Inspection of adaptive patterning workpieces by dynamic design and deep learning-based presentation

Abstract

A reference optical image of the die is determined based on the design file by a deep convolutional neural network for inter-image translation. The reference optical image is subtracted from a target image, thereby producing a difference image. The difference image may be binarized after application of an area-of-care mask. The obtained binarized defect image can be used for optical inspection.

Description

[0001] CROSS REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Indian Patent Application No. 202241056424 filed on September 30, 2022 and U.S. Application No. 63 / 426,803 filed on November 21, 2022, the disclosures of which are hereby incorporated by reference. Technical Field

[0003] The present disclosure relates to optical inspection of workpieces such as semiconductor wafers. Background Art

[0004] The evolution of the semiconductor manufacturing industry places increasing demands on yield management and specifically metrology and inspection systems. Critical dimensions continue to shrink, but the industry needs to reduce the time it takes to achieve high-yield, high-value production. Minimizing the total time from detecting a yield issue to resolving the issue maximizes the return on investment for semiconductor manufacturers.

[0005] The manufacture of semiconductor devices such as logic and memory devices typically includes processing a workpiece such as a semiconductor wafer using a number of manufacturing processes to form various features and multiple levels of the semiconductor device. For example, photolithography is a semiconductor manufacturing process that involves transferring a pattern from a mask to a photoresist arranged on a semiconductor wafer. Additional examples of semiconductor manufacturing processes include, but are not limited to, chemical mechanical polishing (CMP), etching, deposition, and ion implantation. An arrangement of multiple semiconductor devices manufactured on a single semiconductor wafer may be separated into individual semiconductor devices.

[0006] Inspection processes are used at various steps during semiconductor manufacturing to detect defects on wafers to promote higher yields and therefore higher profits in the manufacturing process. Inspection has always been an important part of manufacturing semiconductor devices such as integrated circuits (ICs). However, as the size of semiconductor devices decreases, inspection becomes more important for the successful manufacture of acceptable semiconductor devices because smaller defects can cause device failures. For example, as the size of semiconductor devices decreases, the detection of defects of reduced size has become necessary because even relatively small defects can still cause undesirable aberrations in semiconductor devices.

[0007] Some workpieces include dies with multiple chips. Certain die designs with multiple chips, such as system-in-package (SIP), film-frame carrier (FFC), or 3D integrated circuit (3D IC), can make inspection difficult. Each die may be printed differently, which limits the ability to inspect by comparing adjacent devices.

[0008] Existing techniques use neighbor die subtraction or reference-based subtraction. These techniques require images to be structurally similar and aligned to each other. These techniques tend to generate a large number of disturbing points when performed on a die with multiple chips. It may be difficult to suppress such high levels of disturbing points. Disturbing points can be reduced by improved alignment, but this typically uses cumbersome and time-consuming manual alignment methods.

[0009] Therefore, improved methods and systems are needed. Summary of the invention

[0010] In a first embodiment, a method is provided. The method includes receiving a target image of a workpiece including a die having a plurality of chips and a design file including a design of the die at a processor. Using the processor, a reference optical image of the die is generated based on the design file by a deep convolutional neural network for image-to-image translation. Using the processor, the reference optical image is subtracted from the target image, thereby generating a difference image. Using the processor, an execution-time care region mask of the die is generated based on the design file including the design of the die. Using the processor, the execution-time care region mask is applied to the difference image, thereby generating a masked difference image. Using the processor, a threshold is applied to the masked difference image, thereby generating a binarized defect image. The die may include at least one SIP device, FFC, or 3D IC.

[0011] The deep convolutional neural network can be a recurrent generative adversarial network.

[0012] The design file may be a graphic design system file.

[0013] The method may include aligning the target image and the reference optical image prior to the subtracting.

[0014] The method may include extracting, using the processor, an illuminated region image from the target image using the execution-time illuminated region mask.

[0015] The method may include using an optical inspection system to generate the target image.

[0016] A non-transitory computer-readable medium storing a program may be configured to instruct the processor to execute the method of the first embodiment.

[0017] In a second embodiment, a system is provided. The system includes: a light source that generates a light beam; a stage that is configured to hold a workpiece in the path of the light beam; a detector that is configured to receive the light beam reflected from the workpiece; and a processor that is in electronic communication with the detector. The workpiece includes a die having a plurality of chips. The processor is configured to: generate a target image of the workpiece based on information from the detector; receive a design file that includes a design of the die; generate a reference optical image of the die based on the design file by a deep convolutional neural network for image-to-image translation; subtract the reference optical image from the target image, thereby generating a difference image; generate a run-time care area mask of the die based on the design file that includes the design of the die; apply the run-time care area mask to the difference image, thereby generating a masked difference image; and apply a threshold to the masked difference image, thereby generating a binarized defect image. The die may include at least one SIP device, FFC, or 3DIC.

[0018] The deep convolutional neural network can be a recurrent generative adversarial network.

[0019] The design file may be a graphic design system file.

[0020] The processor may be further configured to align the target image and the reference optical image prior to the subtracting.

[0021] The processor may be further configured to extract an illuminated region image from the target image using the execution-time illuminated region mask. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] For a fuller understanding of the nature and purpose of the present disclosure, reference should be made to the following detailed description taken in conjunction with the accompanying drawings, wherein:

[0023] Figure 1 It is an exemplary SIP device;

[0024] Figure 2 is a flow chart of a method according to the present disclosure; and

[0025] Figure 3 is a diagram of an embodiment of an optical inspection system according to the present disclosure. DETAILED DESCRIPTION

[0026] Although the claimed subject matter will be described in terms of certain embodiments, other embodiments including embodiments that do not provide all of the advantages and features set forth herein are also within the scope of the present disclosure. Various structural, logical, process step, and electronic changes may be made without departing from the scope of the present disclosure. Therefore, the scope of the present disclosure is defined solely by reference to the appended claims.

[0027] Through adaptive patterning, the print and appearance of each die are different. The embodiments disclosed herein use different design files for each die and workpiece to detect defects on the workpiece using deep learning methods for optical image rendering and attention area alignment. The alignment time limit can be removed by using a cycle generative adversarial network (CycleGAN) network for physics-based optical image rendering. These embodiments can help identify defects on the workpiece through adaptive die patterns by dynamically loading design files and using deep learning-based rendering. Defect detection on adaptive die patterned workpieces is possible. Using a physically-based reference generation with a design file avoids the training step.

[0028] Figure 1 is an exemplary SIP device. Multiple chips are included in a die. High-speed chip placement can cause translational and rotational offsets, such as Figure 1 As shown in Figure 1 As seen in FIG, Chip A and Chip B are positioned differently in Die A, Die B, and Die C. The redistribution lines (RDLs) connecting these chips can be structurally adjusted based on chip placement. When comparing two dies, the redistribution lines may be different.

[0029] Figure 1 The die-to-die pattern variations seen in cause problems during inspection. First, static care areas cannot be set during recipe creation. Second, previous die-to-die techniques of comparing the target die to adjacent dies will not work due to the non-repeating structure. For example, Figure 1 As shown in Figure 1, Die A and Die B are not identical. The angle difference means that it is not possible to draw a statically lit area for both dies.

[0030] Figure 2 is a flow chart of method 100. Some or all of the steps in method 100 may be performed using a processor.

[0031] Receive a target image 101. The target image 101 may be generated by, for example, an optical inspection system. The target image 101 may have a workpiece including a die with multiple chips, such as a SIP device, an FFC device, a 3D IC device, or another type of device. For ease of illustration, Figure 2 A generic SIP device is shown in FIG.

[0032] A design file 102 corresponding to the die in the target image 101 is also received. The design file 102 represents the planar geometry, text labels, and other information about the layout of the die in a hierarchical form. In an example, the design file 102 is a graphic design system (GDS) file, such as GDSII or OASIS. A GDS is a graphic design system containing design information for a workpiece or die. For a standard wafer, the GDS may contain a die-level design, but for a workpiece with adaptive die patterning, the GDS may contain a workpiece-level design that is different for each workpiece or has different die designs at the die level. The design file 102 may be a model die or a specific die that needs to be used in the method 100.

[0033] Using the design file 102, a deep convolutional neural network with inter-image translation is used to generate a reference optical image 103. Inter-image translation maps the input image and the output image so that the output image can be used to perform a specific task. For example, an image can be transformed from one domain to another to learn the mapping between the two images. Inter-image translation can be used for image synthesis, noise reduction, or other purposes. A cyclic generative adversarial network can be used to perform these activities. Each die can generate a different reference optical image 103. The cyclic generative adversarial network can be a CycleGAN, Pix2Pix, or other type of generative adversarial network model. Pix2Pix is ​​trained using aligned and paired data. CycleGAN can be trained with an unpaired data set. Pairing means that the images of the source domain and the target domain should have the same position and the number of images of the two domains should be the same.

[0034] Pix2Pix is ​​a Generative Adversarial Network (GAN) model designed for general image-to-image translation. The Pix2Pix model is a conditional GAN. The generation of the output image is conditioned on an input such as a source image. The discriminator is provided with a source image and a target image and determines whether the target is a credible transformation of the source image. The generator is trained via an adversarial loss, which encourages the generator to produce credible images in the target domain. The Pix2Pix GAN can perform image-to-image translation tasks, such as generating an image from a different image (e.g., a detailed image from a sketch, design, or blueprint).

[0035] CycleGAN is an image-to-image translation model, just like Pix2Pix. Unlike Pix2Pix, CycleGAN uses an unpaired dataset to learn the mapping between input and output images. Like Pix2Pix, CycleGAN can generate images from different images (e.g., detailed images from sketches, designs, or blueprints).

[0036] Deep learning, which has its roots in neural network technology, is a probabilistic graphical model with many layers of neurons, which is often referred to as a deep architecture. Deep learning technology processes information such as images, text, speech, etc. in a layered manner. When using deep learning in the present disclosure, feature extraction is automatically done using learning from data. For example, a deep learning classification module can be used to classify, sort, or bin defects based on one or more extracted features.

[0037] In general, deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data. In a simple case, there may be two sets of neurons: one that receives input signals and one that sends output signals. When an input layer receives input, it passes a modified version of the input to the next layer. In a deep network, there are many layers between the input and output, allowing the algorithm to use multiple processing layers consisting of multiple linear and nonlinear transformations.

[0038] Deep learning is part of a broader family of machine learning methods based on representations of learning data. Observations (e.g., features to be extracted for reference) can be represented in many ways (e.g., vectors of per-pixel intensity values ​​or in more abstract ways, such as a set of edges, regions of a specific shape, etc.). Some representations are better than others when simplifying learning tasks (e.g., face recognition or facial expression recognition). Deep learning can provide efficient algorithms for unsupervised or semi-supervised feature learning and hierarchical feature extraction.

[0039] In an embodiment, the deep learning module is configured as a neural network. In further embodiments, the deep learning module may be a deep neural network having a set of weights that model the world according to the data that the deep neural network has been fed to train it. Neural networks can be generally defined as computational methods based on relatively large collections of neural units that loosely model the way biological brains solve problems with relatively large clusters of biological neurons connected by axons. Each neural unit is connected to many other neural units, and connections can be either mandatory or inhibitory in terms of their effect on the activation state of the connected neural units. These systems are self-learning and trained rather than explicitly programmed, and excel in areas where solutions or feature detection are difficult to express in traditional computer programs.

[0040] Neural networks typically include multiple layers, with signal paths traversing from front to back. The goal of a neural network is to solve problems in the same way as the human brain, but neural networks are far more abstract. Modern neural network projects typically work with thousands to millions of neural units and millions of connections. A neural network can have any suitable architecture and / or configuration known in the art.

[0041] GANs use deep learning methods such as convolutional neural networks to provide generative modeling. Generative modeling is an unsupervised learning task in machine learning that involves automatically discovering and learning regularities or patterns in input data so that the model can be used to generate or output new examples that are plausibly determinable from the original data set.

[0042] GANs train a generative model by framing the problem as a supervised learning problem with two sub-models. First, there is a generator model that is trained to generate new instances. Second, there is a discriminator model that attempts to classify instances as real (from the domain) or fake (generated). The two models are trained together (i.e., adversarial) in a zero-sum game until the discriminator model is fooled enough for the generator model to produce credible instances.

[0043] Regions of the design file 102 that correspond to the optical image 101 may be used in the method 100. Features of the die may be in both the design file 102 and the optical image 101.

[0044] A deep convolutional neural network can be trained to map the design file 102 image to the target image 101 and generate a reference optical image 103. This learning generally does not require perfect alignment. Approximate locations in the design file 102 and the target image 101 are generally sufficient to learn the network parameters. During the execution phase, the corresponding GDS portion of the target die is obtained from the design file 102 and then converted to the reference optical image 103 using the model trained during recipe creation. The reference optical image 103 is used as a reference image, aligned, and compared to the target image 101 to detect anomalies. Manual alignment at setup time or between the target image 101 and the reference optical image 103 can be avoided.

[0045] Infrastructure with high-performance computing accelerators (e.g., graphics processing units (GPUs)) can be used to train the network by designing examples of snippets and images.

[0046] The reference optical image 103 may be generated from the design file 102 in real time or otherwise as needed.

[0047] Next, the reference optical image 103 is subtracted from the target image 101 using image subtraction. This produces a difference image 104. The image subtraction can be performed using the same neural network, a different neural network, or a convolutional processor. Figure 2 As shown in , image subtraction results in a difference image 104 having slight differences between the target image 101 and the reference optical image 103 .

[0048] Prior to image subtraction, the target image 101 may optionally be aligned to the reference optical image 103. Based on the reference optical image 103 generated, this may not be necessary.

[0049] The design file 102 is also optionally used to generate an execution-time care area mask 105. The execution-time care area mask 105 includes the design of the die in the target image 101. The execution-time care area mask 105 can also exclude areas with high noise or that are not the focus of inspection. A different execution-time care area mask 105 can be generated for each die.

[0050] Optionally, a run-time attention area mask 105 is applied to the difference image 104, which produces a masked difference image 106. Masked difference image 106 may mask out (or exclude) portions of the difference image 104, such as areas that are not inspected, areas that rarely contain defects, or areas known to cause false positives during inspection.

[0051] A threshold may be applied to the masked difference image 106. This step produces a binarized defect image 107 that can be used for optical inspection. The binarized defect image may include pixels having two colors, such as black and white. A threshold may be used to determine which pixels are assigned to one of the two colors. In an example, post-processing such as filtering and merging candidate defect pixels in the binarized defect image 107 may provide defect information for all sites.

[0052] Optionally, a run-time region of interest image can be extracted from the target image 101 using a run-time region of interest mask 105. The resulting image can be used for optical inspection or comparison with a binarized defect image 107. The run-time region of interest mask 105 can also be applied prior to image subtraction of one or both of the images resulting in a difference image 104.

[0053] In the example, the user creates a recipe using layer design information with overall alignment. A deep convolutional neural network learns network parameters to render a reference optical image.

[0054] At runtime, the design file is dynamically loaded for each workpiece and an optical reference image is rendered using a deep convolutional neural network. A high-performance computing node configuration or a central processing unit (CPU) cluster can be used. This rendered optical reference image is used to update the design-based attention area to the runtime optical image and is also used directly to detect anomalies.

[0055] In an embodiment, a deep learning based classifier may reduce disturbing points and separate defects of interest (DOIs) in the binarized defect image 107 .

[0056] Figure 31 shows one embodiment of an optical inspection system 200. System 200 includes an optical-based subsystem 201. In general, optical-based subsystem 201 is configured to generate an optical-based output of sample 202 by directing light to sample 202 (or scanning light over sample 202) and detecting light from sample 202. In one embodiment, sample 202 includes a wafer. The wafer may include any wafer known in the art, such as the wafer used in method 100. In another embodiment, sample 202 includes a reticle. The reticle may include any reticle known in the art.

[0057] exist Figure 3 In the embodiment of the system 200 shown in FIG. 2 , the optical-based subsystem 201 includes an illumination subsystem configured to direct light to the sample 202. The illumination subsystem includes at least one light source. For example, Figure 3 , the illumination subsystem includes a light source 203. In one embodiment, the illumination subsystem is configured to direct light to the sample 202 at one or more incident angles, which may include one or more oblique angles and / or one or more normal angles. Figure 3 , light from light source 203 is directed at an oblique angle of incidence through optical element 204 and then through lens 205 to sample 202. The oblique angle of incidence may include any suitable oblique angle of incidence that may vary depending on, for example, the characteristics of sample 202.

[0058] The optical-based subsystem 201 can be configured to direct light to the sample 202 at different incident angles at different times. For example, the optical-based subsystem 201 can be configured to change one or more characteristics of one or more elements of the illumination subsystem so that light can be directed to the sample 202 at different incident angles. Figure 3 202. In one such example, the optical-based subsystem 201 can be configured to move the light source 203, the optical element 204, and the lens 205 so that the light is directed to the sample 202 at different oblique angles of incidence or normal (or near-normal) angles of incidence.

[0059] In some examples, the optical-based subsystem 201 can be configured to direct light to the sample 202 at more than one incident angle at the same time. For example, the illumination subsystem can include more than one illumination channel, one of which can include, for example, Figure 3, and another of the illumination channels (not shown) may include similar elements that may be configured differently or identically or may include at least a light source and possibly one or more other components (such as those described further herein). If this light is directed to the sample at the same time as other light, one or more characteristics (e.g., wavelength, polarization, etc.) of the light directed to the sample 202 at different incident angles may be different, so that light generated by illuminating the sample 202 at different incident angles can be distinguished from each other at a detector.

[0060] In another example, the illumination subsystem may include only one light source (e.g., Figure 3 ) and the light from the light source may be separated into different optical paths (e.g., based on wavelength, polarization, etc.) by one or more optical elements (not shown) of the illumination subsystem. The light in each of the different optical paths may then be directed to the sample 202. The multiple illumination channels may be configured to direct light to the sample 202 at the same time or at different times (e.g., when different illumination channels are used to sequentially illuminate the sample). In another example, the same illumination channel may be configured to direct light with different characteristics to the sample 202 at different times. For example, in some examples, the optical element 204 may be configured as a spectral filter and the properties of the spectral filter may be changed in a variety of different ways (e.g., by swapping out the spectral filter) so that light of different wavelengths may be directed to the sample 202 at different times. The illumination subsystem may have any other suitable configuration known in the art for directing light with different or the same characteristics to the sample 202, either sequentially or simultaneously, at different or the same angles of incidence.

[0061] In one embodiment, the light source 203 may include a broadband plasma (BBP) source. In this way, the light generated by the light source 203 and directed to the sample 202 may include broadband light. However, the light source may include any other suitable light source, such as a laser. The laser may include any suitable laser known in the art and may be configured to generate light of one or more any suitable wavelengths known in the art. In addition, the laser may be configured to generate monochromatic or near-monochromatic light. In this way, the laser may be a narrow-band laser. The light source 203 may also include a polychromatic light source that generates multiple discrete wavelengths or bands of light.

[0062] Light from optical element 204 can be focused onto sample 202 by lens 205. Figure 3 2 is shown as a single refractive optical element, but it should be understood that in reality, lens 205 may include several refractive and / or reflective optical elements that combine to focus light from the optical element to the sample. Figure 3The illumination subsystem shown in and described herein may include any other suitable optical elements (not shown). Examples of such optical elements include, but are not limited to, polarizing components, spectral filters, spatial filters, reflective optical elements, apodizers, beam splitters (such as beam splitter 213), apertures, and the like which may include any such suitable optical elements known in the art. Additionally, the optics-based subsystem 201 may be configured to alter one or more of the elements of the illumination subsystem based on the type of illumination to be used to generate the optics-based output.

[0063] The optical-based subsystem 201 may also include a scanning subsystem configured to scan light over the sample 202. For example, the optical-based subsystem 201 may include a stage 206 on which the sample 202 is placed during the generation of the optical-based output. The scanning subsystem may include any suitable mechanical and / or robotic assembly (including the stage 206) that may be configured to move the sample 202 so that the light may be scanned over the sample 202. Additionally or alternatively, the optical-based subsystem 201 may be configured so that one or more optical elements of the optical-based subsystem 201 perform some scanning of the light over the sample 202. The light may be scanned over the sample 202 in any suitable manner, such as in a serpentine path or in a spiral path.

[0064] The optical-based subsystem 201 further includes one or more detection channels. At least one of the one or more detection channels includes a detector configured to detect light from the sample 202 due to illumination of the sample 202 by the subsystem and to generate an output in response to the detected light. For example, Figure 3 The optical-based subsystem 201 shown in FIG. 1 includes two detection channels, one detection channel formed by a light collector 207, an element 208, and a detector 209, and another detection channel formed by a light collector 210, an element 211, and a detector 212. Figure 3 , the two detection channels are configured to collect and detect light at different collection angles. In some examples, the two detection channels are configured to detect scattered light, and the detection channels are configured to detect light scattered at different angles from the sample 202. However, one or more of the detection channels may be configured to detect another type of light (e.g., reflected light) from the sample 202.

[0065] like Figure 3As further shown in , the two detection channels are shown as being positioned in the plane of the paper and the illumination subsystem is also shown as being positioned in the plane of the paper. Thus, in this embodiment, the two detection channels are positioned in (e.g., centered in) the plane of incidence. However, one or more of the detection channels may be positioned outside the plane of incidence. For example, the detection channel formed by the light collector 210, the element 211, and the detector 212 may be configured to collect and detect light scattered out of the plane of incidence. Thus, this detection channel may generally be referred to as a "side" channel, and this side channel may be centered in a plane that is substantially perpendicular to the plane of incidence.

[0066] although Figure 3 An embodiment of an optical-based subsystem 201 is shown that includes two detection channels, but the optical-based subsystem 201 may include a different number of detection channels (e.g., only one detection channel or two or more detection channels). In one such example, the detection channel formed by the light collector 210, the element 211, and the detector 212 may form one side channel as described above, and the optical-based subsystem 201 may include an additional detection channel (not shown) formed as another side channel positioned on the opposite side of the incident plane. Thus, the optical-based subsystem 201 may include a detection channel that includes the light collector 207, the element 208, and the detector 209 and is centered in the incident plane and is configured to collect and detect light at a scattering angle that is at or close to normal to the surface of the sample 202. Thus, this detection channel may generally be referred to as a "top" channel, and the optical-based subsystem 201 may also include two or more side channels configured as described above. Thus, the optical-based subsystem 201 may include at least three channels (i.e., one top channel and two side channels), and each of the at least three channels has its own collector, each of which is configured to collect light at a scattering angle different from each of the other collectors.

[0067] As further described above, each of the detection channels included in the optical-based subsystem 201 can be configured to detect scattered light. Figure 3 The optical-based subsystem 201 shown in can be configured for dark field (DF) output generation of the sample 202. However, the optical-based subsystem 201 may additionally or alternatively include a detection channel configured for bright field (BF) output generation of the sample 202. In other words, the optical-based subsystem 201 may include at least one detection channel configured to detect light specularly reflected from the sample 202. Thus, the optical-based subsystem 201 described herein may be configured for DF only, BF only, or both DF and BF imaging. Although in Figure 3Each of the light collectors is shown in as a single refractive optical element, but it should be understood that each of the light collectors may include one or more refractive optical dies and / or one or more reflective optical elements.

[0068] One or more detection channels may include any suitable detector known in the art. For example, the detector may include a photomultiplier tube (PMT), a charge coupled device (CCD), a time delay integration (TDI) camera, and any other suitable detector known in the art. The detector may also include a non-imaging detector or an imaging detector. In this way, if the detector is a non-imaging detector, each of the detectors may be configured to detect certain characteristics of the scattered light (such as intensity), but may not be configured to detect such characteristics that vary depending on the position within the imaging plane. Thus, the output generated by each of the detectors included in each of the detection channels of the optical-based subsystem may be a signal or data, rather than an image signal or image data. In such examples, a processor (such as processor 214) may be configured to generate an image of the sample 202 from the non-imaging output of the detector. However, in other examples, the detector may be configured as an imaging detector configured to generate an imaging signal or image data. Thus, the optical-based subsystem may be configured to generate an optical image or other optical-based output described herein in a number of ways.

[0069] It should be noted that this article provides Figure 3 To generally illustrate configurations of an optical-based subsystem 201 that may be included in or that may produce an optical-based output used by system embodiments described herein. The optical-based subsystem 201 configurations described herein may be altered to optimize the performance of the optical-based subsystem 201 as is typically performed when designing a commercial output acquisition system. Additionally, the systems described herein may be implemented using existing systems (e.g., by adding the functionality described herein to the existing system). For some such systems, the methods described herein may be provided as optional functionality of the system (e.g., in addition to other functionality of the system). Alternatively, the systems described herein may be designed as completely new systems.

[0070] The processor 214 may be coupled to the components of the system 200 in any suitable manner (e.g., via one or more transmission media, which may include wired and / or wireless transmission media) such that the processor 214 may receive output. The processor 214 may be configured to use the output to perform a number of functions. The system 200 may receive instructions or other information from the processor 214. The processor 214 and / or the electronic data storage unit 215 may optionally be in electronic communication with a wafer inspection tool, a wafer metrology tool, or a wafer review tool (not illustrated) to receive additional information or send instructions. For example, the processor 214 and / or the electronic data storage unit 215 may be in electronic communication with a scanning electron microscope.

[0071] The processor 214, other systems, or other subsystems described herein may be part of a variety of systems, including a personal computer system, an image computer, a mainframe computer system, a workstation, a network appliance, an Internet appliance, or other devices. The subsystem or system may also include any suitable processor known in the art, such as a parallel processor. In addition, the subsystem or system may include a platform with high-speed processing and software as a stand-alone tool or a networked tool.

[0072] Processor 214 and electronic data storage unit 215 may be disposed in or otherwise be part of system 200 or another device. In an example, processor 214 and electronic data storage unit 215 may be part of an independent control unit or in a centralized quality control unit. Multiple processors 214 or electronic data storage units 215 may be used.

[0073] Processor 214 may be implemented in practice by any combination of hardware, software, and firmware. In addition, its functions as described herein may be performed by one unit or divided among different components, each of which may be implemented by any combination of hardware, software, and firmware. Program code or instructions for processor 214 to implement various methods and functions may be stored in a readable storage medium such as a memory in electronic data storage unit 215 or other storage.

[0074] If the system 200 includes more than one processor 214, different subsystems may be coupled to each other so that images, data, information, instructions, etc. may be sent between the subsystems. For example, one subsystem may be coupled to an additional subsystem via any suitable transmission media that may include any suitable wired and / or wireless transmission media known in the art. Two or more such subsystems may also be operatively coupled via a shared computer-readable storage medium (not shown).

[0075] The processor 214 may be configured to use the output of the system 200 or other output to perform a number of functions. For example, the processor 214 may be configured to send the output to the electronic data storage unit 215 or another storage medium. The processor 214 may be configured according to any of the embodiments described herein. The processor 214 may also be configured to use the output of the system 200 or use images or data from other sources to perform other functions or additional steps.

[0076] The various steps, functions and / or operations of the system 200 and the methods disclosed herein are implemented by one or more of the following: electronic circuits; logic gates; multiplexers; programmable logic devices; ASICs; analog or digital controls / switches; microcontrollers; or computing systems. Program instructions implementing methods such as those described herein may be transmitted via or stored on a carrier medium. The carrier medium may include storage media such as read-only memory, random access memory, magnetic or optical disks, non-volatile memory, solid-state memory, magnetic tape, and the like. The carrier medium may include transmission media such as wires, cables, or wireless transmission links. For example, the various steps described throughout the present disclosure may be implemented by a single processor 214 or, alternatively, multiple processors 214. In addition, different subsystems of the system 200 may include one or more computing or logic systems. Therefore, the above description should not be construed as limiting the present disclosure but is merely illustrative.

[0077] In an example, processor 214 is in communication with system 200. Processor 214 is configured to generate a target image of a workpiece based on information from detector 209 and / or detector 212. The die may include at least one SIP device, FFC device, or 3DIC device. Processor 214 also receives a design file including a design of the die. The design file may be a graphic design system file. A reference optical image of the die is generated based on the design file by processor 214 using a deep convolutional neural network for image-to-image translation. The deep convolutional neural network may be a CycleGAN. Processor 214 is also configured to: subtract the reference optical image from the target image, thereby generating a difference image; generate a run-time care region mask of the die based on the design file including the design of the die; apply the run-time care region mask to the difference image, thereby generating a masked difference image; and apply a threshold to the masked difference image, thereby generating a binarized defect image.

[0078] The processor may be further configured to align the target image and the reference optical image prior to subtracting.

[0079] The processor may be further configured to extract the illuminated region image from the target image using the execution-time illuminated region mask.

[0080] Additional embodiments relate to a non-transitory computer-readable medium storing program instructions executable on a controller for performing a computer-implemented method for testing, as disclosed herein. In particular, Figure 3 , the electronic data storage unit 215 or other storage medium may contain non-transitory computer-readable media including program instructions executable on the processor 214. The computer-implemented method may include any steps of any method described herein, including method 100.

[0081] Program instructions may be implemented in any of a variety of ways, including procedural, component-based, and / or object-oriented techniques, etc. For example, program instructions may be implemented using ActiveX controls, C++ objects, JavaBeans, Microsoft Foundation Classes (MFC), Streaming SIMD Extensions (SSE), or other techniques or methodologies as desired.

[0082] Each of the steps of the method may be performed as described herein. The method may also include any other steps that may be performed by a processor and / or computer subsystem or system described herein. The steps may be performed by one or more computer systems that may be configured according to any of the embodiments described herein. In addition, the method described above may be performed by any of the system embodiments described herein.

[0083] Although the present disclosure has been described with respect to one or more specific embodiments, it will be appreciated that other embodiments of the present disclosure may be made without departing from the scope of the present disclosure. Therefore, the present disclosure is deemed to be limited only by the appended claims and reasonable interpretation thereof.

Claims

1. A method comprising: receiving, at a processor, a target image of a workpiece including a die having a plurality of chips; receiving, at the processor, a design file including a design for the die; generating, using the processor, a reference optical image of the die based on the design file using a deep convolutional neural network for image-to-image translation; subtracting the reference optical image from the target image using the processor, thereby producing a difference image; generating, using the processor, a run-time aware region mask for the die based on the design file including the design of the die; applying, using the processor, the execution-time-sensitive region mask to the difference image, thereby generating a masked difference image; and A threshold is applied to the masked difference image using the processor, thereby producing a binarized defect image.

2. The method of claim 1, wherein the die comprises at least one system-in-package device.

3. The method of claim 1 wherein the die comprises a film frame carrier. The method of claim 1 , wherein the die comprises a 3D integrated circuit.

5. The method of claim 1, wherein the deep convolutional neural network is a recurrent generative adversarial network. The method of claim 1 , wherein the design file is a graphic design system file.

7. The method of claim 1, further comprising aligning the target image and the reference optical image prior to the subtracting.

8. The method of claim 1, further comprising extracting, with the processor, an illuminated region image from the target image using the execution-time illuminated region mask.

9. The method of claim 1, further comprising using an optical inspection system to generate the target image.

10. A non-transitory computer-readable medium storing a program configured to instruct the processor to execute the method according to claim 1.

11. A system comprising: a light source that generates a light beam; a stage configured to hold a workpiece in a path of the light beam, wherein the workpiece comprises a die having a plurality of chips; a detector configured to receive the light beam reflected from the workpiece; and a processor in electronic communication with the detector, wherein the processor is configured to: generating a target image of the workpiece based on information from the detector; receiving a design file including a design of the die; generating a reference optical image of the die based on the design file by a deep convolutional neural network for image-to-image translation; subtracting the reference optical image from the target image, thereby producing a difference image; generating a run-time care region mask for the die based on the design file including the design of the die; applying the execution-time attention region mask to the difference image, thereby generating a masked difference image; and A threshold is applied to the masked difference image, thereby producing a binarized defect image.

12. The system of claim 11, wherein the die comprises at least one system-in-package device, a film frame carrier, or a 3D integrated circuit.

13. The system of claim 11, wherein the deep convolutional neural network is a recurrent generative adversarial network.

14. The system of claim 11, wherein the design file is a graphic design system file.

15. The system of claim 11, wherein the processor is further configured to align the target image and the reference optical image prior to the subtracting.

16. The system of claim 11, wherein the processor is further configured to extract an illuminated region image from the target image using the execution-time illuminated region mask.

Images