Large-format continuous imaging system

Abstract

Scanning inspection devices detect anomalies on the surface of objects such as substrates, substrates with bonded chiplets, and carriers having individualized chiplets. In some embodiments, the inspection device includes a time-delay integral (TDI) linear sensor having an optical input and a data output, wherein two or more optical assemblies are arranged adjacent to each other, with the optical output focusing on different sections of the TDI linear sensor and the optical input receiving a portion of the surface to be inspected. The device may further include a platform having a top surface for supporting the object having the surface to be inspected, a motion assembly for moving the platform, and a controller that communicates with the motion assembly to move the platform in relation to the optical input of the optical assembly.

Description

Technical Field

[0001] Embodiments of the present principle generally relate to semiconductor processing of semiconductor substrates.

Background Art

[0002] Conventional wafer inspection tools are used to detect the unmounted surface of a wafer. Since the surface of the wafer is quite uniform, particles and defects can be detected by conventional methods. However, the inventors have observed that in the case of a mounted wafer where the chiplets are bonded to the surface, the height between the wafer surface and the chiplet surface is substantially different, so that the conventional wafer inspection tool cannot properly inspect the wafer surface. The inventors have also observed that after singulation and other processes, the chiplets may tilt obliquely and shift in position relative to each other, and as a result, the chiplets are no longer placed in the expected locations, which dramatically increases the difficulty of chiplet inspection. Furthermore, the inventors have observed that the conventional step-and-repeat acquisition type inspection process substantially affects the throughput of the inspection process.

[0003] Therefore, the inventors have provided an inspection method and apparatus that can quickly and accurately acquire inspection data regardless of height variations or misalignments.

Summary of the Invention

[0004] A method and apparatus for fast and continuous inspection scanning are provided herein.

[0005] In some embodiments, the inspection device for detecting anomalies may include at least one TDI linear sensor having a time-delay integral (TDI) linear sensor length, a TDI linear sensor optical input, and a TDI linear sensor data output, and a plurality of optical assemblies arranged adjacent to each other, each of which is positioned such that the optical output focuses on a portion of the TDI linear sensor length and the optical input having an input field of view (FOV) receives a portion of the surface of the object being inspected.

[0006] In some embodiments, the inspection apparatus includes a TDI linear sensor approximately 80 mm in length, a platform having a top surface for supporting an object having a surface to be inspected, a motion assembly configured to move the platform, a controller that communicates with the motion assembly to move the platform in relation to the optical inputs of a plurality of optical assemblies, a controller that moves the platform to adjust the distance between the optical inputs and the surface to be inspected, a controller that communicates with the motion assembly, the plurality of optical assemblies, and the TDI linear sensor and is configured to scan the entire surface of a substrate to which chiplets are bonded, and a carrier having individualized chiplets that communicates with the motion assembly, the optical assemblies, and the TDI linear sensor. The system may further include a controller configured to scan a surface, an integrated Z profiler having a laser-based autofocus module configured to determine the Z profile of the surface under inspection before acquiring a surface image by a TDI linear sensor, or an external Z profiler communicating with the controller and configured to provide the Z profile of the surface under inspection before acquiring a surface image by a TDI linear sensor, and a plurality of optical assemblies having similar focus and FOV, wherein at least one of the plurality of optical assemblies incorporates reflected illumination or dark-field illumination, and / or at least one of the plurality of optical assemblies has an adjustable lens configured to focus at least one of the plurality of optical assemblies.

[0007] In some embodiments, an inspection device for detecting anomalies may comprise: at least one TDI linear sensor having a time-delay integral (TDI) linear sensor length of approximately 50 mm to approximately 160 mm, a TDI linear sensor optical input, and a TDI linear sensor data output; a plurality of optical assemblies arranged adjacent to each other, each of which is positioned such that the optical output focuses on a portion of the TDI linear sensor length and the optical input having an input field of view (FOV) receives a portion of the surface of the object to be inspected; a platform having a top surface for supporting an object having a surface to be inspected, and having a four-axis motion assembly configured to move the platform; and a controller configured to communicate with the four-axis motion assembly, the plurality of optical assemblies, and the TDI linear sensor, and to scan the entire surface of a substrate or carrier.

[0008] In some embodiments, the inspection apparatus further includes a substrate to which a chiplet is bonded, a plurality of optical assemblies having similar focus and FOV, wherein at least one of the plurality of optical assemblies incorporates reflected illumination or dark-field illumination, and at least one of the plurality of optical assemblies has an adjustable lens configured to focus at least one of the plurality of optical assemblies, and / or an integrated Z profiler having a laser-based autofocus module configured to determine the Z profile of the surface under inspection before acquiring a surface image by a TDI linear sensor, or an external Z profiler communicating with a controller configured to provide the Z profile of the surface under inspection before acquiring a surface image by a TDI linear sensor.

[0009] In some embodiments, a method for acquiring inspection data may include scanning the entire surface of a substrate or carrier using a plurality of optical assemblies arranged adjacent to each other, and receiving image data from the surface focused on a single time-delay integral (TDI) linear sensor into the plurality of optical assemblies, wherein each optical assembly receives the image data so that it is focused on a different portion of the TDI linear sensor, reconstructs the image data from the TDI linear sensor to form reconstructed image data, maps the reconstructed image data to a position on the surface, analyzes the reconstructed image data and position data to determine the location of anomalies and form a surface defect map, and forms inferences based on the defect map and performs corrective actions based on the inferences.

[0010] In some embodiments, the method may further include corrective measures, which include marking a chiplet as defective and removing the chiplet from the bonder's selection pool, and / or the corrective measures include enhancing the pre-bonding process to mitigate future anomalies.

[0011] Further embodiments are disclosed below.

[0012] Embodiments of the present principle, briefly summarized above and described in more detail below, can be understood by referring to exemplary embodiments of the principle shown in the accompanying drawings. However, the accompanying drawings only show typical embodiments of the present principle and should not be considered limiting in scope, as the principle may accept other equally effective embodiments. [Brief explanation of the drawing]

[0013] [Figure 1] These are top and cross-sectional views of several types of substrates according to some embodiments of this principle. [Figure 2] This figure shows a continuous image detection array in operation for a time-delayed integral (TDI) linear sensor according to some embodiments of this principle. [Figure 3] This is a cross-sectional view of an inspection system according to one embodiment of this principle. [Figure 4] This is a top view of the scanning pattern of an inspection system according to one embodiment of this principle. [Figure 5] This is a cross-sectional view of an inspection system according to one embodiment of this principle. [Figure 6] This is a cross-sectional view of an inspection system having an internal laser-based Z-profiler according to some embodiments of this principle. [Figure 7] This is a top view of an external Z-profiler according to one embodiment of this principle. [Figure 8] This is a cross-sectional view of an inspection system with dark-field illumination according to one embodiment of this principle. [Figure 9] This is a cross-sectional view of an inspection system according to one embodiment of this principle. [Figure 10] This is a method for obtaining inspection data using some embodiments of this principle. [Figure 11] This is a schematic diagram of a hybrid bonding system based on some embodiments of this principle. [Figure 12] This is a cross-sectional view of an inspection system with dark-field illumination according to one embodiment of this principle. [Figure 13] This is a cross-sectional view of an inspection system having a single optical input and split outputs, according to some embodiments of this principle. [Modes for carrying out the invention]

[0014] For ease of understanding, the same reference numerals are used to indicate identical elements common to the drawings, where possible. The drawings are not drawn to scale and may be simplified for clarity. Elements and features of one embodiment may be incorporated into other embodiments for the benefit of the latter without further detail.

[0015] This method and apparatus provides improved inspection scanning of wafers and / or chiplets, enabling high-speed and continuous scanning of, for example, bonded or unbonded wafers and fragmented chiplets, thereby minimizing the impact of the inspection process on throughput. The technology disclosed herein is particularly useful as a high-speed, large-format continuous image inspection system capable of detecting defects and particles in the context of chip-to-wafer hybrid bonding and other inspection scenarios. This inspection system advantageously covers the entire wafer and addresses various challenges inherent in, for example, the hybrid bonding process and misaligned chiplets. Chip-to-wafer hybrid bonding requires defect inspection of three unique sample types: flat substrate wafers, fragmented chiplets, and bonded chiplets on substrate wafers. Chiplet height variations and misalignments present challenges for rapid measurement that are not present in standard flat substrate wafers. To overcome these challenges, the inspection system of this principle possesses unique capabilities not found in conventional optical inspection tools. The inspection system of this principle is also applicable for use in, for example, large display panel inspection applications.

[0016] Hybrid bonding is a packaging and chip stacking technique that precisely fuses individual chiplets onto a larger substrate wafer or bonding wafer. Metal interconnects embedded in the dielectrics of both the chiplet and the substrate are bonded together with the dielectric material, forming a "hybrid bond." Upon contact, the dielectric layers of the chiplet and the substrate weakly bond to each other almost instantaneously. A subsequent high-temperature annealing step is then performed to fuse the metal interconnects and strengthen the bonding strength of the dielectric layers. The presence of particles, chips, cracks, or excessive topographic variations on the chiplet / substrate surface negatively impacts bonding quality and leads to post-bonding defects. Post-bonding defects typically manifest as voids or gaps of varying sizes that prevent proper interconnection formation, negatively impacting yield and resulting in significant waste costs for completely manufactured chiplets / substrate dies. This waste is particularly serious in use cases where multiple chiplets (stacked side-by-side on the substrate or stacked vertically) may be mounted on a single substrate die.

[0017] For an inspection system to be useful for such bonding applications, for example, it must have the ability to scan a complete unbonded wafer (i.e., a bonding substrate to which chiplets can be bonded), a fragmented (potentially misaligned) unbonded chiplet on a carrier, and a bonded chiplet on a bonding substrate. Figure 100A of Figure 1 shows a bonding substrate 102. The bonding substrate 102 may include redistribution layers (RDLs) or other circuits on or below the surface, which can interface with chiplets bonded to the bonding substrate 102. The top surface 120 of the bonding substrate 102 is generally a uniform, flat surface. Figure 100B of Figure 1 shows a carrier 106, such as a tape frame, which holds a chiplet or die 104. The die 104 is fragmented (separated) and bonded to the carrier 106 so that a bonder can pick up the die 104 during the bonding process and place it on the bonding substrate 102. Although the carrier 106 adequately holds the die 104 for processing, the die 104 is not always positioned in an aligned state due to the fragmentation process which destroys the alignment defined by lithography, and due to the flexibility of the carrier 106.

[0018] Therefore, a portion of die 104 may be inclined obliquely (108), or deviate from the alignment of rows or columns (110). When carrier 106 flexes, the top surfaces 122 of die 104 also change in height relative to each other. In FIG. 100C of FIG. 1, bonding substrate 102 is shown with die 104 bonded to the upper surface 120 of bonding substrate 102. There is a substantial height difference between the top surface 122 of die 104 and the upper surface 120 of bonding substrate 102 that must be overcome when scanning the bonded wafer. In cross-sectional view 100D of FIG. 1, the upper example is bonding substrate 102 with die 104 bonded to upper surface 120 and smaller die 112 bonded to upper surface 120. In the lower example, additional die 114 are bonded to die 104, further increasing the height difference between upper surface 120 and the surface of smaller die 112. As used herein, the terms "chiplet" or "die" may be used interchangeably and refer to semiconductor circuits that are singulated and then bonded to a wafer or substrate. In some cases, a die may include two or more chiplets or only one chiplet.

[0019] The inventors observed that the standard approach involves using a step-and-repeat acquisition method, where the optical system is moved to the target position, autofocus is performed, an image is acquired, and then the system is moved to a new position. However, when the number of target positions approaches several thousand sites, as in advanced semiconductors, the step-and-repeat approach becomes far too slow. However, the inventors found that by using a time-delayed integral (TDI) linear sensor, it is possible to continuously acquire rectangular strips of image data without stopping at each individual image position. However, the inventors further found that handling individualized dies and bonded wafers using a TDI linear sensor presents a significant challenge because there is no optical system with microscopic resolution and a field of view (FOV) suitable for a large TDI linear sensor. The inventors have discovered that by using a multi-head, multi-illumination optical inspection system combined with a large TDI linear sensor, they can overcome challenges (multiple optical heads with composite FOVs compatible with the large-format TDI linear sensor provide the necessary microscopic resolution to achieve high scanning speeds) and achieve the ideal combination of throughput, magnification, and sensitivity required for pre-bonding and post-bonding inspections. In addition, using multiple optical heads allows for a cost-effective inspection system with the flexibility to scan with different resolutions, focal heights, and fields of view by individually controlling each optical head (a single optical solution with microscopic resolution and a very large FOV is prohibitively expensive and fragile).

[0020] In this technology, by using a large-format TDI linear sensor, it becomes possible to use a single sensor for multiple optical assemblies. In FIG. 2, the operation 200 of the TDI linear sensor is shown. The TDI linear sensor has a length M (204) and a width N (206). The number of pixels 216 along the length M (204) (X-axis 210) corresponds to the available lateral field of view along an object (e.g., die surface, wafer surface, etc.). The number of pixels along the width N (206) corresponds to the number of stages (1 to N) within the TDI linear sensor. The sample moves (212) along the Y dimension 208 through the stages (1 to N), and the sample is imaged N times by the sensor, thereby increasing the number of photons collected for a given sample and improving the SNR (signal-to-noise ratio) without hesitation (continuous imaging) even if the sample does not stop. In some embodiments, the length M (204) can have a length (214) of about 10 mm to about 160 mm or more.

[0021] Figure 3 shows an inspection system 300 using a TDI linear sensor 302, multiple optical assemblies 304, a stage 338, a motion assembly 324, and a controller 330. The motion assembly 324 moves the sample on the stage 338 along multiple axes to ensure a complete inspection scan of the sample. In some embodiments, the motion assembly is a four-axis stage (i.e., movement in the X, Y, Z, and theta directions). The motion assembly 324 allows the optical heads 352 of the optical assembly 304 to be positioned at any location on the substrate 322 under test. Stage movement, sensor acquisition, and illumination are coordinated by a combination of software commands / queries in the controller 330, and / or digital trigger signals intended to more precisely facilitate timing and synchronization between the object under test and the scanning system. Each optical head 352 may optionally include an independent Z positioner (e.g., motor-driven, piezo, etc.). The length 354 of the TDI linear sensor 302 may range from approximately 10 mm to approximately 160 mm. In some embodiments, the TDI linear sensor 302 may be replaced with a non-TDI linear sensor. When a non-TDI linear sensor is used instead of the TDI linear sensor 302, special care must be taken to ensure a good signal-to-noise ratio for the inspection measurement (e.g., using a brighter light source, a more efficient optical system, and / or a more sensitive image detector in the sensor).

[0022] Larger TDI linear sensors improve throughput by enabling continuous imaging of a wider field of view. However, the inventors have found two distinct challenges in using longer sensor lengths. First, there is a lack of available optics that can balance the magnification, numerical aperture (i.e., resolution), and FOV required for inspection systems in bonding applications. Existing lenses specifically designed for large-format TDI sensors may have the necessary magnification (e.g., around 5x) but suffer from extremely low resolution (e.g., NA=0.01). Microscope objective lenses, on the other hand, have both magnification (e.g., 5-10x) and resolution (e.g., NA=0.15-0.30), but generally cannot support sensor sizes exceeding 30mm. Second, when imaging a wider FOV on individualized dies or bonded substrates, multiple chiplets may be in the field of view simultaneously. Each chiplet in the field of view may have a different focal plane, and a single lens cannot focus on all chiplets in the field of view simultaneously. Reacquiring data from multiple focal planes through rescanning significantly reduces throughput. The inventors have found that high throughput can be achieved by using a large-format TDI linear sensor with multiple densely arranged optical assemblies 304 composed of high-magnification, high-resolution microscope bodies. The microscope bodies of the optical assemblies 304 are oriented along the longitudinal direction of the TDI linear sensor. Each optical assembly 304 includes an illumination system (e.g., internal or external to the optical assembly 304) which is described in detail below. In some embodiments, the optical assembly 304 also includes a Z-positioner (e.g., based on a motor, piezo, liquid lens, or motor-driven lens unit, etc.) to compensate for different height ranges of features on the sample under inspection.

[0023] As shown in Figure 3, the optical assembly 304 of the inspection system 300 has a sample FOV 316 adjusted so that the FOV width 320 on the sample is approximately 1 mm to approximately 5 mm in some embodiments. The Z distance 318 between the optical head 352 and the sample surface can be adjusted as needed by the motion assembly 324 and controller 330 (e.g., to maintain the Z distance 318 within the focusing range of the optical assembly 304, to clear the maximum Z height on the wafer, etc.). In some embodiments, the width 310 of the optical assembly 304 may be approximately 30 mm to 80 mm. The width 310 is typically determined by the internal space required to house internal optical components and optional internal light sources, etc. Any number of optical assemblies from 2 to N(312) can be used. The spacing distance 314 between optical assemblies may be approximately 1 mm to approximately 20 mm in some embodiments. While reducing the spacing between optical assemblies can increase the number of assemblies per TDI linear sensor, sufficient spacing is required to avoid vibration or other operational problems (optical assemblies must not come into contact with each other during operation).

[0024] The optical assembly 304 transmits an image of a portion of the sample surface to a portion of the M pixels (see Figure 2) of the TDI linear sensor 302 via the focused TDI FOV 306 through the optical assembly 304. In some embodiments, the width 308 of this portion of the TDI linear sensor 302 may be approximately 5 mm to approximately 30 mm, depending on the magnification performance of the optical assembly 304. The sample surface may be at different Z heights (e.g., tiplet 326, lower tiplet 328, no tiplet, etc.), which may require different focal planes or adjustments to the Z axis of the stage 338. In some embodiments, each optical head 352 corresponds to a different position along the length of the TDI linear sensor (i.e., across the M pixels shown in Figure 2) and a different position on the sample, and each optical assembly independently focuses on the sample directly below it using its own independent illumination / focusing system.

[0025] An example of an embodiment of the optical assembly 304 is shown in Figure 5. The optical head 510 has an optical system 516 that helps guide the sample FOV to the sample surface in order to acquire image data 508. In some embodiments, the optical assembly has an epi-illumination function. The sample surface is illuminated from within by a light source 502 that generates an illumination beam 504 that is reflected by the lens 506 and incident on the sample surface. The Z-position assembly 512 is used to perform rapid repositioning so that the sample FOV can be focused on the sample surface as the sample moves under the inspection system 500. In some embodiments, the Z-position assembly can be moved using a motor and can use electrically adjustable lenses or piezos, etc. The image data 508 passes through the optical assembly to the focusing lens 514 and then to a portion of the M pixels on the TDI linear sensor 302.

[0026] Referring again to Figure 3, the controller 330 can fully control the movement of the stage 338 via the motion assembly 324. The controller 330 can also communicate with the optical assembly 304 to facilitate adjustments such as illumination, focus, and other optical assembly parameters. The controller 330 can also receive raw image data from the TDI linear sensor 302 and / or control the TDI linear image sensor 302 as needed. The controller 330 may directly control the inspection system 300 or control a computer (or controller) associated with the inspection system 300. In operation, the controller 330 enables data acquisition and feedback from the inspection system 300 to optimize the performance of the inspection system 300 and control the processing flow according to the methods described herein. The controller 330 generally includes a central processing unit (CPU) 332, memory 334, and support circuitry 336. The CPU 332 may be any form of general-purpose computer processor that can be used in an industrial environment. The support circuitry 336 may conventionally be coupled to the CPU 332 and include a cache, clock circuitry, input / output subsystems, power supply, etc. Software routines, such as the methods described herein, may be stored in memory 334 and, when executed by CPU 332, can be converted into a special-purpose computer (controller 330). The software routines may also be stored and / or executed by a second controller (not shown) located remotely from the inspection system 300.

[0027] Memory 334, when executed by CPU 332, is a form of computer-readable storage medium containing instructions for facilitating the operation of semiconductor processes and equipment. The instructions in memory 334 are in the form of a program product, such as a program that implements the methods of the present principle. The program code may conform to one of several different programming languages. In one example, the present disclosure may be implemented as a program product stored on a computer-readable storage medium for use in a computer system. The program in the program product defines the function of the embodiments (including the methods described herein). Exemplary computer-readable storage media include, but are not limited to, non-writable storage media on which information is permanently stored (e.g., read-only memory devices in a computer, such as CD-ROM disks, flash memory, ROM chips, or any type of solid-state non-volatile semiconductor memory, readable by a CD-ROM drive) and writable storage media on which modifiable information is stored (e.g., floppy disks in a diskette drive, or hard disk drives, or any type of solid-state random-access semiconductor memory). Such computer-readable storage media is an embodiment of the present principle if it carries computer-readable instructions that direct the function of the methods described herein.

[0028] The inspection system 300 provides a high-speed, large-format continuous image inspection system capable of handling various sample types shown in Figure 1. In some embodiments, the inspection system 300 consists of multiple independently configured optical assemblies, each corresponding to a different section of a single TDI linear sensor. Each optical assembly 304 can also independently focus on different parts of the sample (e.g., wafer surface, chiplet surface, etc.). Because the optical assemblies can be configured independently, the inspection system 300 is highly flexible. For example, if high throughput is desired, each optical assembly 304 can be made identical and configured to image different parts of the sample (e.g., different chiplets, different surfaces, etc.). Using identically configured optical assemblies allows for simultaneous measurement of larger areas of the sample, reducing the overall time required to perform individual measurements. If inspection information at multiple depths is desired, each optical assembly 304 can be configured to correspond to different types of samples at different depths (e.g., optical assembly 1 corresponds to chiplet type 1 at height 1, optical assembly 2 corresponds to a substrate at height 2, etc.).

[0029] In some embodiments, the inspection system 1304 may have one or more optical assemblies 304, each having a single optical element 1302 focused on the substrate 322, as shown in Figure 1300 of Figure 13. The sample FOV 316 provides an image of the substrate 322 onto the single optical element 1302. The image of the substrate 322 from the single optical element 1302 is then split into multiple output FOVs 1306 on different regions of the TDI linear sensor 302, which are generated by one or more optical assemblies 304 (1308). By splitting the image from the single optical element 1302, it becomes possible to focus a variable number (1310) of output FOVs 1306 on different positions on the TDI linear sensor 302 for a single input FOV. In some embodiments, the output FOV 1306 may be divided as a function of wavelength and / or polarization, and / or based on a portion of the image from the input FOV (e.g., 50% going into path 1 (1312), 25% going into path 2 (1314), 25% going into path N (1316), etc.). In some embodiments, one or more optical assemblies 304 of the inspection system 1304 include a plurality of optical assemblies 1 to N (304A) (including optical assembly 1 (304B), optical assembly 2 (304C), optical assembly N (304D), etc.), each optical assembly 1 to N (304A) receives one optical path of a divided FOV (optical path 1 (1312), optical path 2 (1314), optical path N (1316), etc.) from a single optical element 1302, and each optical assembly 1 to N (304A) provides one of a plurality of output FOVs 1306 incident at different positions on the TDI linear sensor 302. In some embodiments, one or more optical assemblies 304 of the inspection system 1304 is a single optical assembly 304E, which receives an image from a single optical element 1302 and divides the output of the single optical assembly 304E into a plurality of output FOVs 1306.

[0030] Referring again to Figure 3, if multiple magnifications are desired, each optical assembly 304 may consist of different magnifications tuned for individual applications (for example, optical assembly 1 having a 50x magnification for searching for small defects, optical assembly 2 having a 1x magnification for searching for large cracks and large displacements, etc.). If multiple types of illumination are desired, as will be described in more detail below, each optical assembly 304 may consist of different types of illumination to accommodate different types of applications. For example, optical assembly 1 may be standard reflected illumination for searching for cracks, optical assembly 2 may be dark-field illumination to increase sensitivity to particles, and / or optical assembly 3 may be configured to have a Nomarski DIC arm or asymmetric illumination, etc., to increase sensitivity to topographic variations across the entire sample. Any wavelength can be used in the optical assembly as long as the spectral range is detectable by the TDI linear sensor 302. The inspection system 300 can also be configured by any combination of modifications (e.g., high throughput, multiple depths, multiple magnifications, multiple illumination types, etc.).

[0031] In one exemplary use case, at least three optical assemblies can be placed adjacent to each other for a large TDI linear sensor approximately 160 mm in length, thereby improving throughput by a corresponding factor. As a simplified example, and not intended to be limiting, the field of view (FOV) of each optical assembly over the sample can be equal to the field number (FN) of the microscope objective lens divided by the magnification of the microscope objective lens. Thus, a standard 5x microscope objective lens (FN=25) supports an FOV of 25 / 5x = 5 mm on the object side. In this case, the three optical assemblies would have access to a range of approximately 15 mm on the sample side and 75 mm on the sensor side. Figure 400 in Figure 4 shows the acquisition pattern of the multi-head inspection system based on the exemplary use case. Each optical assembly 304 sweeps a first swath 408 with a distinct rectangular shape across the entire substrate 322 as the substrate moves in the Y direction 402. The inspection system has three optical assemblies 304, so that three swaths are formed on the substrate 322 each time it passes over it, improving throughput. After the first swath 408 is acquired, the substrate 322 moves in the X direction 406 by the width of the swath, and the second swath 410 is acquired immediately next to the first swath 408. After the second swath 410 is completed, the substrate 322 moves in the X direction 406 by the width of the swath, and the third swath 412 is acquired immediately next to the second swath 410. This process is repeated until the entire sample is acquired. Because only a single TDI linear sensor is used in combination with multiple optical assemblies, this inspection system is more cost-effective than a system that uses multiple adjacent TDI linear sensors, where each sensor has an optical system for each of the multiple TDI linear sensors. While multiple TDI linear sensors require additional synchronization between sensors, the method and apparatus of this principle, which uses only a single TDI linear sensor, does not require such synchronization.

[0032] In some embodiments, it is necessary to independently focus each individual optical assembly of the inspection system. In such cases, each optical assembly can be focused on the appropriate Z-plane, thereby enabling the acquisition of a focused image. Passive autofocus methods based on image analysis are too slow for TDI-based inspection systems where the sample is constantly moving. We have found that by using an independent Z-profiler or array of Z-profilers to provide a Z-height map of the sample, the Z-positioner of each optical assembly can be rapidly adjusted during image acquisition of the sample. We have also found that the presence of a separate, independent Z-positioner on each optical assembly enables two operating modes: a high-throughput mode and a multi-depth mode. In the former embodiment, multiple chiplets of varying thicknesses can be measured simultaneously by the inspection system, improving the acquisition speed. Alternatively, each optical assembly can focus on a completely different sample target located at a different depth within the sample (for example, the first optical assembly focuses on a bonded chiplet at depth Z1, and the second optical assembly focuses on the substrate at depth Z2).

[0033] Z-height information for focusing each optical assembly can be obtained from an internal or external source (e.g., an integrated device within the inspection system, or a separate device that provides the inspection system with Z-profile input). The Z-height of the sample must be measured and communicated to the optical assembly before acquiring an image of its position on the sample. Figure 600A in Figure 6 shows an example of an integrated focus assembly based on a laser autofocus assembly 602. The laser autofocus assembly 602 generates a laser beam 604A that passes through the optical assembly and is incident on the sample (e.g., a tiplet 326). This laser beam 604A is reflected from the sample surface, and the reflected laser beam 604B returns to the laser autofocus assembly 602. The laser autofocus assembly 602 uses the timing of the laser beam 604A and the reflected laser beam 604B to construct the Z-profile (Z-height) of the sample surface to be subsequently sampled.

[0034] As shown in the enlarged top view 600B of Figure 6, enclosed by a dashed circle, the laser autofocus assembly 602 focuses the laser beam 604A onto a spot 606 on the sample prior to the acquisition of image data 508 by the inspection system as the sample moves in the Y direction 608. The sample passes under the mounting area of ​​the TDI linear sensor after the laser autofocus assembly 602 has acquired Z height information and delay information (timing data indicating the measurement position when the scan is performed), so that the height measurement position can be synchronized with the mounting area and image acquisition position of the TDI linear sensor. Figure 700 of Figure 7 shows an external Z profiler assembly 704. In some embodiments, the external Z profiler assembly 704 may be a laser-based Z profiler, etc. The external Z profiler assembly 704 may be located near the inspection system 706, such as inside or outside the inspection chamber. The external Z profiler assembly 704 can scan the substrate 102 for the height of the die 104, etc., using one or more lasers 702. Subsequently, the Z-profile (height profile) of the entire substrate or carrier can be transferred to the inspection system 706 before image scanning by the inspection system 706. The inspection system 706 can then use the Z-profile of the substrate to adjust the focus of each optical assembly before image acquisition by the optical assembly.

[0035] Illumination of the sample is primarily achieved using light sources in the ultraviolet (UV) / visible light / near-infrared (NIR) wavelength range. The wavelength is largely dependent on the performance of the TDI linear sensor. Other wavelengths can be used if they are compatible with the TDI linear sensor used in the inspection system. Silicon-based TDI linear sensors have good sensitivity from wavelengths of approximately 300 nm to approximately 1100 nm. In some embodiments, each optical assembly has an illumination source. The illumination source can correspond to one of two illumination modes. Referring again to Figure 5, optical assembly 304 uses the first reflected illumination mode. Reflected illumination occurs when the sample is illuminated through the optical head 510. The illumination angle coincides with the optical axis of the optical head 510. Reflected illumination can take various options to enhance contrast on the sample plane. For example, asymmetric pupil illumination or Nomarski differential interference contrast (DIC) illumination can emphasize topographic variations on the sample plane. Variations of reflected illumination include asymmetric illumination, polarization, or differential contrast imaging, which do not use a flat intensity field in the pupil.

[0036] The second illumination mode is oblique incidence or dark-field illumination. Dark-field illumination occurs when the sample is illuminated at an angle located outside the focusing aperture of the optical head 510. Dark-field illumination is particularly sensitive to particles on the chiplet / substrate surface and is also sensitive to large height changes, which can cause noticeable artifacts near the edges of the sample piece. The inventors have found that the problems of dark-field illumination can be mitigated when the height profile changes gradually when performing upper edge inspection by using bilateral dark-field illumination, as shown in Figure 800A of Figure 8. In some embodiments, dark-field illumination is provided by a first external light source 802A and a second external light source 802B. Dark-field illumination is sensitive to abrupt height changes caused by the edges of the chiplet 326. The first external light source 802A is used to illuminate the left side 326A of the chiplet 326, and the second external light source 802B is used to illuminate the right side 326B of the chiplet 326. In enlarged view 800B, the corner defect 808 exhibits overexposure and loss of detail when light is incident from the right 804, whereas when light is incident from the left 806, it appears with better contrast and detail. The inventors also found that the problem of dark-field illumination when the height profile changes abruptly and significantly during sidewall or lower edge inspection can be mitigated by using bilateral dark-field illumination as shown in Figure 1200A of Figure 12. In some embodiments, dark-field illumination is provided by a first external light source 1202A and a second external light source 1202B. The first external light source 1202A is used to illuminate the right side 326B of the tiplet 326, and the second external light source 1202B is used to illuminate the left side 326A of the tiplet 326. In the enlarged view 1200B, the lower edge defect 1208 casts a shadow on the right side of the tiplet 326 when light is incident from the left side 1206, whereas when light is incident from the right side 1204, the contrast and details are more clearly visible.

[0037] In some embodiments, the optical assemblies used in the inspection system may have the same or different illumination types / systems. For example, in some embodiments, an inspection system using a large-format TDI linear sensor may have four optical assemblies. The first optical assembly may have a dark-field illumination system with or without different wavelengths to better detect particles, the second and third optical assemblies may have reflected illumination with or without different wavelengths to enhance crack or chip inspection, and the fourth optical assembly may have a high-magnification Nomarski DIC illumination system for topographic inspection. Subsequently, a multi-illumination type inspection system can address various use cases for defect inspection without requiring multiple optical assembly designs or multiple inspection chambers within the bonding tool, significantly reducing costs and improving the flexibility of the inspection system.

[0038] Figure 9 shows an inspection system 900 having a motion assembly 324 with a stage 338, a plurality of optical assemblies 304, a TDI linear sensor 302, and a controller 330 that communicates with an inspection data processor 902 having an inspection data model 904 (optional). In some embodiments, the inspection data processor 902 may reside entirely or partially within the controller 330, and / or the controller 330 may reside entirely or partially within the inspection data processor 902. Image data processing generally consists of three stages: reconstruction, analysis, and inference (machine learning). The reconstruction stage involves taking raw data (in this case, a stream of pixels from the TDI linear sensor 302) directly and / or via the controller 330 into the inspection data processor 902. The inspection data processor 902 maps each pixel to the correct position of that pixel in wafer coordinate space (X,Y,Z). In some embodiments, calibration of each optical assembly is performed during system setup, and real-time information from the stage 338, motion assembly 324, and Z-positioners within each optical assembly may be utilized.

[0039] The analysis step performed by the inspection data processor 902 includes extracting information from the reconstructed image. In some embodiments, areas on the wafer / chiplet that are likely to have a high probability of defects may be recorded. The analysis can be achieved by segmenting the region of interest (e.g., chiplet) using various standard image matching techniques and highlighting deviations by comparing them with a reference die. For example, particles on a dielectric surface may appear as particularly bright pixels under dark-field illumination. Once defect areas are highlighted, these areas can be further classified into their own subcategories (e.g., particles, cracks, chips, etc.). A defect map can then be generated across the entire sample.

[0040] The inference phase performed by the inspection data processor 902 references numerous ways in which the defect map may affect the operation of the hybrid bonding tool, for example, without limitation. For instance, cracks or scratches found on a chiplet or substrate can be used to mark a die as defective and remove it from the bonder's selection pool. Individual particles may require additional pre-treatment steps, or the systematic presence of particles in specific areas of the wafer can be used to adjust pre-treatment steps for future batches (e.g., performing additional pre-cleaning before bonding). For example, excessive variation in surface topography may trigger changes to the preceding polishing step before bonding. Thus, the defect map provides a broad index of various aspects of tool functionality, which can help reduce run-to-run variability, provide machine learning for long-term process tuning, and shorten tool maintenance and service times. Defect data can also be used in the binning process and / or to identify the extent of defective dies, etc.

[0041] In some examples, the inspection data model 904 can be used to assist in providing inference. The inspection data model 904 may be based on historical data and / or on real-time data obtained from the ongoing inspection process. The inspection data model 904 can be continuously updated by feedback from the post-bonding process 906, directly or indirectly, via the inspection data processor 902. The feedback information may include annealing information, gap filling information, and / or chemical mechanical polishing (CMP) information related to the post-bonding process 906 or the bonding process. The inspection data model 904 can also be used directly and / or indirectly via the inspection data processor to assist in providing inference for pre-bonding processes 908, such as the pre-cleaning process, CMP process, and chiplet fragmentation process, and for the bonding process (e.g., a hybrid bonding process).

[0042] In Figure 10, in some embodiments, but not limited to, the method 1000 for acquiring inspection data can be performed before or after the bonding process. In block 1002, the surface of the substrate or carrier is scanned using a plurality of optical assemblies arranged adjacent to each other. Using multiple optical assemblies can improve throughput. The optical assemblies may have identical or different configuration parameters, such as magnification, illumination mode, and focal plane, as described herein, but not limited to. In some embodiments, the optical assemblies may have an internal light source operating in reflected illumination mode. In some embodiments, the optical assemblies may have an external light source operating in dark-field illumination mode. Contrast enhancement may also be used. In block 1004, image data from the surface is received by the plurality of optical assemblies, and this image data is then focused onto a single TDI linear sensor. Each optical assembly focuses individual image data onto a different portion of the TDI linear sensor. In some embodiments, three or more optical assemblies may be used. In some embodiments, the surface FOV of each optical assembly may be about 1 mm to about 5 mm. In some embodiments, the magnification of the optical assembly may be about 5x to about 50x. In some embodiments, the length (M) of the TDI linear sensor may be about 10mm to about 160mm. In some embodiments, each optical assembly can provide a surface image field of view (FOV) of about 10mm to about 30mm to a portion of the TDI linear sensor.

[0043] In block 1006, image data from the TDI linear sensor is reconstructed and mapped to a location on the surface. In block 1008, the reconstructed image data and location data are used to determine the location of anomalies and form a surface defect map. In block 1010, inferences are formed based on the defect map, and corrective actions are performed based on the inferences. In some embodiments, an inspection model (see, e.g., Figure 9) can be used to assist in establishing inferences and the like. In some embodiments, corrective actions may include marking a chiplet as defective and removing it from the bonder selection pool. In some embodiments, corrective actions may include enhancing pre-bonding, bonding, or post-bonding processes to mitigate anomalies. For example, additional pre-cleaning can be performed before bonding to remove additional particles, the bonding pressure on damaged chiplets can be reduced during the bonding process to reduce chiplet cracks, and / or additional gap-filling material can be used after bonding to ensure that varying chiplet heights are covered.

[0044] Referring again to Figure 9, the inspection system 900 can be independent of or integrated into various hardware structures. In some embodiments, the inspection system 900 may be a standalone inspection station and / or integrated into a hybrid bonding tool, such as the integrated hybrid bonding tool in Figure 11. For example, Figure 11 shows a schematic top view of an integrated hybrid bonding tool 1100 for bonding a die / chiplet to a target (bonding substrate), according to at least some embodiments. The methods described above and below may be performed within the integrated hybrid bonding tool 1100 to enhance hybrid bonding and / or to enhance pre-bonding or post-bonding processes. The integrated hybrid bonding tool 1100 generally includes an apparatus front-end module (EFEM) 1102 and a number of automation modules 1110 coupled in series to the EFEM 1102. Multiple automation modules 1110 are configured to transport one or more types of substrates from the EFEM 1102 through the integrated hybrid bonding tool 1100 and to perform one or more processing steps on one or more types of substrates (e.g., a source with chiplets, a source with fragmented chiplets, a target or bonding substrate to which chiplets are bonded, a bonding substrate to which chiplets have previously been bonded, etc.). Each of the multiple automation modules 1110 generally includes a transfer chamber 1116 and one or more process chambers 1106 coupled to the transfer chamber 1116 for performing one or more processes. The multiple automation modules 1110 are bonded to each other via their respective transfer chambers 1116, thereby providing modular expandability and customizability of the integrated hybrid bonding tool 1100.As shown in Figure 11, the multiple automation modules 1110 each comprise three automation modules: the first automation module 1110a is coupled to the EFEM 1102, the second automation module 1110b is coupled to the first automation module 1110a, and the third automation module 1110c is coupled to the second automation module 1110b.

[0045] The EFEM1102 includes a plurality of load ports 1114 for receiving one or more types of substrates. In some embodiments, the one or more types of substrates include 200 mm wafers, 300 mm wafers, 450 mm wafers, tape frame substrates, carrier substrates with or without reconfigured dies / chiplets, silicon substrates, glass substrates, and the like. In some embodiments, the plurality of load ports 1114 include at least one of one or more first load ports 1114a for receiving a first type of substrate 1112a, or one or more second load ports 1114b for receiving a second type of substrate 1112b. In some embodiments, the first type of substrate 1112a has a different size from the second type of substrate 1112b. In some embodiments, the second type of substrate 1112b includes a tape frame substrate or a carrier substrate. In some embodiments, the second type of substrate 1112b includes a plurality of dies / chiplets arranged on a tape frame or carrier plate. In some embodiments, the second type of substrate 1112b can hold dies / chiplets of different types and sizes. Therefore, one or more second load ports 1114b may have different sizes or receiving surfaces configured to load second type substrates 1112b having different sizes. In some embodiments, multiple load ports 1114 are arranged along a common side of the EFEM 1102. Figure 11 shows a pair of first load ports 1114a and a pair of second load ports 1114b, but the EFEM 1102 may include other combinations of load ports, such as one first load port 1114a and three second load ports 1114b. Furthermore, the integrated hybrid bonding tool 1100 may also incorporate a buffer 1190 that provides temporary storage or buffering for both the source and target. The buffer 1190 helps enable different size / type dies / chiplets to meet timing and other factors and / or constraints by making the target and / or source readily available for processing without the need to remove them from external sources.

[0046] In some embodiments, the EFEM 1102 includes a scanning station 1108 having a substrate ID reader for scanning one or more types of substrates to identify information. In some embodiments, the substrate ID reader includes a barcode reader or an optical character recognition (OCR) reader. The integrated hybrid bonding tool 1100 is configured to use any identification information from the scanned one or more types of substrates to determine processing based on that identification information, e.g., different processes and / or placements for a first type of substrate 1112a and a second type of substrate 1112b. In some embodiments, the scanning station 1108 may be configured to rotate to align the first type of substrate 1112a or the second type of substrate 1112b. In some embodiments, one or more of a plurality of automation modules 1110 include the scanning station 1108. The EFEM robot 1104 is positioned in the EFEM 1102 and is configured to transport first type substrates 1112a and second type substrates 1112b between a plurality of load ports 1114 and a scanning station 1108. The EFEM robot 1104 may include a substrate end effector for handling first type substrates 1112a and a second end effector for handling second type substrates 1112b. The EFEM robot 1104 can rotate or move linearly while rotating.

[0047] The transfer chamber 1116 includes a buffer 1120 configured to hold one or more first type substrates 1112a. In some embodiments, the buffer 1120 is configured to hold one or more first type substrates 1112a and one or more second type substrates 1112b. The transfer chamber 1116 includes a transfer robot 1126 configured to transfer the first type substrates 1112a and second type substrates 1112b between the buffer 1120, one or more process chambers 1106, and a buffer located in an adjacent automation module among a plurality of automation modules 1110. For example, the transfer robot 1126 in the first automation module 1110a is configured to transfer the first type substrates 1112a and second type substrates 1112b between the first automation module 1110a and the buffer 1120 in the second automation module 1110b. In some embodiments, the buffer 1120 is located within the internal volume of the transfer chamber 1116, which advantageously reduces the overall footprint of the tool. In addition, the buffer 1120 can be left open to the internal volume of the transfer chamber 1116 to facilitate access by the transfer robot 1126.

[0048] One or more process chambers 1106 may include atmospheric pressure chambers configured to operate under atmospheric pressure and vacuum chambers configured to operate under vacuum pressure. Examples of atmospheric pressure chambers include wet cleaning chambers, radiation chambers, heating chambers, inspection chambers, bonding chambers, etc. An example of a vacuum chamber is a plasma activation chamber. The atmospheric pressure chambers of the above types may also be configured to operate under vacuum if necessary. One or more process chambers 1106 may be any process chamber or module necessary to perform bonding processes, cleaning processes, radiation processes, inspection processes (e.g., the methods described above and below), etc. In some embodiments, each of the multiple automation modules 1110, one or more process chambers 1106, includes at least one of the following: a wet cleaning chamber 1122, a plasma activation chamber 1130, a degassing chamber 1132, a radiation chamber 1134, an inspection chamber, or a bonder chamber 1140. Thus, the integrated hybrid bonding tool 1100 includes at least one wet cleaning chamber 1122, at least one plasma activation chamber 1130, at least one degassing chamber 1132, at least one radiation chamber 1134, at least one inspection chamber, and at least one bonder chamber 1140. The one or more process chambers 1106 may be located at any suitable position within the integrated hybrid bonding tool 1100.

[0049] A wet cleaning chamber is configured to perform a wet cleaning process that cleans one or more types of substrates through a fluid such as water. The wet cleaning chamber may include a first wet cleaning chamber 1122a for cleaning a first type of substrate 1112a, and / or a second wet cleaning chamber 1122b for cleaning a second type of substrate 1112b. A degassing chamber is configured to perform a degassing process that removes moisture, for example, through a high-temperature baking process. In some embodiments, the degassing chamber includes a first degassing chamber 1132a and / or a second degassing chamber 1132b. A plasma activation chamber may be configured to perform an activation process on a substrate as preparation for hybrid bonding. Activation helps to increase the bonding strength between surfaces. In some embodiments, the plasma activation chamber includes a first plasma activation chamber 1130a and a second plasma activation chamber 1130b. The radiation chamber 1134 is configured to perform a radiation process to reduce the adhesion between dies on a source, such as a tape frame substrate or a carrier substrate having a reconfigured die. For example, the radiation chamber 1134 may be an ultraviolet radiation chamber configured to direct ultraviolet light towards the source, or a heating chamber configured to heat the source. Reducing the adhesion between the die and the source makes it easier to remove the die from the source. The inspection chamber 1118 can be used to inspect the source and target before bonding and to detect defects. The bonder chamber 1140 is configured to transfer at least a portion of the die from the source to the target for bonding. The bonder chamber 1140 generally includes a first support 1142 for supporting one of the first type substrates 1112a and a second support 1144 for supporting one of the second type substrates 1112b. After bonding, the inspection chamber 1118 can be used to inspect the bonded substrate or target and to detect defects resulting from the bonding.

[0050] In some embodiments, the last automation module of a plurality of automation modules 1110, for example, the third automation module 1110c in Figure 11, includes one or more bonder chambers 1140 (two are shown in Figure 11). In some embodiments, any of the plurality of automation modules 1110 may include an inspection chamber 1118 configured to perform measurements of one or more types of substrates. In Figure 11, the inspection chamber 1118 is shown as part of a second automation module 1110b coupled to a transfer chamber 1116 of the second automation module 1110b. However, the inspection chamber 1118 may be coupled to any transfer chamber 1116 or may be located within a transfer chamber 1116.

[0051] The controller 1180 controls the operation of any of the integrated hybrid bonding tools described herein, including the integrated hybrid bonding tool 1100. The controller 1180 may do so by directly controlling the integrated hybrid bonding tool 1100, or by controlling a computer (or controller) associated with the integrated hybrid bonding tool 1100. In operation, the controller 1180 enables data acquisition and feedback from the integrated hybrid bonding tool 1100, optimizes the performance of the integrated hybrid bonding tool 1100, and controls the processing flow in accordance with the methods described herein, such as using an inspection system to detect defects in the target or source wafer before or after bonding. The controller 1180 generally includes a central processing unit (CPU) 1182, memory 1184, and support circuitry 1186. The CPU 1182 may be any form of general-purpose computer processor that can be used in an industrial environment. The support circuitry 1186 is conventionally coupled to the CPU 1182 and may include a cache, clock circuitry, input / output subsystems, power supply, etc. Software routines, such as those described herein, may be stored in memory 1184 and executed by CPU 1182, which can then be converted into a special-purpose computer (controller 1180). The software routines may also be stored and / or executed by a second controller (not shown) located remotely from the integrated hybrid bonding tool 1100.

[0052] Memory 1184, when executed by CPU 1182, is a form of computer-readable storage medium containing instructions for facilitating the operation of semiconductor processes and equipment. The instructions in Memory 1184 are in the form of a program product, such as a program that implements the methods of the present principle. The program code may conform to one of several different programming languages. In one example, the present disclosure may be implemented as a program product stored on a computer-readable storage medium for use in a computer system. The program in the program product defines the function of the embodiments (including the methods described herein). Exemplary computer-readable storage media include, but are not limited to, non-writable storage media on which information is permanently stored (e.g., read-only memory devices in a computer, such as CD-ROM disks, flash memory, ROM chips, or any type of solid-state non-volatile semiconductor memory, readable by a CD-ROM drive) and writable storage media on which modifiable information is stored (e.g., floppy disks in a diskette drive, or hard disk drives, or any type of solid-state random-access semiconductor memory). Such computer-readable storage media is an embodiment of the present principle if it carries computer-readable instructions that direct the function of the methods described herein.

[0053] Embodiments of this principle may be implemented in hardware, firmware, software, or any combination thereof. Embodiments may also be implemented as instructions stored using one or more computer-readable media that can be read and executed by one or more processors. The computer-readable media may include any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computing platform, or a “virtual machine” running on one or more computing platforms). For example, the computer-readable media may include any suitable form of volatile or non-volatile memory. In some embodiments, the computer-readable media may include non-transient computer-readable media.

[0054] The above describes embodiments of the present principle, but other embodiments and further embodiments of the present principle can be devised without departing from the basic scope of the present principle.

Claims

1. A TDI linear sensor having a time-delay integral (TDI) linear sensor length, a TDI linear sensor optical input, and a TDI linear sensor data output, A plurality of optical assemblies arranged adjacent to each other, wherein each of the plurality of optical assemblies is configured such that the optical output focuses on a portion of the TDI linear sensor length, and the optical input having an input field of view (FOV) receives a portion of the surface of the object to be inspected. An inspection device equipped with features for detecting abnormalities.

2. The inspection apparatus according to claim 1, wherein the length of the TDI linear sensor is approximately 80 mm.

3. A platform having an upper surface for supporting an object having the surface to be inspected, the platform having a motion assembly configured to move the platform, A controller that communicates with the motion assembly to move the platform in response to the optical inputs of the plurality of optical assemblies, The inspection apparatus according to claim 1, further comprising:

4. The inspection apparatus according to claim 3, wherein the controller moves the platform to adjust the distance between the optical input and the surface of the object to be inspected.

5. The inspection apparatus according to claim 3, wherein the controller is configured to communicate with the motion assembly, the plurality of optical assemblies, and the TDI linear sensor and to scan the entire surface of the substrate.

6. The inspection apparatus according to claim 5, wherein the substrate has chiplets bonded to the substrate.

7. The inspection apparatus according to claim 3, wherein the controller is configured to communicate with the motion assembly, the optical assembly, and the TDI linear sensor and to scan a carrier having individualized chiplets.

8. An integrated Z profiler having a laser-based autofocus module configured to determine the Z profile of the surface to be inspected before acquiring a surface image by the TDI linear sensor, or An external Z profiler configured to provide a Z profile of the surface to be inspected before acquiring a surface image by the TDI linear sensor, the external Z profiler communicating with the controller The inspection apparatus according to claim 3, further comprising:

9. The inspection apparatus according to claim 1, wherein the plurality of optical assemblies have similar focus and FOV.

10. The inspection apparatus according to claim 1, wherein at least one of the plurality of optical assemblies incorporates reflected illumination or dark-field illumination.

11. The inspection apparatus according to claim 1, wherein at least one of the plurality of optical assemblies has an adjustable lens configured to focus the at least one of the plurality of optical assemblies.

12. A TDI linear sensor having a time-delay integral (TDI) linear sensor length of approximately 50 mm to approximately 160 mm, a TDI linear sensor optical input, and a TDI linear sensor data output, A plurality of optical assemblies arranged adjacent to each other, wherein each of the plurality of optical assemblies is configured such that the optical output focuses on a portion of the TDI linear sensor length, and the optical input having an input field of view (FOV) receives a portion of the surface of the object to be inspected. A platform having an upper surface for supporting an object having the aforementioned surface to be inspected, the platform having a four-axis motion assembly configured to move the platform, A controller configured to communicate with the four-axis motion assembly, the plurality of optical assemblies, and the TDI linear sensor, and to scan the entire surface of the substrate or carrier, An inspection device equipped with features for detecting abnormalities.

13. The inspection apparatus according to claim 12, wherein the substrate has chiplets bonded to the substrate.

14. The inspection apparatus according to claim 12, wherein the plurality of optical assemblies have similar focus and FOV.

15. The inspection apparatus according to claim 12, wherein at least one of the plurality of optical assemblies incorporates reflected illumination or dark-field illumination.

16. The inspection apparatus according to claim 12, wherein at least one of the plurality of optical assemblies has an adjustable lens configured to focus the at least one of the plurality of optical assemblies.

17. An integrated Z profiler having a laser-based autofocus module configured to determine the Z profile of the surface to be inspected before acquiring a surface image by the TDI linear sensor, or An external Z profiler configured to provide a Z profile of the surface to be inspected before acquiring a surface image by the TDI linear sensor, the external Z profiler communicating with the controller The inspection apparatus according to claim 12, further comprising:

18. Scanning the entire surface of a substrate or carrier using multiple optical assemblies arranged adjacent to each other, Receiving image data from the surface focused on a single time-delayed integral (TDI) linear sensor into a plurality of optical assemblies, wherein each optical assembly receives the image data and focuses it on a different portion of the TDI linear sensor. The image data from the TDI linear sensor is reconstructed to form reconstructed image data, and the reconstructed image data is mapped to a position on the surface. The reconstructed image data and position data are analyzed to determine the location of the anomaly and to form a defect map of the surface. Forming inferences based on the defect map and implementing corrective measures based on the inferences, A method for obtaining test data, including [specific data].

19. The method according to claim 18, wherein the corrective action includes marking the chiplet as defective and removing the chiplet from the bonder selection pool.

20. The method according to claim 18, wherein the corrective action includes enhancing the pre-bonding process to mitigate future anomalies.

Images