 300 ends at operation 311.

[0056] If at least one of the center and rotational orientation is incorrect, the method 300 continues at operation 313 to transfer the correction coordinates (e.g., x offset, y offset, and / or theta offset) to, for example, the translation stage 205 or a robotic transfer mechanism for the substrate, as described above. At operation 315, the substrate is repositioned based on the transferred values ​​of the correction coordinates. The method 300 ends at operation 311.

[0057] The disclosed subject matter can be used, for example, in a machine learning framework using a deep convolutional neural network. The deep convolutional neural network characterizes an image of a substrate captured by the one or more cameras based on multiple images. The characterized images can then be used to describe features or characteristics of the substrate (e.g., distinguishing the imaged substrate from wax and a carrier substrate (if present), determining the location of one or more fiducials, detecting and quantifying the amount of bowing in the substrate (if present), and detecting correction coordinates if necessary). To avoid having to collect millions of images, transfer learning can be used. In an example, the number of images typically used to train a network can range from about 5000 images to about 50,000 images. Training begins with a pre-trained network. This training scheme limits the computational requirements often encountered in more complex deep convolutional neural network systems.

[0058] Training can begin with a mixture of exemplary images, including images of various types of substrates (with various types of fiducials), bonding agents (e.g., wax if the substrate is a bonding wafer), and substrate carriers. As described above, the images can include multiple images of a given substrate. Other desired features of the substrate, such as substrate thickness (e.g., to account for z-offset), substrate bowing, substrate warping, and substrate sagging (e.g., from thinning the substrate) can be noted to adjust the positioning position (e.g., pick position) between the robot's end effector and the substrate due to one or more of the substrate features or characteristics. These characteristics can be used, for example, to avoid collisions between the robot's end effector and the substrate due to one or more of the substrate characteristics. Thus, training can begin with generalizations of the captured images. Furthermore, in addition to defining the characteristics of the substrate, as noted above, an identification number (e.g., ID number) or code associated with each substrate can be recorded.

[0059] As described below, a generalized substrate characterization system including at least a portion of a machine learning network can be used with Figure 5 The generalized substrate characterization system may be used with or embedded in a machine of the invention. Thus, the generalized substrate characterization system may first be used in a training mode to train a machine learning framework, and then may be used in a normal operating mode to detect features or characteristics of each substrate. In various embodiments, the training mode may be performed by a manufacturer of the substrate characterization system. The data obtained from the training mode may then be used, for example, at a manufacturing facility (e.g., a semiconductor device manufacturer or "fab") to determine the characteristics of each substrate used within the facility.

[0060] Figure 4 A method for detecting fiducial markers and substrates (e.g., Figure 2BIn various embodiments, the plurality of cameras may include an overhead camera 201 and a Figure 2B At least one or one or more of the cameras 231A, 231B, 231C, 231D. In various embodiments, the top camera 201 may not be used. Instead, at least two of the cameras 231A, 231B, 231C, 231D are used.

[0061] In an advanced example, for use Figure 2B The method 400 of using the multi-camera pre-aligner system 230 to detect fiducial marks and other characteristics of a substrate may include various steps. However, not all of the steps shown may be required for a given operation. Exemplary steps may include at least some of the following steps described below.

[0062] At operation 401, an end effector of a robot arm places a substrate (e.g., substrate 233) on a pre-aligner chuck (e.g., a substrate holder such as translation platform 205). At operation 403, at least one unique image of the substrate is captured by each of a plurality of cameras (e.g., at least two cameras, including overhead camera 201 and / or one or more of cameras 231A, 231B, 231C, 231D). Optionally, and depending at least in part on the ambient light level on the substrate, the substrate may be viewed directly by a plurality of the cameras using, for example, front-side lighting, back-side lighting, and / or a light reflector toward the substrate.

[0063] At operation 405, the captured image is processed (e.g., by a machine learning framework or convolutional neural network, as described below with reference to Figure 5 All images captured are transferred to a machine learning framework to calculate at least a center shift (e.g., at least in the x-direction and the y-direction) and a rotation correction (if necessary).

[0064] After the images are processed, at operation 407, comparison of the multiple images of the substrate to one another allows one or more edges of the substrate to be determined. If the edges from the first image to the additional image are not aligned when the two images are stacked on each other (e.g., virtually stacked in a machine learning framework or a processor that compares the processed images), there is a translation error in the position of the substrate. Comparison of the multiple images to one another further allows the presence and position of the fiducial marker to be determined. Comparison of one image from the first camera with at least one additional image from the additional camera also allows the reference marker to be better delineated relative to the actual position of the rest of the substrate and the one or more edges of the substrate.

[0065] In addition, as described herein, at least some of the following information may be extracted from the processed image at operation 409: the location of the fiducials on the substrate; determining the substrate size (e.g., diameter); the amount of misalignment of the substrate on the substrate platform (e.g., misalignment in the x-direction, y-direction, z-direction, and / or theta-direction), and the amount of bow or warp in the substrate. If at least one top-side camera (e.g., top-mounted camera 201 and / or one or both of cameras 231A, 231B) has sufficient resolution, a top-side substrate identification (ID) number or code may optionally be read from the image at operation 411 (e.g., in an optical character recognition (OCR) operation).

[0066] At operation 413, the pixel counts from the processed image may optionally be converted to physical units (e.g., units of millimeters or fractions of millimeters) to determine the size of the substrate (e.g., diameter), the translation distance that the substrate deviates from the center on the pre-aligner chuck, or other physical units of interest. The pixels may be converted to physical units via an algorithm such as a direct linear transform (DLT) transformation matrix. The DLT transformation matrix may be predetermined and embedded in Figure 5 A two-dimensional (2D) or three-dimensional (3D) transformation matrix can be calculated to determine the translation and / or angular (θ) transformation used to move the observed substrate position to the desired substrate position.

[0067] At operation 415, translation corrections and / or theta corrections are calculated from the processed images. At operation 415, if misalignment of the substrate is detected from the processed images, the system can calculate translation corrections (e.g., x-direction, y-direction, z-direction) and / or theta corrections. The corrections can be transferred, for example, to the translation stage 205 or the transfer robot to correct the placement of the substrate 233.

[0068] The angular misalignment of the substrate can be derived by observing, for example, a grid on the substrate 233 (if present, such as on a wafer containing printed and / or etched dies). Additionally or alternatively, the misalignment of the substrate 233 can be derived from a pre-learned pattern recognition image that can include an alignment relative to a fiducial (although when another alignment technique described herein is employed, a fiducial need not be used to determine the angular misalignment). If desired, lateral corrections (e.g., in the x- and y-directions) can also be calculated from the captured images.

[0069] At operation 417, a determination is made as to whether the center of the substrate and the rotational orientation of the substrate are correct. For example, based on the detection of the edges of the substrate, it can be determined whether the position of each edge is symmetrical relative to the calculated center of the substrate, which is relative to the center of the pre-alignment chuck. If the center and rotational orientation are correct, the method 400 ends at operation 423.

[0070] If at least one of the center and the rotational orientation is incorrect, the method 400 continues at operation 419 to transfer the correction coordinates (e.g., x offset, y offset, z offset, and / or theta offset) to, for example, the translation stage 205 or a robotic transfer mechanism for the substrate, as described above. At operation 421, based on the transferred values ​​of the correction coordinates, the substrate is repositioned to adjust for the correction coordinates. The method 400 ends at operation 423.

[0071] Continuing with the example of method 400 for using the multi-camera pre-aligner system 230, for each image captured, the following operations may also be performed. Figure 4 , but will be appreciated by those of ordinary skill in the art upon reading and understanding the disclosed subject matter, that radial lens correction may be inverted, if desired, to correct for lens distortion. For example, radial lens distortion (e.g., curvilinear distortion, such as pincushion or barrel distortion) is often present in captured images as an artifact of the imaging system (e.g., distortion originating from the lens of the camera). Estimating radial lens distortion and preparing radial lens correction therefrom may be implemented using a convnet within the machine learning framework described below to prepare radial lens correction. Those of ordinary skill in the art will recognize that radial lens correction may be inverted using a convnet within the machine learning framework described below to prepare radial lens correction. Figure 5 Various CNN architectures other than those shown are used to perform image classification.

[0072] Figure 5 An example of a machine learning framework 500 for detecting fiducial marks (e.g., notches) on a substrate, as well as other desired features of the substrate discussed herein, is shown in accordance with various embodiments of the disclosed subject matter. Figure 2A and Figure 2B ) and images obtained by the above-mentioned substrate characterization system to determine the location of the fiducial marks and other substrate characteristics. As described above, the machine learning framework 500 can be used in a training mode to train, for example, a convnet within the machine learning framework 500, and can then be used in a normal operating mode to characterize a substrate in a manufacturing environment (e.g., for the location of the fiducials).

[0073] like Figure 5As shown, the machine learning framework 500 is shown to include a preprocessor 503 and a machine learning network 530. The original image 501 is provided to the preprocessor 503. In this example, the preprocessor 503 filters or otherwise processes the original image 501 to, for example, crop, scale, or otherwise change or enhance the original image 501, and generates a preprocessed image 505.

[0074] The pre-processed image 505 can then be input into a machine learning network 530. The machine learning network 530 can be provided as a multi-layer machine learning model. For example, the machine learning network 530 can include the following four layers, including an input layer 507, a feature extraction layer 509, a feature relationship layer 511, and a decision layer 513. In this exemplary model, the decision layer 513 can have multiple outputs. The outputs can include, for example, key points, bounding boxes, masks for reference position identification, and masks for determining one or more edges of the substrate.

[0075] Continue to refer Figure 5 , pixel information from the preprocessed image 505 can be sent to the input layer 507. Each node in the input layer 507 can correspond to a pixel in the preprocessed image 505. The machine learning network 530 can be trained in one or more layers in layers 507 to 513 in an iterative manner. As described above, the decision layer 513 can output decisions about various substrate characteristics of a given substrate. The substrate characteristics are then generated in the output box 515. Therefore, the output box 515 can store the substrate characteristics extracted from the original image 501. In various embodiments, the output box 515 can provide a text indication showing the substrate characteristics (e.g., substrate offset in the x-direction, y-direction, z-direction and / or theta-direction, reference type and position; substrate bow and warpage; substrate ID number and other characteristics of the substrate). In various embodiments, the values ​​and / or characteristics in the output box 515 can be input as commands to, for example, instruct a translation stage or robot to reposition the substrate in the desired x, y, z and / or theta position.

[0076] As noted above, the machine learning framework 500 can first be used in a training mode to train the machine learning network 530 to identify the characteristics of the substrate. The machine learning framework 500 can then be used in a normal operating mode to identify the characteristics of the substrate in a manufacturing environment. The training of the machine learning network 530 can be a supervised process and can be performed outside the location where the identification of the characteristics of the substrate is performed. The training can use a set of training images (e.g., one or more training images) with known substrate characteristics or determined features to train the machine learning network 530.

[0077] The techniques shown and described herein can be used as follows: Figure 6The discussed machine 600 may be executed in part or in whole. Figure 6 An exemplary block diagram is shown including a machine 600 on which any one or more of the techniques (e.g., methods) discussed herein may be performed. In various examples, the machine 600 may operate as a standalone device or may be connected (e.g., networked) to other machines.

[0078] In a networked deployment, the machine 600 can operate in the capacity of a server machine, a client machine, or both in a server-client network environment. In an example, the machine 600 can act as a peer machine in a peer-to-peer (P2P) (or other distributed) network environment. The machine 600 can be a personal computer (PC), a tablet device, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a network router, a switch or a bridge, or any machine capable of (sequentially or otherwise) executing instructions for the actions to be taken by the specified machine. In addition, although only a single machine is shown, the term "machine" should also be understood to include any collection of machines that execute one (or more) instruction sets to perform any of the methods described herein, such as cloud computing, software as a service (SaaS), and other computer cluster configurations, either individually or jointly.

[0079] Examples as described herein may include logic or multiple components or mechanisms, or may be operated by logic or multiple components or mechanisms. A circuit system is a collection of circuits implemented in a tangible entity including hardware (e.g., simple circuits, gates, logic, etc.). Circuit system membership may change over time and potential hardware changes. A circuit system includes members that can perform specified operations individually or in combination when operating. In one example, the hardware of the circuit system can be immutably designed to perform specific operations (e.g., hardwired). In one example, the hardware including the circuit system may include variably connected physical components (e.g., execution units, transistors, simple circuits, etc.), which include computer-readable media that are physically modified (e.g., magnetically, electronically, such as via a change in physical state or a transformation of another physical property, etc.) to encode instructions for specific operations. When connecting physical components, the basic electrical properties of the hardware composition may, for example, change from insulating properties to conductive properties or vice versa. Instructions enable embedded hardware (e.g., execution units or loading mechanisms) to form members of the circuit system in hardware via variable connections to perform parts of specific operations when operating. Thus, when the device is operating, the computer-readable medium is communicatively coupled to other components of the circuit system. In one example, any physical component can be used in more than one member of more than one circuit system. For example, in operation, an execution unit can be used in a first circuit of a first circuit system at one point in time and reused by a second circuit in the first circuit system or by a third circuit in the second circuit system at a different time.

[0080] The machine 600 (e.g., a computer system) may include a hardware-based processor 601 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), a hardware processor core, or any combination thereof), a main memory 603, and a static memory 605, some or all of which may communicate with each other via an interconnect 630 (e.g., a bus). The machine 600 may also include a display device 609, an input device 611 (e.g., an alphanumeric keyboard), and a user interface (UI) navigation device 613 (e.g., a mouse). In an example, the display device 609, the input device 611, and the UI navigation device 613 may include at least a portion of a touch screen display. In addition, the machine 600 may also include a storage device 620 (e.g., a drive unit), a signal generating device 617 (e.g., a speaker), a network interface device 650, and one or more sensors 615, such as a global positioning system (GPS) sensor, a compass, an accelerometer, or other sensors. The machine 600 may include an output controller 619, such as a serial controller or interface (e.g., a universal serial bus (USB)), a parallel controller or interface, or other wired or wireless (e.g., an infrared (IR) controller or interface, near field communication (NFC), etc.), which is coupled to communicate or control one or more peripheral devices (e.g., a printer, a card reader, etc.).

[0081] The storage device 620 may include a machine-readable medium on which is stored one or more sets of data structures or instructions 624 (e.g., software or firmware) embodying or utilized by any one or more of the techniques or functions described herein. The instructions 624 may also reside, completely or at least partially, within the main memory 603, within the static memory 605, within the mass storage device 607, or within the hardware-based processor 601 during execution of instructions therein by the machine 600. In an example, one or any combination of the hardware-based processor 601, the main memory 603, the static memory 605, or the storage device 620 may constitute a machine-readable medium.

[0082] Although the machine-readable medium is considered a single medium, the term "machine-readable medium" may include a single medium or multiple media (eg, a centralized or distributed database, and / or associated caches and servers) configured to store one or more instructions 624.

[0083] The term "machine-readable medium" may include any medium that can store, encode, or carry instructions for execution by the machine 600 and cause the machine 600 to perform any one or more of the techniques in the present disclosure, or any medium that can store, encode, or carry data structures used by or associated with such instructions. Non-limiting examples of machine-readable media may include solid-state memory and optical and magnetic media. Therefore, machine-readable media are not transient propagating signals. Specific examples of large-scale machine-readable media may include: non-volatile memory, such as semiconductor memory devices (e.g., electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM)) and flash memory devices; magnetic or other phase change or state change memory circuits; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.

[0084] The instructions 624 may also be transmitted or received over the communication network 621 using a transmission medium via the network interface device 650 using any of a variety of transmission protocols (e.g., frame relay, Internet Protocol (IP), transmission control protocol (TCP), user datagram protocol (UDP), hypertext transfer protocol (HTTP), etc.). Example communication networks may include, among other things, a local area network (LAN), a wide area network (WAN), a packet data network (e.g., the Internet), a mobile telephone network (e.g., a cellular network), a plain old telephone (POTS) network, and a wireless data network (e.g., a wireless network called a cellular network). The Institute of Electrical and Electronics Engineers (IEEE) 802.22 family of standards, known as The IEEE 802.26 family of standards), the IEEE 802.25.4 family of standards, a peer-to-peer (P2P) network. In an example, the network interface device 650 may include one or more physical jacks (e.g., Ethernet, coaxial, or telephone jacks) or one or more antennas to connect to the communication network 626. In an example, the network interface device 650 may include multiple antennas to wirelessly communicate using at least one of single-input multiple-output (SIMO), multiple-input multiple-output (MIMO), or multiple-input single-output (MISO) technology. The term "transmission medium" should be understood to include any intangible medium that can store, encode, or carry instructions for execution by the machine 600, and includes digital or analog communication signals or other intangible media to facilitate the communication of such software.

[0085] As used herein, the term "or" may be interpreted as being inclusive or exclusive. In addition, based on reading and understanding the disclosure provided, one of ordinary skill in the art will appreciate other embodiments. In addition, one of ordinary skill in the art will readily appreciate that various combinations of the techniques and examples provided herein may all be applied in various combinations.

[0086] Throughout the specification, multiple instances can implement the parts, operations or structures described as single instances. Although each operation is shown and described as a separate operation, one or more of each operation can be performed simultaneously, and unless otherwise specified, it is not required that these operations must be performed in the order shown. The structure and function presented as a separate component in the example configuration can be implemented as a combined structure or component. Similarly, the structure and function presented as a single component can be implemented as a separate component. These and other variations, modifications, additions and improvements all fall within the scope of the subject matter described herein.

[0087] In addition, although not explicitly shown, it will be appreciated by those skilled in the art that each of the various arrangements, quantities, and number of elements may be different (e.g., the number of cameras). In addition, each example shown and described herein represents only one possible configuration and should not be considered to limit the scope of the present invention.

[0088] Although various embodiments are discussed separately, these separate embodiments are not intended to be considered independent techniques or designs. As indicated above, each of the various parts can be interrelated, and each can be used alone or in combination with other embodiments discussed herein. For example, although various embodiments of operations, systems, and processes have been described, these methods, operations, systems, and processes can be used alone or in various combinations.

[0089] Therefore, as will be apparent to those of ordinary skill in the art when reading and understanding the disclosure provided herein, many modifications and variations can be made. According to the foregoing description, in addition to those methods and devices listed herein, functionally equivalent methods and devices within the scope of the present disclosure will be apparent to those skilled in the art. Portions and features of some embodiments may be included in portions and features of other embodiments, or replace portions and features of other embodiments. Such modifications and variations are intended to fall within the scope of the appended claims. Therefore, the present disclosure is limited only by the terms of the appended claims and the full scope of equivalents to which these claims are entitled. It should also be understood that the terms used herein are only for the purpose of describing specific embodiments and are not intended to be limiting.

[0090] The abstract of the specification is provided to allow the reader to quickly determine the essence of the present technical disclosure. The abstract is submitted with the understanding that it will not be used to interpret or limit the claims. In addition, in the foregoing detailed description, it can be seen that various features can be combined together in a single embodiment for the purpose of simplifying the present disclosure. This method of disclosure should not be interpreted as limiting the claims. Therefore, the following claims are hereby incorporated into the detailed description, with each claim acting as a separate embodiment by itself.

[0091] The description provided herein includes illustrative examples, devices and apparatuses that embody various aspects of the subject matter described in the present invention document. In this description, for the purpose of explanation, many specific details are set forth in order to provide an understanding of the various embodiments of the subject matter discussed. However, it is apparent to one of ordinary skill in the art that various embodiments of the disclosed subject matter can be practiced without these specific details. In addition, well-known structures, materials and techniques have not been shown in detail to avoid blurring the various illustrated embodiments. As used herein, the terms "approximately", "approximately" and "substantially" may refer to values ​​within ±10% of a given value or range of values, for example.

[0092] The following numbered examples are specific embodiments of the disclosed subject matter.

[0093] Embodiment 1: An apparatus for detecting a reference located on a substrate and pre-aligning it to a predetermined position on a substrate holder. The apparatus includes at least one camera, which is positioned above the substrate when the substrate is located on the substrate holder. The at least one camera is configured to capture multiple images of at least a top surface of the substrate, at least two of the multiple images showing at least partially unique views of the substrate. A hardware-based processor is configured to determine the position of an edge of the substrate based on the captured multiple images. The hardware-based processor is further configured to determine a characteristic dimension of the substrate across the top surface, determine the position of the reference on the substrate, calculate at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder, and determine the position of the reference based on the captured multiple images.

[0094] Embodiment 2: According to the device of embodiment 1, the device also includes a rotating platform, which is coupled to the substrate holder to rotate the substrate at least about 180 degrees from an original 0 degree position placed on the substrate holder, and the at least one camera is configured to acquire at least one image of the substrate at each of 0 degrees and about 180 degrees, and the at least one image of the substrate acquired at each of 0 degrees and about 180 degrees thereby displays the at least partially unique view of the substrate.

[0095] Embodiment 3: The apparatus of Embodiment 2, wherein the hardware-based processor is configured to determine a location of an edge of the substrate based on the at least one image obtained at each of 0 degrees and approximately 180 degrees.

[0096] Embodiment 4: The device of any preceding embodiment, wherein the fiducial comprises at least one type of fiducial, including a type selected from a notch and a flat surface on the substrate.

[0097] Embodiment 5: The apparatus according to any one of the preceding embodiments, wherein the apparatus is configured to pre-align a transparent substrate.

[0098] Embodiment 6: The apparatus according to any one of the preceding embodiments, wherein the apparatus is configured to recognize a substrate identification code based on optical character recognition, wherein the substrate identification code is located on the top surface of the substrate.

[0099] Embodiment 7: The apparatus of any preceding embodiment, further comprising at least one camera located below the substrate during the pre-alignment process.

[0100] Embodiment 8: The apparatus of any preceding embodiment, wherein the hardware-based processor comprises a convolutional neural network.

[0101] Embodiment 9: The apparatus according to any of the preceding embodiments, wherein the hardware-based processor is further configured to determine an offset of the reference on the substrate relative to the predetermined position on the substrate holder in the z direction.

[0102] Embodiment 10: The device of any preceding embodiment, wherein the characteristic dimension is a diameter of the substrate.

[0103] Embodiment 11: The apparatus of any preceding embodiment, wherein the hardware-based processor is further configured to determine an amount of bowing in the substrate based on the captured plurality of images.

[0104] Embodiment 12: The apparatus according to any one of the preceding embodiments, wherein the x-direction offset, the y-direction offset and the theta-direction offset of the substrate relative to the predetermined position on the substrate holder are to be transferred to a substrate transfer robot.

[0105] Embodiment 13: The apparatus according to Embodiment 12, wherein the substrate transfer robot is configured to reposition the substrate to the predetermined position on the substrate holder based on the transfer values ​​of the x-direction offset, the y-direction offset, and the theta-direction offset.

[0106] Embodiment 14: The apparatus according to any of the preceding embodiments, wherein the x-direction offset, the y-direction offset and the theta-direction offset of the substrate relative to the predetermined position on the substrate holder are transferred to the substrate holder.

[0107] Embodiment 15: The apparatus according to Embodiment 14, wherein the substrate holder is configured to move the substrate to the position of the predetermined position on the substrate holder based on the transfer values ​​of the x-direction offset, the y-direction offset, and the theta-direction offset.

[0108] Embodiment 16: The apparatus of any preceding embodiment, wherein the hardware-based processor is configured to apply a neural network to the plurality of images captured from the substrate to determine the location of the fiducial on the bonded substrate.

[0109] Embodiment 17: A method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder. The method comprises capturing at least one first image of at least a top surface of the substrate and capturing at least one second image of at least the top surface of the substrate, wherein the at least one first image and the at least one second image show at least a portion of a unique view of the substrate. Processing the at least one first image and the at least one second image to determine the position of the fiducial, wherein the processing further comprises calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder.

[0110] Embodiment 18: According to the method of Embodiment 17, the method also includes rotating the substrate from an original 0 degree position placed on the substrate holder by at least about 180 degrees between capturing the first image and capturing the second image, and the processor is further configured to determine the position of the reference based on at least one image obtained at each of 0 degrees and 180 degrees.

[0111] Embodiment 19: The method according to any one of Embodiment 17 or Embodiment 18, the method further comprising determining the position of the reference on both the bonded substrate and the unbonded substrate based on the at least one first image and the at least one second image.

[0112] Embodiment 20: The method according to any one of Embodiments 17 to 19, further comprising processing the at least one first image and the at least one second image to determine a characteristic dimension of the substrate across the top surface of the substrate.

[0113] Embodiment 21: According to the method described in any one of Embodiments 17 to 20, the method further comprises identifying a substrate identification code from one of the at least one first image and the at least one second image based on optical character recognition, wherein the substrate identification code is located on the top surface of the substrate.

[0114] Embodiment 22: The method of any one of Embodiments 17 to 21, wherein the processing comprises using a convolutional neural network to determine characteristics of the substrate.

[0115] Embodiment 23: A computer-readable medium comprising instructions that, when executed by a machine, cause the machine to perform operations including: receiving at least one first image of at least a top surface of the substrate and receiving at least one second image of at least the top surface of the substrate. The at least one first image and the at least one second image display at least a portion of a unique view of the substrate. Processing the at least one first image and the at least one second image to determine a position of the reference on the substrate. The processing also includes calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder, and determining the position of the reference based on the at least one first image and the at least one second image.

[0116] Embodiment 24: The computer readable medium of Embodiment 23, wherein the operations further comprise determining the location of the fiducial on both the bonded substrate and the unbonded substrate based on the at least one first image and the at least one second image.

[0117] Embodiment 25: A computer-readable medium according to any one of Embodiment 23 or Embodiment 24, wherein the operation further comprises processing the at least one first image and the at least one second image to determine a characteristic dimension of the substrate across the top surface of the substrate.

[0118] Embodiment 26: The computer readable medium of any one of Embodiments 23 to 25, wherein the operations further comprise determining a characteristic of the substrate using a convolutional neural network.

[0119] Embodiment 27: A method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder. The method includes capturing, by each of a plurality of cameras, at least one unique image of the substrate after the substrate is placed on the substrate holder; processing each of the captured images; determining at least one edge of the substrate based on the processed images; determining the position of the fiducial on the substrate based on the processed images; and calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate relative to the predetermined position on the substrate holder based on the processed images.

[0120] Embodiment 28: The method of Embodiment 27, further comprising determining the location of the fiducial on both the bonded substrate and the unbonded substrate based on the processed image.

[0121] Embodiment 29: The method of any one of Embodiment 27 or Embodiment 28, further comprising processing the captured image to determine a characteristic dimension of the substrate across a top surface of the substrate.

[0122] Embodiment 30: The method according to any one of Embodiments 27 to 29, further comprising identifying a substrate identification code from at least one of the captured images based on optical character recognition, wherein the substrate identification code is located on the top surface of the substrate.

[0123] Embodiment 31: The method according to any one of Embodiments 27 to 30, further comprising converting pixel units from the captured image into physical units.

Claims

1. An apparatus for detecting a reference located on a substrate and pre-aligning the reference to a predetermined position on a substrate holder, the apparatus comprising: at least one camera to be positioned above the substrate when the substrate is on the substrate holder, the at least one camera to capture a plurality of images of at least a top surface of the substrate, at least two of the plurality of images showing distinct views of at least a portion of the substrate; and a hardware-based processor for determining a position of an edge of the substrate based on the plurality of captured images, the hardware-based processor further for determining a characteristic dimension of the substrate across the top surface, determining a position of the fiducial on the substrate, calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate from the predetermined position on the substrate holder, and determining a position of the fiducial based on the plurality of captured images.

2. The apparatus of claim 1 , further comprising a rotation platform coupled to the substrate holder to rotate the substrate at least approximately 180 degrees from an original 0 degree position placed on the substrate holder, the at least one camera being configured to acquire at least one image of the substrate at each of 0 degrees and approximately 180 degrees, the at least one image of the substrate acquired at each of 0 degrees and approximately 180 degrees thereby displaying the at least partially unique view of the substrate.

3. The apparatus of claim 2, wherein the hardware-based processor is configured to determine a location of an edge of the substrate based on the at least one image obtained at each of 0 degrees and approximately 180 degrees.

4. The apparatus of claim 1, wherein the fiducial comprises at least one type of fiducial, including a type selected from a notch and a flat surface on the substrate. The apparatus of claim 1 , wherein the apparatus is configured to pre-align a transparent substrate. 6 . The apparatus of claim 1 , wherein the apparatus is configured to recognize a substrate identification code based on optical character recognition, wherein the substrate identification code is located on the top surface of the substrate.

7. The apparatus of claim 1, further comprising at least one camera positioned below the substrate during a pre-alignment process.

8. The apparatus of claim 1, wherein the hardware-based processor comprises a convolutional neural network.

9. The apparatus of claim 8, wherein the hardware-based processor is further configured to determine an offset of the fiducial on the substrate from the predetermined position on the substrate holder in the z-direction.

10. The apparatus of claim 1, wherein the characteristic dimension is a diameter of the substrate.

11. The apparatus of claim 1, wherein the hardware-based processor is further configured to determine an amount of bow in the substrate based on the plurality of images captured.

12. The apparatus of claim 1, wherein the x-direction offset, the y-direction offset, and the theta-direction offset of the substrate from the predetermined position on the substrate holder are to be transferred to a substrate transfer robot. 13 . The apparatus of claim 12 , wherein the substrate transfer robot is configured to reposition the substrate to the predetermined position on the substrate holder based on transfer values ​​of the x-direction offset, the y-direction offset, and the theta-direction offset.

14. The apparatus of claim 1, wherein the x-direction offset, the y-direction offset, and the theta-direction offset of the substrate from the predetermined position on the substrate holder are to be transferred to the substrate holder. 15 . The apparatus of claim 14 , wherein the substrate holder is configured to move the substrate to a position of the predetermined position on the substrate holder based on the transfer values ​​of the x-direction offset, the y-direction offset, and the theta-direction offset.

16. The apparatus of claim 1, wherein the hardware-based processor is configured to apply a neural network to the plurality of images captured from the substrate to determine a location of the fiducial on a bonded substrate.

17. A method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder, the method comprising: capturing at least one first image of at least a top surface of the substrate; capturing at least one second image of at least the top surface of the substrate, the at least one first image and the at least one second image showing at least partially distinct views of the substrate; The at least one first image and the at least one second image are processed to determine the position of the fiducial, the processing further comprising calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate from the predetermined position on the substrate holder.

18. The method of claim 17, further comprising rotating the substrate at least approximately 180 degrees from an original 0 degree position placed on the substrate holder between capturing the first image and capturing the second image, the processor being further configured to determine a position of the reference based on at least one image obtained at each of 0 degrees and 180 degrees.

19. The method of claim 17, further comprising determining the position of the fiducial on both a bonded substrate and an unbonded substrate based on the at least one first image and the at least one second image.

20. The method of claim 17, further comprising processing the at least one first image and the at least one second image to determine a characteristic dimension of the substrate across the top surface of the substrate.

21. The method of claim 17, further comprising discerning a substrate identification code from one of the at least one first image and the at least one second image based on optical character recognition, wherein the substrate identification code is located on the top surface of the substrate.

22. The method of claim 17, wherein the processing comprises using a convolutional neural network to determine characteristics of the substrate.

23. A computer-readable medium comprising instructions that, when executed by a machine, cause the machine to perform operations comprising: receiving at least one first image of at least a top surface of the substrate; receiving at least one second image of at least the top surface of the substrate, the at least one first image and the at least one second image showing at least partially distinct views of the substrate; The at least one first image and the at least one second image are processed to determine the position of the reference on the substrate, the processing also comprising calculating at least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate from the predetermined position on the substrate holder, and determining the position of the reference based on the at least one first image and the at least one second image.

24. The computer-readable medium of claim 23, wherein the operations further comprise determining the location of the fiducial on both a bonded substrate and an unbonded substrate based on the at least one first image and the at least one second image.

25. The computer readable medium of claim 23, wherein the operations further comprise processing the at least one first image and the at least one second image to determine a characteristic dimension of the substrate across the top surface of the substrate.

26. The computer-readable medium of claim 23, wherein the operations further comprise determining a characteristic of the substrate using a convolutional neural network.

27. A method for pre-aligning a fiducial located on a substrate to a predetermined position on a substrate holder, the method comprising: capturing, by each camera of a plurality of cameras, at least one unique image of the substrate after the substrate is placed on the substrate holder; processing each of the captured images; determining at least one edge of the substrate based on the processed image; determining a position of the fiducial on the substrate based on the processed image; as well as At least an x-direction offset, a y-direction offset, and a theta-direction offset of the substrate from a predetermined position on the substrate holder are calculated from the processed images.

28. The method of claim 27, further comprising determining the location of the fiducial on both a bonded substrate and an unbonded substrate from the processed images.

29. The method of claim 27, further comprising processing the captured image to determine a characteristic dimension of the substrate across a top surface of the substrate.

30. The method of claim 27, further comprising discerning a substrate identification code from at least one of the captured images based on optical character recognition, wherein the substrate identification code is located on a top surface of the substrate.

31. The method of claim 27, further comprising converting pixel units from the captured image into physical units.

Images