WO2020246149A1 - 半導体検査装置及び半導体検査方法 - Google Patents
半導体検査装置及び半導体検査方法 Download PDFInfo
- Publication number
- WO2020246149A1 WO2020246149A1 PCT/JP2020/016739 JP2020016739W WO2020246149A1 WO 2020246149 A1 WO2020246149 A1 WO 2020246149A1 JP 2020016739 W JP2020016739 W JP 2020016739W WO 2020246149 A1 WO2020246149 A1 WO 2020246149A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- cad
- optical
- optical image
- semiconductor device
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—Two-dimensional [2D] image generation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/39—Circuit design at the physical level
- G06F30/392—Floor-planning or layout, e.g. partitioning or placement
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30148—Semiconductor; IC; Wafer
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2210/00—Indexing scheme for image generation or computer graphics
- G06T2210/32—Image data format
Definitions
- the present disclosure relates to a semiconductor inspection device and a semiconductor inspection method for inspecting a semiconductor device.
- Patent Document 1 an image has been acquired by using a semiconductor device as an inspection target device (DUT: device under test), and various analyzes such as analysis of a failure location have been performed based on the image (Patent Document 1 below and Patent Document below). 2).
- an optical image such as an LSM image is made high in resolution to generate a reconstructed image
- a second CAD image is generated based on a plurality of layers of CAD data
- the reconstructed image is a second. Alignment with respect to 2 CAD images is disclosed. According to such a method, accurate alignment of optical images becomes possible.
- the embodiment has been made in view of such a problem, and provides a semiconductor inspection apparatus capable of accurately aligning an optical image acquired for a semiconductor device and a CAD image corresponding to the semiconductor device.
- the task is to do.
- the semiconductor inspection device of one embodiment of the present disclosure includes an optical detector that detects light from a semiconductor device and outputs a detection signal, an optical system that guides the light to the photodetector, and a semiconductor device based on the detection signal.
- An image generation unit that generates a first optical image, which is an optical image, a reception unit that accepts input of a first CAD image, and a first CAD image conversion process using the optical image as teacher data by machine learning.
- An image conversion unit that learns and converts the first CAD image into a second CAD image that resembles an optical image by a conversion process based on the learning result, and a position based on the optical image and the second CAD image. It is provided with an alignment unit for performing alignment.
- another aspect of the semiconductor inspection method of the present disclosure is a step of detecting light from a semiconductor device via an optical system and outputting a detection signal, and an optical image of the semiconductor device based on the detection signal.
- the step of generating the optical image of 1, the step of accepting the input of the first CAD image, and the conversion process of the first CAD image using the optical image as the teacher data are learned by machine learning, and the result of the learning is obtained.
- the conversion process based on the above includes a step of converting the first CAD image into a second CAD image resembling an optical image, and a step of aligning the optical image and the second CAD image.
- an optical image reflecting light from a semiconductor device and a first CAD image are acquired, and the first CAD image is converted based on the result of learning by machine learning. Is converted into a second CAD image that resembles an optical image, and the second CAD image and the optical image are aligned with each other.
- the pattern on the CAD image is converted so as to be closer to the optical image, and then the alignment is performed with the optical image. The accuracy can be improved.
- Other forms of the semiconductor inspection apparatus of the present disclosure include an optical detector that detects light from a semiconductor device and outputs a detection signal, an optical system that guides the light to the optical detector, and a semiconductor device based on the detection signal.
- An image generation unit that generates a first optical image, which is an optical image of the above, a reception unit that accepts an input of the first CAD image, and a reconstruction of the first optical image using the first CAD image as teacher data.
- An image conversion unit that learns the processing by machine learning and reconstructs the first optical image into a second optical image that resembles the first CAD image by the reconstruction processing based on the learning result, and a second The optical image of the above and the alignment unit for aligning based on the first CAD image are provided.
- another aspect of the semiconductor inspection method of the present disclosure is a step of detecting light from a semiconductor device via an optical system and outputting a detection signal, and an optical image of the semiconductor device based on the detection signal.
- the step of generating the optical image of 1, the step of accepting the input of the first CAD image, and the reconstruction process of the first optical image using the first CAD image as teacher data are learned by machine learning.
- the optical image reflecting the light from the semiconductor device and the first CAD image are acquired, and the optical image is reconstructed based on the learning result by machine learning to perform the first CAD. It is converted into a second optical image that resembles an image, and the second optical image and the first CAD image are aligned.
- the unclear part on the optical image is converted so as to be closer to the CAD image, and then the position is aligned with the CAD image. The accuracy of alignment can be improved.
- the optical image acquired for the semiconductor device and the CAD image corresponding to the semiconductor device can be accurately aligned.
- FIG. 1 It is a schematic block diagram of the observation system 1 which concerns on embodiment.
- A is a diagram showing an example of a first optical image that is the target of the reconstruction process, and (b) is a second view generated by the reconstruction process of the first optical image of (a).
- the figure which shows an example of the optical image of (c) is a figure which shows an example of the first CAD image corresponding to the first optical image of (a).
- (A) is a diagram showing an example of the first optical image stored in the storage unit 27 of FIG. 1, and (b) is the first optical image of (a) stored in the storage unit 27 of FIG. It is a figure which shows the 1st CAD image corresponding to.
- FIG. 1 is a schematic configuration diagram of an observation system 1 which is a semiconductor inspection device according to an embodiment.
- the observation system 1 shown in FIG. 1 is an optical system that acquires and processes images such as heat-generating images of semiconductor devices in order to inspect semiconductor devices such as logic LSIs, ICs (integrated circuits) such as memories, and power devices. ..
- the observation system 1 includes an optical device (optical system) 13, an objective lens 15, and a stage 17 incorporating a plurality of detectors 3, a two-dimensional camera 5, a lighting device 7, a dichroic mirror 9, and a beam splitter 11 such as a half mirror. , Computer (Personal Computer) 19, tester 21, input device 23, and display device 25.
- Computer Personal Computer
- Each of the plurality of detectors 3 is a photodetector that detects light from the semiconductor device S mounted on the stage 17.
- the detector 3 may be an imaging device such as an InGaAs (indium gallium arsenide) camera or an InSb (indium antimonide) camera having sensitivity to infrared wavelengths.
- the detector 3 acquires an LSM (Laser Scanning Microscope) image or an EOFM (Electro Optical Frequency Mapping) image by detecting the reflected light while scanning the laser light two-dimensionally on the semiconductor device S. It may be a detection system that outputs a detection signal of.
- Each of the plurality of detectors 3 is switched to and optically connectable to the optical device 13, and detects the light from the semiconductor device S via the objective lens 15 and the dichroic mirror 9 in the optical device 13. ..
- the two-dimensional camera 5 is a camera incorporating a CCD (Charge Coupled Device) image sensor, a CMOS (Complementary Metal-Oxide Semiconductor) image sensor, and the like, and detects the reflected light from the semiconductor device S mounted on the stage 17. It is an optical detector that outputs a detection signal of a two-dimensional pattern image of a semiconductor device.
- the two-dimensional camera 5 detects a two-dimensional pattern image of the semiconductor device S via the objective lens 15, the dichroic mirror 9 in the optical device 13, and the beam splitter 11.
- the objective lens 15 is provided so as to face the semiconductor device S, and sets the magnification of the image formed on the plurality of detectors 3 and the two-dimensional camera 5.
- the objective lens 15 includes a plurality of built-in lenses having different magnifications, and has a function of switching the built-in lens that forms an image on the detector 3 or the two-dimensional camera 5 between a high-magnification lens and a low-magnification lens.
- the dichroic mirror 9 transmits light in a predetermined wavelength range in order to guide an image such as an emission image, a heat generation image, or a reflection image of the semiconductor device S to the detector 3, and two-dimensionally transmits a two-dimensional pattern image of the semiconductor device S.
- Light having a wavelength other than the predetermined wavelength range is reflected in order to guide the light to the camera 5.
- the beam splitter 11 transmits the pattern image reflected by the dichroic mirror 9 toward the two-dimensional camera 5, and reflects the illumination light for generating the two-dimensional pattern image emitted from the lighting device 7 toward the dichroic mirror 9.
- the illumination light is applied to the semiconductor device S via the dichroic mirror 9 and the objective lens 15.
- the tester 21 applies a predetermined electric signal test pattern, a predetermined voltage, or a predetermined current to the semiconductor device S. By applying this test pattern, light emission or heat generation due to a failure of the semiconductor device S is generated.
- the computer 19 is an image processing device (processor) that processes the detection signals acquired by the detector 3 and the two-dimensional camera 5.
- the computer 19 has a storage unit 27, an image processing unit (reception unit, image generation unit) 29, an image analysis unit (image conversion unit, alignment unit, output unit) 31, and control as functional components. It is composed of a unit 33.
- the computer 19 includes an input device 23 such as a mouse and a keyboard for inputting data to the computer 19, and a display device (output unit) 25 such as a display device for displaying the image processing result by the computer 19. It is attached.
- an arithmetic processing device such as a CPU of the computer 19 executes a computer program (image processing program) stored in a storage medium such as an internal memory of the computer 19 or a hard disk drive. It is a function to be realized.
- the arithmetic processing unit of the computer 19 causes the computer 19 to function as each functional unit of FIG. 1 by executing this computer program, and sequentially executes the semiconductor inspection processing described later.
- Various data necessary for executing the computer program and various data generated by executing the computer program are all stored in an internal memory such as a ROM or RAM of the computer 19 or a storage medium such as a hard disk drive.
- the storage unit 27 is the first in which a light emission image acquired by the detector 3, a measurement image in which a heat generation image or the like is detected, and a pattern image of the semiconductor device S acquired by the detector 3 or the two-dimensional camera 5 are detected.
- the optical image and the first CAD image showing the high-resolution pattern of the semiconductor device S created based on the CAD data acquired from the outside are stored.
- the first optical image is an image showing the optical measurement result of the two-dimensional pattern of the semiconductor device S, and may be an image of the two-dimensional pattern detected by the two-dimensional camera 5, or a detector. It may be an LSM image based on the detection signal detected by 3.
- the image processing unit 29 sequentially generates a measurement image and a first optical image based on a detection signal received from the detector 3 or the two-dimensional camera 5, and sequentially stores the measurement image and the first optical image.
- the image processing unit 29 receives input of CAD data from an external storage unit 35 constructed in an external PC, a server device, or the like via the network NW, generates a first CAD image from the CAD data, and stores the storage unit.
- This CAD data is based on design information regarding the layout of each layer such as the diffusion layer, the metal layer, the gate layer, and the element separation layer of the semiconductor device S, and is stored in an external PC or server device such as software called a layout viewer. Generated by.
- This CAD data is used as a first CAD image showing a pattern image of the semiconductor device S.
- the control unit 33 controls data processing in the computer 19 and processing of a device connected to the computer 19. Specifically, the control unit 33 emits illumination light by the illumination device 7, images by the plurality of detectors 3 and the two-dimensional camera 5, switches the connection of the plurality of detectors 3 to the optical device 13, and sets the objective lens 15. It controls switching of magnification, application of test patterns by tester 21, display of observation results by display device 25, and the like.
- the image analysis unit 31 performs reconstruction processing, pattern conversion processing, and alignment processing on various images sequentially stored in the storage unit 27. Details of the functions of each process of the image analysis unit 31 will be described below.
- the image analysis unit 31 generates a second optical image that resembles the first CAD image based on the first optical image described in the storage unit 27 (reconstruction processing).
- reconstruction processing a plurality of first optical images are used as training data, and the first CAD image corresponding to the semiconductor device S targeted by those images is used as teacher data, which is a kind of machine learning, deep learning. It is executed using the learning model obtained as a result of pre-training by.
- the learning model data obtained by the prior learning is stored in the storage unit 27 and is referred to during the subsequent reconstruction process.
- CNN Convolutional Neural Network
- FCN Frully Convolutional Networks
- U-Net ResNet (Residual Network)
- ResNet Residual Network
- FIG. 2 shows an example of an image processed by the image analysis unit 31, in which (a) is the first optical image GA 1 and (b) which are LSM images to be reconstructed.
- the second optical images GB 1 and (c) generated by the reconstruction process for the first optical image GA 1 of (a) are the first optical images corresponding to the first optical image of (a).
- the CAD image GC 1 is shown respectively. In this way, the low-resolution and unclear portion of the first optical image GA 1 is converted to be closer to the first CAD image GC 1 by the reconstruction process, and is resembled to the first CAD image GC 1 as a whole.
- the optical image GB 1 of 2 is generated.
- the image analysis unit 31 generates a second CAD optical image that resembles the first optical image based on the first CAD image described in the storage unit 27 (pattern conversion processing).
- pattern conversion processing a plurality of first CAD images were used as training data, and the first optical images acquired for the semiconductor device S corresponding to those images were used as teacher data, and were learned in advance by deep learning. It is executed using the resulting learning model.
- a first optical image obtained by cutting out an arbitrary range on the semiconductor device S and performing enlargement processing and a first CAD image obtained by cutting out a position corresponding to the range and performing enlargement processing. Use multiple combinations with.
- the data of the learning model obtained by such prior learning is stored in the storage unit 27 and referred to in the subsequent pattern conversion processing.
- CNN, FCN, U-Net, ResNet, etc. are used as the learning model for deep learning, but the learning model is not limited to a specific one, and the number of nodes and the number of layers of the learning model are arbitrary. Can be set to.
- the (a), shows a first example of an optical image GA 1 is stored LSM image in the storage unit 27, the (b), the first optical image GA 1 (a)
- the first CAD image GC 1 representing the pattern image of the semiconductor device S corresponding to is shown.
- not all the patterns of the first CAD image GC 1 appear in the image observed in the first optical image GA 1 , and some layers (gate layer or under the gate layer) do not appear.
- the pattern of the diffusion layer and the element separation layer, etc.) mainly appears.
- a shading image may appear in the corresponding portion of the pattern depending on how crowded the pattern of the first CAD image GC 1 is.
- FIG. 4 shows an image of an image generated by the image analysis unit 31 by the pattern conversion process for the first CAD image GC 1 .
- the image GC 4 is an image when a predetermined image correction process is directly applied to the first CAD image GC 1 .
- FIG. 5 shows an image of the pattern conversion process by the image analysis unit 31 using the first CAD image GC 1 .
- the first optical image is an optical observation from the surface of the semiconductor device S is generally although pattern P G of the gate layer is reflected, the pattern P D of the dummy is the process rule of the semiconductor device S first hardly appear in the optical image GA 1.
- the image analysis unit 31 in advance by learning, the process of extracting the pattern P D of the dummy from the first CAD image GC 1, may be constructed as a learning model. Accordingly, the image analysis unit 31 may convert the second to the CAD image GC 2 by removing the dummy pattern P D on the first CAD image GC 1.
- the diffusion layer and the element isolation layer disposed below the gate is sometimes causing the contrast of the pattern P G.
- the optical image is formed from different layers according to the change in the structure of the semiconductor device for each generation.
- the difference between the diffusion layer and the element separation layer may cause pattern contrast in the optical image, so the CAD image is preprocessed after appropriately extracting the information of the layer to be machine-learned. There is a need.
- the image analysis unit 31 includes a range on the semiconductor device S of the second optical image GB 1 generated by the reconstruction process described above and the second optical image GB 1 generated by the pattern conversion process described above. Based on the third CAD image GC 3 corresponding to the above, alignment with respect to each other's image positions is performed by pattern matching between images. Then, the image analysis unit 31 measures the measurement image in which light emission, heat generation, etc. are detected by measuring the same range as the first optical image GA 1 using the alignment result, and the first CAD image GC. 1 is superimposed and displayed on the display device 25.
- FIG. 6 is a flowchart showing the flow of pre-learning to generate a learning model of reconstruction processing
- FIG. 7 is a flowchart showing a flow of pre-learning to generate a learning model of pattern conversion processing
- FIG. It is a flowchart which shows the flow of the analysis process of a semiconductor device S.
- Step S01 when the computer 19 starts learning the reconstruction process at an arbitrary timing such as a user operation, the detector 3, the two-dimensional camera 5, the lighting device 7, and the like are controlled. As a result, a plurality of first optical images of the semiconductor device S, which are training data, are acquired and stored in the storage unit 27 (step S01). Next, the computer 19 acquires a plurality of CAD data corresponding to the first optical image from the outside, and based on the CAD data, the first CAD image to be the teacher data is acquired and stored in the storage unit 27. (Step S02).
- the image analysis unit 31 of the computer 19 constructs a learning model of the reconstruction process by deep learning. (Step S03). As a result, the data of the learning model acquired by the image analysis unit 31 is stored in the storage unit 27 (step S04).
- a plurality of CAD data corresponding to the plurality of types of semiconductor devices S are acquired from the outside, and these are obtained.
- a plurality of first CAD images to be training data are acquired based on the CAD data of the above and stored in the storage unit 27 (step S101).
- the computer 19 controls the detector 3, the two-dimensional camera 5, the lighting device 7, and the like, so that a plurality of first optical images corresponding to the first CAD images, which are teacher data, are acquired and stored in the storage unit. It is stored in 27 (step S102).
- the image analysis unit 31 of the computer 19 constructs a learning model of the pattern conversion process by deep learning. (Step S103). As a result, the data of the learning model acquired by the image analysis unit 31 is stored in the storage unit 27 (step S104).
- Step S201 the CAD data of the semiconductor device S is acquired by the computer 19, and the first CAD image is generated based on the CAD data and stored in the storage unit 27 (step S203).
- step S203 the process of steps S204 to S205 or the process of steps S206 to S207 is executed according to the user's selection input to the computer 19.
- the image analysis unit 31 of the computer 19 refers to the data of the learning model stored in the storage unit 27 and performs pattern conversion processing on the first CAD image, so that the second CAD image is performed. Is acquired (step S204). After that, the image analysis unit 31 performs alignment based on the first optical image and the third CAD image obtained by applying a predetermined image correction process to the second CAD image. Finally, the image analysis unit 31 superimposes the measurement image on the first CAD image and displays it on the display device 25 using the alignment result (step S205).
- the image analysis unit 31 refers to the data of the learning model stored in the storage unit 27 and performs reconstruction processing on the first optical image, so that the second optical image is acquired. (Step S206). After that, the image analysis unit 31 performs alignment based on the second optical image and the first CAD image. Finally, the image analysis unit 31 superimposes the measurement image on the first CAD image and displays it on the display device 25 using the alignment result (step S207).
- the first optical image reflecting the light from the semiconductor device S and the first CAD image showing the pattern image of the semiconductor device S are acquired, and the first CAD image is acquired.
- the CAD image 1 is converted into a second CAD image that resembles the first optical image by a conversion process based on the result of learning by machine learning, and the second CAD image and the first optical image are combined. Aligned.
- the semiconductor device S to which various process rules are applied is targeted for inspection, the pattern on the CAD image is converted so as to be closer to the optical image and then aligned with the optical image. The accuracy of the image can be improved.
- the first optical image is converted into a second optical image resembling the first CAD image by a reconstruction process based on the learning result by machine learning, and the second optical image is converted into the second optical image.
- the optical image and the first CAD image are aligned.
- the unclear portion on the optical image is converted so as to be closer to the CAD image, and then the image is aligned with the CAD image. The accuracy of alignment can be improved.
- deep learning is adopted as machine learning.
- the semiconductor device S to which various process rules are applied is targeted for inspection, either one of the CAD image and the optical image can be effectively converted so as to be closer to the other.
- the image analysis unit 31 of the computer 19 learns using the first optical image and the first CAD image at the positions corresponding to each other on the semiconductor device S. With such a function, it is possible to efficiently construct a conversion process or a reconstruction process for making either one of the CAD image and the optical image resemble the other, and it is possible to efficiently realize the alignment.
- a second CAD image can be generated by extracting a dummy pattern as a specific pattern from the first CAD image.
- the density of the dummy pattern and the pattern other than the dummy are different, so that the contrast in the optical image may be different.
- the generation or type of semiconductor device is learned by learning how the above dummy pattern is optically displayed. Even if the image changes, the optimum conversion for bringing the image closer to the optical image can be realized, so that the image analysis unit 31 that executes this conversion can efficiently realize highly accurate alignment.
- the image analysis unit 31 of the present embodiment has a function of superimposing and outputting the measurement image obtained corresponding to the first optical image and the first CAD image based on the alignment result. .. With such a function, the inspection position in the measurement image can be easily visually recognized.
- the image analysis unit 31 of the above embodiment superimposes and displays the measurement image and the first CAD image using the alignment result, but the first CAD image or the first CAD image is displayed. Only the CAD data on which the image is based may be displayed.
- the computer 19 may have a function of setting an analysis position to irradiate light based on the result of alignment on the displayed first CAD image or CAD data. Further, the computer 19 may have a function of superimposing and displaying a signal such as a light emission signal detected from the semiconductor device S on the CAD data based on the alignment result.
- a specific pattern may be extracted from the first CAD image by a conversion process to convert it into a second CAD image.
- the CAD image can be converted so as to effectively resemble the optical image, and highly accurate alignment can be efficiently realized.
- the above specific pattern may be a dummy pattern. In this case, more accurate alignment can be realized.
- the machine learning may be deep learning.
- the image can be effectively converted so that either one of the CAD image and the optical image is brought closer to the other.
- learning may be performed using the first optical image and the first CAD image at positions corresponding to each other on the semiconductor device.
- an output unit or a step may be further provided to superimpose and output the image obtained corresponding to the first optical image and the first CAD image based on the alignment result. If such an output unit or step is provided, the inspection position in the optical image can be easily visually recognized.
- a semiconductor inspection device for inspecting a semiconductor device and a semiconductor inspection method are used, and an optical image acquired for the semiconductor device and a CAD image corresponding to the semiconductor device can be accurately aligned. It is a thing.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Architecture (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
- Image Processing (AREA)
Abstract
Description
Claims (14)
- 半導体デバイスからの光を検出して検出信号を出力する光検出器と、
前記光を前記光検出器に導く光学系と、
前記検出信号に基づいて、前記半導体デバイスの光学画像である第1の光学画像を生成する画像生成部と、
第1のCAD画像の入力を受け付ける受付部と、
前記光学画像を教師データとして用いて前記第1のCAD画像の変換処理を機械学習によって学習し、当該学習の結果に基づいた前記変換処理により、前記第1のCAD画像を前記光学画像に似せた第2のCAD画像に変換する画像変換部と、
前記光学画像と前記第2のCAD画像に基づいて位置合わせを行う位置合わせ部と、
を備える半導体検査装置。 - 前記画像変換部は、前記変換処理により、前記第1のCAD画像から特定のパターンを抜き出すことで前記第2のCAD画像に変換する、
請求項1に記載の半導体検査装置。 - 前記特定のパターンはダミーパターンである、
請求項2に記載の半導体検査装置。 - 半導体デバイスからの光を検出して検出信号を出力する光検出器と、
前記光を前記光検出器に導く光学系と、
前記検出信号に基づいて、前記半導体デバイスの光学画像である第1の光学画像を生成する画像生成部と、
第1のCAD画像の入力を受け付ける受付部と、
前記第1のCAD画像を教師データとして用いて前記第1の光学画像の再構成処理を機械学習によって学習し、当該学習の結果に基づいた前記再構成処理により、前記第1の光学画像を前記第1のCAD画像に似せた第2の光学画像に再構成する画像変換部と、
前記第2の光学画像と前記第1のCAD画像に基づいて位置合わせを行う位置合わせ部と、
を備える半導体検査装置。 - 前記機械学習は、ディープラーニングである、
請求項1~4のいずれか1項に記載の半導体検査装置。 - 前記画像変換部は、前記半導体デバイス上の互いに対応する位置における前記第1の光学画像及び前記第1のCAD画像を用いて学習する、
請求項1~5のいずれか1項に記載の半導体検査装置。 - 前記位置合わせの結果を基に、前記第1の光学画像に対応して得られた画像と前記第1のCAD画像とを重畳して出力する出力部をさらに備える、
請求項1~6のいずれか1項に記載の半導体検査装置。 - 光学系を介して半導体デバイスからの光を検出して検出信号を出力するステップと、
前記検出信号に基づいて、前記半導体デバイスの光学画像である第1の光学画像を生成するステップと、
第1のCAD画像の入力を受け付けるステップと、
前記光学画像を教師データとして用いて前記第1のCAD画像の変換処理を機械学習によって学習し、当該学習の結果に基づいた前記変換処理により、前記第1のCAD画像を前記光学画像に似せた第2のCAD画像に変換するステップと、
前記光学画像と前記第2のCAD画像に基づいて位置合わせを行うステップと、
を備える半導体検査方法。 - 前記変換処理により、前記第1のCAD画像から特定のパターンを抜き出すことで前記第2のCAD画像に変換する、
請求項8に記載の半導体検査方法。 - 前記特定のパターンはダミーパターンである、
請求項9に記載の半導体検査方法。 - 光学系を介して半導体デバイスからの光を検出して検出信号を出力するステップと、
前記検出信号に基づいて、前記半導体デバイスの光学画像である第1の光学画像を生成するステップと、
第1のCAD画像の入力を受け付けるステップと、
前記第1のCAD画像を教師データとして用いて前記第1の光学画像の再構成処理を機械学習によって学習し、当該学習の結果に基づいた前記再構成処理により、前記第1の光学画像を前記第1のCAD画像に似せた第2の光学画像に再構成するステップと、
前記第2の光学画像と前記第1のCAD画像に基づいて位置合わせを行うステップと、
を備える半導体検査方法。 - 前記機械学習は、ディープラーニングである、
請求項8~11のいずれか1項に記載の半導体検査方法。 - 前記半導体デバイス上の互いに対応する位置における前記第1の光学画像及び前記第1のCAD画像を用いて学習する、
請求項8~12のいずれか1項に記載の半導体検査方法。 - 前記位置合わせの結果を基に、前記第1の光学画像に対応して得られた画像と前記第1のCAD画像とを重畳して出力するステップをさらに備える、
請求項8~13のいずれか1項に記載の半導体検査方法。
Priority Applications (6)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| SG11202113170XA SG11202113170XA (en) | 2019-06-03 | 2020-04-16 | Semiconductor inspection device and semiconductor inspection method |
| CN202080041571.3A CN113994372A (zh) | 2019-06-03 | 2020-04-16 | 半导体检查装置及半导体检查方法 |
| JP2021524694A JP7413375B2 (ja) | 2019-06-03 | 2020-04-16 | 半導体検査装置及び半導体検査方法 |
| EP20818448.1A EP3955208A4 (en) | 2019-06-03 | 2020-04-16 | SEMICONDUCTOR INSPECTION DEVICE AND SEMICONDUCTOR INSPECTION PROCEDURE |
| KR1020217028489A KR20220016022A (ko) | 2019-06-03 | 2020-04-16 | 반도체 검사 장치 및 반도체 검사 방법 |
| US17/608,833 US12094138B2 (en) | 2019-06-03 | 2020-04-16 | Semiconductor inspection device and semiconductor inspection method |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2019103830 | 2019-06-03 | ||
| JP2019-103830 | 2019-06-03 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020246149A1 true WO2020246149A1 (ja) | 2020-12-10 |
Family
ID=73652778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2020/016739 Ceased WO2020246149A1 (ja) | 2019-06-03 | 2020-04-16 | 半導体検査装置及び半導体検査方法 |
Country Status (8)
| Country | Link |
|---|---|
| US (1) | US12094138B2 (ja) |
| EP (1) | EP3955208A4 (ja) |
| JP (1) | JP7413375B2 (ja) |
| KR (1) | KR20220016022A (ja) |
| CN (1) | CN113994372A (ja) |
| SG (1) | SG11202113170XA (ja) |
| TW (1) | TWI862591B (ja) |
| WO (1) | WO2020246149A1 (ja) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022122596A (ja) * | 2021-02-10 | 2022-08-23 | 浜松ホトニクス株式会社 | 検査装置及び検査方法 |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102365169B1 (ko) * | 2021-09-17 | 2022-02-18 | 주식회사 애자일소다 | 사용자 학습 환경 기반의 강화학습 장치 및 방법 |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009162718A (ja) * | 2008-01-10 | 2009-07-23 | Olympus Corp | 基板検査装置および検査領域設定方法 |
| WO2015098342A1 (ja) | 2013-12-26 | 2015-07-02 | 浜松ホトニクス株式会社 | 画像処理方法、画像処理装置、画像処理プログラム、及び画像処理プログラムを記憶した記憶媒体 |
| US20170148226A1 (en) * | 2015-11-19 | 2017-05-25 | Kla-Tencor Corporation | Generating simulated images from design information |
| US20170193680A1 (en) * | 2016-01-04 | 2017-07-06 | Kla-Tencor Corporation | Generating high resolution images from low resolution images for semiconductor applications |
| US20170345140A1 (en) * | 2016-05-25 | 2017-11-30 | Kla-Tencor Corporation | Generating simulated images from input images for semiconductor applications |
| US20180293346A1 (en) | 2017-04-07 | 2018-10-11 | Fei Company | Alignment of cad data to images in high resolution optical fault analysis |
| JP2019129169A (ja) * | 2018-01-22 | 2019-08-01 | 株式会社日立ハイテクノロジーズ | 画像評価方法及び画像評価装置 |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9390486B2 (en) | 2010-09-29 | 2016-07-12 | Neeraj Khurana | System and method for automatic orientation of a chip to the CAD layout with sub-optical resolution |
| JP2015098342A (ja) | 2013-11-19 | 2015-05-28 | 株式会社エンプラス | 密閉容器および密閉容器セット |
| JP5996687B2 (ja) * | 2015-02-10 | 2016-09-21 | 浜松ホトニクス株式会社 | 検査装置及び検査方法 |
| JP6820184B2 (ja) * | 2016-10-26 | 2021-01-27 | 浜松ホトニクス株式会社 | 半導体デバイス検査方法及び半導体デバイス検査装置 |
| US10733744B2 (en) | 2017-05-11 | 2020-08-04 | Kla-Tencor Corp. | Learning based approach for aligning images acquired with different modalities |
| CN111051993B (zh) * | 2017-09-08 | 2022-06-28 | Asml荷兰有限公司 | 用于机器学习辅助的光学邻近误差校正的训练方法 |
-
2020
- 2020-04-16 WO PCT/JP2020/016739 patent/WO2020246149A1/ja not_active Ceased
- 2020-04-16 KR KR1020217028489A patent/KR20220016022A/ko not_active Ceased
- 2020-04-16 JP JP2021524694A patent/JP7413375B2/ja active Active
- 2020-04-16 SG SG11202113170XA patent/SG11202113170XA/en unknown
- 2020-04-16 CN CN202080041571.3A patent/CN113994372A/zh active Pending
- 2020-04-16 US US17/608,833 patent/US12094138B2/en active Active
- 2020-04-16 EP EP20818448.1A patent/EP3955208A4/en active Pending
- 2020-05-15 TW TW109116148A patent/TWI862591B/zh active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009162718A (ja) * | 2008-01-10 | 2009-07-23 | Olympus Corp | 基板検査装置および検査領域設定方法 |
| WO2015098342A1 (ja) | 2013-12-26 | 2015-07-02 | 浜松ホトニクス株式会社 | 画像処理方法、画像処理装置、画像処理プログラム、及び画像処理プログラムを記憶した記憶媒体 |
| US20170148226A1 (en) * | 2015-11-19 | 2017-05-25 | Kla-Tencor Corporation | Generating simulated images from design information |
| US20170193680A1 (en) * | 2016-01-04 | 2017-07-06 | Kla-Tencor Corporation | Generating high resolution images from low resolution images for semiconductor applications |
| US20170345140A1 (en) * | 2016-05-25 | 2017-11-30 | Kla-Tencor Corporation | Generating simulated images from input images for semiconductor applications |
| US20180293346A1 (en) | 2017-04-07 | 2018-10-11 | Fei Company | Alignment of cad data to images in high resolution optical fault analysis |
| JP2019129169A (ja) * | 2018-01-22 | 2019-08-01 | 株式会社日立ハイテクノロジーズ | 画像評価方法及び画像評価装置 |
Non-Patent Citations (1)
| Title |
|---|
| See also references of EP3955208A4 |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2022122596A (ja) * | 2021-02-10 | 2022-08-23 | 浜松ホトニクス株式会社 | 検査装置及び検査方法 |
| JP7629742B2 (ja) | 2021-02-10 | 2025-02-14 | 浜松ホトニクス株式会社 | 検査装置及び検査方法 |
Also Published As
| Publication number | Publication date |
|---|---|
| TWI862591B (zh) | 2024-11-21 |
| SG11202113170XA (en) | 2021-12-30 |
| JPWO2020246149A1 (ja) | 2020-12-10 |
| EP3955208A4 (en) | 2023-05-24 |
| EP3955208A1 (en) | 2022-02-16 |
| TW202114007A (zh) | 2021-04-01 |
| JP7413375B2 (ja) | 2024-01-15 |
| US12094138B2 (en) | 2024-09-17 |
| KR20220016022A (ko) | 2022-02-08 |
| CN113994372A (zh) | 2022-01-28 |
| US20220301197A1 (en) | 2022-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI706485B (zh) | 光學晶粒對資料庫檢查 | |
| JP6707616B2 (ja) | 画像処理方法、画像処理装置、画像処理プログラム、及び画像処理プログラムを記憶した記憶媒体 | |
| JP7413376B2 (ja) | 半導体検査方法及び半導体検査装置 | |
| TW201928541A (zh) | 用於大型偏移晶粒對晶粒檢測之多步驟影像對準方法 | |
| KR102768819B1 (ko) | 반도체 검사의 다중 모드 결함 분류 | |
| KR20190004000A (ko) | 반도체 애플리케이션을 위한 신경망 및 순방향 물리적 모델을 포함하는 방법 및 시스템 | |
| CN115516295A (zh) | 使用深度学习方法的缺陷尺寸测量 | |
| US10591539B2 (en) | Automated scan chain diagnostics using emission | |
| US10983162B2 (en) | Semiconductor device inspection method and semiconductor device inspection device | |
| US12094138B2 (en) | Semiconductor inspection device and semiconductor inspection method | |
| JP6232476B2 (ja) | サブ光学解像度によるcadレイアウトへのチップの自動配向システムおよび方法 | |
| JP6961846B1 (ja) | Euvマスク検査装置、euvマスク検査方法、euvマスク検査プログラム及びeuvマスク検査システム | |
| US9958502B2 (en) | Defect isolation methods and systems | |
| US12340504B2 (en) | Semiconductor inspecting method and semiconductor inspecting device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20818448 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 2021524694 Country of ref document: JP Kind code of ref document: A |
|
| ENP | Entry into the national phase |
Ref document number: 20217028489 Country of ref document: KR Kind code of ref document: A |
|
| ENP | Entry into the national phase |
Ref document number: 2020818448 Country of ref document: EP Effective date: 20211111 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |