Method and system for wafer image generation

Abstract

Embodiments of the present invention provide methods and systems for generating a die tensor of a wafer from a CAD file. It is necessary to have an image of the wafer that can be a faithful reproduction of a perfect reference image. Afterwards, the perfect reference image can be used to detect defects in the wafer by comparing the die in the perfect reference image with the die in the actual image of the wafer. The perfect reference image can be generated from a CAD file that includes a circuit design for the wafer. The generated perfect reference image of the wafer can be referred to as a die tensor. The die tensor can include a predefined number of slices, and the predefined number of slices can be used to improve the detection of defects in the die of the wafer.

Description

Technical Field

[0001] Embodiments herein relate to wafer inspection, and more particularly, to methods and systems for generating a die tensor of a wafer. Background Art

[0002] In order to imprint an integrated circuit (IC) on a wafer, a reticle (mask) may be prepared based on a computer-aided design (CAD) file including the design specifications of the IC. The reticle (mask) containing the chip design may be transferred to the wafer in a cyclic sequence through various semiconductor manufacturing processes such as deposition, photolithography, etching, diffusion, etc. Process variations may be caused by any of these process steps, and therefore, process variations need to be checked to detect defects. The detection of such defects may be referred to as wafer inspection.

[0003] In order to perform the inspection, an image of the wafer can be captured using an optical system. Thereafter, image processing techniques can be used to detect the presence of defects in the wafer. In an example, a wafer having multiple dies, ICs, or chips of the same type arranged in a repeating pattern is considered. The grayscale value of a die can be compared with the grayscale value of another adjacent die. If the comparison value exceeds a predefined threshold, it can be concluded that there is a defect in the die. A similar strategy can be used to detect defects within a specific die. The grayscale values ​​of adjacent dies can be used as reference values ​​for detecting defects. However, if the reference value itself is defective, the detected defect may be erroneous. This may lead to the propagation of false detections. In addition, inspecting a wafer by comparing the grayscale values ​​of all dies in a wafer with the grayscale values ​​of their respective adjacent dies can be a time-consuming process. Summary of the invention

[0004] The main purpose of the embodiments of this document is to disclose a method and system for generating a multi-dimensional image of a wafer based on a CAD file, wherein the CAD file includes properties of the imaging and manufacturing processes used for granular recipe creation, fine defect detection in the wafer, information related to the wafer manufacturing process, optical parameters from an inspection tool or camera, wafer metrology, etc.

[0005] Thus, embodiments provide methods and systems for generating a die tensor of a wafer from a CAD file. Embodiments include generating a reference image that can be used to detect defects in a wafer by comparing the die in the reference image to the die in an actual image of the wafer. The reference image can be generated from a CAD file including a circuit design of the wafer. The generated reference image of the wafer can be referred to as a die tensor. The die tensor can include a predefined number of slices, and using the predetermined number of slices can result in accurate detection of defects in the die of the wafer.

[0006] When considered in conjunction with the following description and the accompanying drawings, these and other aspects of the embodiments herein will be better understood and appreciated. However, it should be understood that although the following description indicates embodiments and many of their specific details, the following description is provided by way of illustration and not limitation. Various changes and modifications may be made within the scope of the embodiments herein without departing from the spirit of the present invention, and the embodiments herein include all such modifications. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present invention is illustrated in the accompanying drawings, and like reference numerals indicate corresponding parts in the various drawings throughout the drawings. The embodiments of the present invention will be better understood from the following description with reference to the accompanying drawings, in which:

[0008] Figure 1 depicts a slice of a die tensor generated according to embodiments disclosed herein; and

[0009] Figure 2 Various units of an apparatus for generating a die tensor according to embodiments disclosed herein are depicted. DETAILED DESCRIPTION

[0010] The embodiments herein and their various features and advantageous details are explained more fully with reference to the non-limiting embodiments shown in the accompanying drawings and described in detail in the following description. Descriptions of known components and processing techniques are omitted so as not to unnecessarily confuse the embodiments herein. The examples herein are intended only to facilitate understanding of the manner in which the embodiments herein may be implemented and to further enable those skilled in the art to implement the embodiments herein. Therefore, the examples should not be construed as limiting the scope of the embodiments herein.

[0011] Embodiments herein disclose methods and systems for generating a die tensor of a wafer from a computer-aided design (CAD) file. Embodiments include generating a reference image that can be used to detect defects in a wafer by comparing the die in the reference image to the die in an actual image of the wafer. The reference image can be generated from a CAD file including a circuit design of the wafer. The generated reference image of the wafer can be referred to as a die tensor. The die tensor can include a predefined number of slices, and using the predefined number of slices can result in accurate detection of defects in the die of the wafer.

[0012] Referring now to the drawings, and more particularly to the Figure 1 and Figure 2 (wherein like reference numerals indicate corresponding features throughout the drawings), a preferred embodiment is shown.

[0013] Figure 1 Depicts a slice of a die tensor generated according to embodiments disclosed herein. Figure 1As shown, in one example, the die tensor includes four slices. The slices are referred to herein as slices 0, 1, 2, and 3. Slice 0 can be a generated reference image. The generated reference image can be used to detect defects in a wafer. Defects can be detected by comparing the die in the die tensor with the die in the actual image of the wafer. As the number of slices of the die tensor increases, the accuracy of wafer inspection increases.

[0014] Multiple images of the wafer can be captured by changing the optical conditions. At a specific location of the wafer, the grayscale value at a specific pixel in each of the multiple images can be different. This can provide a distribution (histogram) of grayscale values ​​of a specific pixel at a specific location of the die tensor. Slice 1 can show the grayscale distribution of the generated image (die tensor).

[0015] The grayscale values ​​of pixels at specific locations of the die tensor can be compared with the grayscale values ​​of pixels of an actual image of the wafer. An embodiment includes setting a predefined threshold difference between the grayscale values ​​of a reference image and an actual image for pixels at different locations. If the grayscale difference for a specific pixel at a certain location is greater than the predefined threshold difference set at that location, it can be concluded that there may be a defect in the die in that location of the wafer. Therefore, it is critical to set a predefined threshold difference for different locations of the die tensor. The predefined threshold difference can be set based on the grayscale distribution of the pixels in the die tensor.

[0016] The grayscale distribution can provide information about the sensitivity of the pixel to various noisy optical conditions, preventing false detection of defects, etc. This helps to set a predefined threshold difference for each pixel. Information obtained from other slices can be considered for setting the predefined threshold difference.

[0017] Slice 2 can provide the location of different dies in the wafer. Wafer die index and shot index can be used to indicate the location of the dies in the wafer. In one example, Figure 1 As shown, the wafer die index of the die can be (1, 0) and the exposure unit index of the die can be (2, 2). Based on the optical conditions, different parts of the reference image depicting different locations of the wafer may look different. This may be due to the difference between the grayscale values ​​of the pixels at the center of the wafer and the grayscale values ​​of the pixels at the edge of the wafer. In addition, the die at the center of the wafer may have different characteristics compared to the die at the edge of the wafer. The characteristic differences may occur during the processing steps of manufacturing. Therefore, different threshold difference values ​​can be set to detect defects in the die at different parts (locations) of the die tensor.

[0018] Slice 3 provides the variation in thickness across the wafer. The variation in thickness may affect the grayscale values ​​of pixels at different locations of the die tensor. A predefined threshold difference may be set to account for the thickness variation. Embodiments allow for the inclusion of additional slices that (in addition to the grayscale value difference between pixels, the location of the die in the wafer, the thickness of the wafer) may account for additional variations that may affect the quality of the wafer.

[0019] Figure 2 Depicted are various units of an apparatus 200 for generating a die tensor 224 according to embodiments disclosed herein. The apparatus 200 includes an image filter 211 that can generate a die tensor 224 having a predefined number of slices based on inputs from a CAD file 201, design specifications 204, process information 205, metrology 210, and optical information 209. In an example, the die tensor 224 can include four slices, namely 101, 102, 103, and 104. The image filter 211 can iteratively generate the die image 224 through recipe adjustments using a feedback unit 225.

[0020] CAD file 201 may be one of the inputs to image filter 211. CAD file 201 may include a complete design architecture of wafer 226. CAD file 201 may also include logic circuits, connections between circuits, locations of dies, etc. Wireframe generator 202 may read CAD file 201 and generate wireframe image 203. Wireframe image 203 may depict a simulated circuit diagram.

[0021] The metrology 210 may be specific to the product being manufactured. The metrology 210 may be input obtained from measurements of metrology parameters on wafers of similar products that have been produced earlier. The metrology 210 may also include the effects of specific process steps (e.g., lithography, diffusion, etching, etc.) on wafers of similar products. Sensor values ​​for wafer process steps, previous metrology readings, process event data (e.g., recipe changes) may be used to predict metrology information. Machine learning methods may be used to combine these values ​​to derive predicted values ​​for metrology readings.

[0022] The design specification 204 may include a design of a circuit of a die to be imprinted on the wafer 226. The design specification 204 may be part of the CAD file 201. When the CAD file 201 is converted to an image, the design specification 204 may be lost. An embodiment includes extracting the design specification 204 and providing the design specification 204 as input to the image filter 211. The design specification 204 may be extracted from the CAD file 201 using a parser tool.

[0023] The graticule may be masked on wafer 226 while die tensor 224 is generated, and an image of wafer 228 may be captured under specific optical conditions in feedback unit 225. There may be a plurality of process equipment that may perform chemical processes on wafer 226. The process may be etching, lithography, deposition, etc. An equipment that performs at least one of these processes may be referred to as a process equipment (process tool). In one example, an etcher tool may drill patterns at different locations on wafer 226, a deposition tool may deposit chemicals or compounds on wafer 226 in some manner, a lithography tool may print different types of patterns on wafer 226, and so on.

[0024] Process information 205 may be information obtained from a plurality of process equipment. When performing a chemical process, the process equipment may collect various types of data. Process information 205 may include process event data 206, data 207 collected by sensors in the process equipment, metadata 208, and the like. Process information 205 may be obtained by an image filter 211 as one of the inputs for generating a die tensor 224. In one example, consider that a wafer is placed in a chamber during one of the processes. Process event data 206 may be an increase or decrease in pressure inside a chamber (process equipment). Data 207 measured by sensors in the process equipment may be temperature, pressure, current, voltage, impedance, and the like. Data 207 collected from sensors may be represented in a time series. Metadata 208 may be information related to wafer products, recipes, equipment, chambers, and the like.

[0025] The optical information 209 includes optical parameters and conditions that may affect the image of the wafer being captured (wafer image) 228. The optical information 209 may be used as an adjustable recipe parameter, which may be obtained by the image filter 211 to generate the die tensor 224. The die tensor 224 may be fed to the feedback unit 225, which may perform image operations to generate an error image 229. The feedback unit 225 may obtain a wafer image 228 from a wafer 226 using an image capture device 227. Image operations may be performed on the wafer image 228 and the die tensor 224 to generate an error image 229. Based on the error image 229 (feedback), recipe adjustments may be performed and the optical information 209 may be updated to regenerate the die tensor 224. Thus, the die tensor 224 may be iteratively generated until the die tensor 224 finally generated is a perfect reference image to be used for inspecting the wafer 226.

[0026] The image filter 211 may include a segmentation engine 212 that may segment the wireframe image 203 into a plurality of unique entities, wherein each unique entity may have a die pattern that may be repetitive in nature. The unique entities may be transformed in parallel before each of the unique entities are stitched together again. The segments (unique entities) may overlap to avoid boundary artifacts when stitching.

[0027] The image filter 211 may include an image transformation engine 213. The image transformation engine 213 may obtain inputs from the CAD file 201, the design specification 204, the process information 205, the metrology 210, and the optical information 209, either directly or through other units in the image filter 211. The image transformation engine 213 may use the inputs to transform the segmented parts in parallel.

[0028] The transformation can be performed using methods designed in turn using machine learning methods. In one embodiment, during training, slice 0 can be compared to a wafer image 228 taken on a physical wafer using an image capture device 227 in a feedback unit 225 to optimize recipe parameters. Based on the comparison, an error image 229 is generated. The error image 229 is fed back to the image filter 211. The error image can be used as a tool during machine learning to generate a method for transforming the segmented portions. The image filter 211 can include a stitching engine 214 that can stitch the transformed segmented portions to generate a die tensor 224.

[0029] Design specifications 204 obtained from CAD file 201, and metrology 210 may be used by image transformation engine 213 to generate relevant features of the die captured in die tensor 224. Image filter 211 may include clustering unit 215 and mapping unit 216. Clustering unit 215 may cluster features of the die that may have similar effects on the appearance of die tensor 224. Mapping unit 216 may assign a code or key ID to each cluster.

[0030] The image filter 211 may include another clustering unit 221 and another mapping unit 222. The another clustering unit 221 may cluster optical parameters that may have a similar effect on the appearance of the die tensor 224. Clustering is performed to combine optical parameters with similar effects. The another mapping unit 222 may assign a code or key ID to each cluster.

[0031] The image filter 211 may include an optical inspector 223. The optical inspector 223 may receive a code or key ID for a cluster formed based on features and optical parameters of dies that have a similar impact on the appearance of the die tensor 224. The optical inspector 223 may also receive a performance history of the generated die tensor 224 and similar imprinted wafers. Based on the received input, the optical inspector 223 may update the optical information 209. Thereafter, the updated optical information 209 may be used to generate an updated die tensor 224 in a subsequent iteration.

[0032] The image filter 211 may include a decorrelator 217 , a time series encoder 218 , a grouping unit 219 , and a code regulator 220 .

[0033] The time series encoder 218 may convert the process event data 206 and the data 207 (time series or signal data) collected by the sensor in the process equipment into a code vector. The time series encoder 218 may encode the process information 205 such as the process event data 206 and the data 207 collected by the sensor into a feature space, that is, perform feature extraction on the process information 205. Encoding may be performed by various machine learning methods such as autoencoder (AE), principal component analysis (PCA), restricted Boltzmann machine (RBM), convolutional neural network (CNN), recurrent neural network (RNN), etc. The grouping unit 219 may group similar process information 205 based on the metadata 208. The grouped process information may be fed to the time series encoder 218.

[0034] The decorrelator 217 may obtain the process information 205, such as the process event data 206 and the data 207 collected by the sensor. The process event data 206 and the data 207 collected by the sensor may then be combined to calculate decorrelative information.

[0035] The code conditioner 220 can adjust the extracted features so that repeated information (if any) in the extracted features can be ignored. This can reduce memory and computational costs. The adjustment can be performed based on the calculated decorrelation information. The code conditioner 220 can report the health of the process to the image transformation engine 213 based on the process information 205 obtained through the time series encoder 218 and the decorrelator 217. The result (health) of the process performed on the wafer 226 is a key factor that may affect the appearance of the generated die tensor 224. In one example, consider the process of lithography; if sufficient temperature cannot be applied, the required resist thickness may not be achieved. This may affect the appearance of the die tensor 224.

[0036] The image transformation engine 213 can obtain the process information 205, the design specification 204, and the optical information 209 through the code conditioner 220; and directly obtain the metrology 210 for transforming the segmented entities in parallel.

[0037] The embodiments disclosed herein may be implemented by at least one software program running on at least one hardware device and performing network management functions to control network elements. Figure 2 The illustrated network elements include blocks that may be at least one of hardware devices or a combination of hardware devices and software modules.

[0038] The embodiments disclosed herein describe methods and systems for generating wafer images from CAD files. Therefore, it should be understood that the scope of protection extends to such a program, and in addition to the computer-readable module having a message therein, such a computer-readable storage module includes: a program code module for implementing one or more steps of the method when the program is running on a server or a mobile device or any suitable programmable device. The method is implemented in a preferred embodiment by a software program written in, for example, a very high speed integrated circuit hardware description language (VHDL) (another programming language) or implemented together with the software program, or the method is implemented by one or more VHDL or several software modules executed on at least one hardware device. The hardware device can be any type of portable device that can be programmed. The device can also include the following modules, which can be, for example: a hardware module (e.g., an ASIC), or a combination of hardware and software modules (e.g., an ASIC and an FPGA, or at least one microprocessor and at least one memory having a software module located therein). The method embodiments described herein can be implemented partially in hardware and partially in software. Alternatively, the present invention can be implemented on different hardware devices (e.g., using multiple CPUs).

[0039] The above description of the specific embodiments fully reveals the general nature of the embodiments herein, so that other technicians can easily modify and / or adjust these specific embodiments for various applications without departing from the overall concept by applying existing knowledge, and therefore such adjustments and modifications should and are intended to be included in the meaning and scope of the equivalents of the disclosed embodiments. It is understood that the phrases or terms used herein are used for descriptive purposes and are not restrictive. Therefore, although the embodiments herein have been described according to preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced by modification within the spirit and scope of the embodiments as described herein.

Claims

1. A method for generating a die tensor of a wafer according to a computer-aided design (CAD) file, the method comprising: The wireframe image obtained according to the CAD file is segmented into a plurality of entities by a segmentation engine; performing, by an image transformation engine, a transformation on each of the plurality of entities based on the optical information; as well as A die tensor including a predefined number of slices is generated by the splicing engine by combining each of the transformed multiple entities, where the slice comprising a reference image, a grayscale distribution of the reference image, locations of dies in the wafer, and thickness variations across the wafer, wherein the image transformation engine further performs a transformation on each of the plurality of entities based on process information, the process information comprising process event data, data collected by a sensor in at least one process device, and metadata of at least one process, the process event data comprising an increase or decrease in pressure inside the process device, the data collected by the sensor comprising temperature, pressure, current, voltage, and impedance, the metadata comprising information related to the process device and recipe, and The process event data and the data collected by the sensor are obtained by a time series encoder to perform feature extraction on the process information; and the process information is grouped based on the metadata.

2. The method according to claim 1, wherein: Features of the die generated according to the design specification are clustered based on their influence on the appearance of the die tensor.

3. The method according to claim 1, wherein: Optical parameters and optical conditions in the optical information are clustered based on their impact on the appearance of the die tensor.

4. The method according to claim 1, wherein: The health status of the at least one process is provided to the image transformation engine by a code conditioner based on the feature extracted and grouped process information.

5. The method according to claim 1, wherein: The optical information is updated by an optical inspector based on an error image received from a feedback unit, wherein the error image is generated by performing an image operation between the generated die tensor and an image of the wafer.

6. A device for generating a die tensor of a wafer according to a computer-aided design (CAD) file, the device comprising: a segmentation engine configured to segment the wireframe image obtained from the CAD file into a plurality of entities; an image transformation engine configured to perform a transformation on each of the plurality of entities based on the optical information; as well as A concatenation engine is configured to generate a die tensor including a predefined number of slices by combining each of the transformed plurality of entities, wherein The slice includes a reference image, a grayscale distribution of the reference image, locations of dies in a wafer, and thickness variations across the wafer, wherein the image transformation engine further performs a transformation on each of the plurality of entities based on process information, the process information comprising process event data, data collected by a sensor in at least one process device, and metadata of at least one process, the process event data comprising an increase or decrease in pressure inside the process device, the data collected by the sensor comprising temperature, pressure, current, voltage, and impedance, the metadata comprising information related to the process device and recipe, and The apparatus further includes a time series encoder for obtaining the process event data and data collected by sensors to perform feature extraction on the process information; and the process information is grouped based on the metadata.

7. The device according to claim 6, wherein: Features of the die generated from the design specification are clustered based on their influence on the appearance of the die tensor.

8. The device according to claim 6, wherein: Optical parameters and optical conditions in the optical information are clustered based on their impact on the appearance of the die tensor.

9. The device according to claim 6, wherein: The health status of the at least one process is provided to the image transformation engine by a code conditioner based on the feature extracted and grouped process information.

10. The device according to claim 6, wherein: The optical information is updated by an optical inspector based on an error image received from a feedback unit, wherein the error image is generated by performing an image operation between the generated die tensor and an image of the wafer.

Images