KR102561093B1 - Apparatus and Method of Filter Extraction for Defect Detection of Semiconductor Apparatus, and Defect Detection System Using the Same - Google Patents

Abstract

An apparatus for extracting a filter according to an embodiment of the present technology generates a difference image based on image data provided from a filter library including a plurality of filters for classifying defects of a semiconductor device and an external device, and a plurality of difference images are generated in the difference image. It may be configured to include an image processing unit configured to generate a filtered image for each filter by applying filters, respectively, and to calculate a signal-to-noise ratio for the filtered image.

Description

Apparatus and Method of Filter Extraction for Defect Detection of Semiconductor Apparatus, and Defect Detection System Using the Same}

The present invention relates to a defect detection technology of a semiconductor pattern, and more particularly, to a filter extraction device and method for detecting defects in a semiconductor device, and a defect detection system using the same.

Defect inspection of a pattern fabricated on a semiconductor substrate is an important technique for improving the production yield of semiconductor devices.

Image inspection techniques using light or electron beams, or both, can be used to inspect patterns for defects. In these techniques, an image acquired using a light or electron beam is compared to itself or to a reference image to inspect defects.

As the design rules of semiconductor devices decrease, patterns formed on semiconductor substrates have increasingly fine widths. Accordingly, the optical resolution of the pattern inspection device inevitably reaches its limit.

Therefore, there is an urgent need for a technique capable of coping with the performance degradation problem for defect detection.

Embodiments of the present technology may provide a filter extraction device and method for detecting defects in a semiconductor device capable of providing an optimal filter for an image-based pattern inspection device, and a defect detection system using the same.

An apparatus for extracting a filter according to an embodiment of the present technology includes a filter library including a plurality of filters for classifying defects of a semiconductor device; and generating a difference image based on image data provided from an external device, applying a plurality of filters to the difference image, generating a filtered image for each filter, and calculating a signal-to-noise ratio for the filtered image. Processing unit; may be configured to include.

A filter extraction method according to an embodiment of the present technology is a filter extraction method of a filter extraction device including a filter library including a plurality of filters for classifying defects of a semiconductor device, wherein the filter extraction device is provided from an external device. generating a difference image based on the image data, and generating a filtered image by applying one of candidate filters selected from the filter library to the difference image; calculating a signal-to-noise ratio for the filtered image; and repeating the steps of generating the filtered image and calculating the signal-to-noise ratio for all of the candidate filters.

A defect detection system according to an embodiment of the present technology includes an image-based measurement device that inspects a semiconductor device and outputs image data; and a filter library including a plurality of filters for classifying defects of the semiconductor device, generating a difference image based on the image data provided from the measurement device, and applying a plurality of filters to the difference image, respectively. A filter extraction device configured to generate a filtered image for each filter, calculate a signal-to-noise ratio for the filtered image, select one filter based on the signal-to-noise ratio, and transplant it into the measuring device; there is.

According to the present technology, it is possible to provide an optimal filter capable of detecting defects through an image-based pattern inspection device. Therefore, it is possible to improve discrimination between defects and noise by overcoming the resolution limit of optical inspection equipment due to the decrease in pattern width.

In addition, as the filter selected as the optimal filter is applied to the pattern inspection device, it is not necessary to perform the filter selection process of the pattern inspection device itself, so the time required for pattern inspection can be reduced.

Furthermore, if the shape of the optimal filter selected for each defect in the pattern is analyzed, a defect generation tendency can be identified, and defect factors can be minimized during subsequent manufacturing of the semiconductor device.

1 is a block diagram of a filter extraction device according to an embodiment.
2 is a configuration diagram of a filtering unit according to an embodiment.
3 is a configuration diagram of an SNR measurement unit according to an embodiment.
4 is a configuration diagram of a selection unit according to an embodiment.
5 is a flowchart illustrating a filter extraction method according to an exemplary embodiment.
6 is a flowchart illustrating a filtering method according to an exemplary embodiment.
7 is a flowchart for explaining an SNR measurement method according to an embodiment.
8 are exemplary diagrams of a filter according to an embodiment.
9 is a diagram for explaining a filtering process and an SNR measurement process according to an embodiment.
10 is a diagram for explaining a user interface of a filter extraction device according to an exemplary embodiment.
11 is a configuration diagram of a defect detection system according to an embodiment.

Hereinafter, embodiments of the present technology will be described in more detail with reference to the accompanying drawings.

1 is a block diagram of a filter extraction device according to an embodiment.

Referring to FIG. 1 , the filter extraction device 10 may include a controller 110, a storage unit 120, a user interface 130, an image processing unit 100, and a selection unit 160.

The controller 110 may be configured to control the overall operation of the filter extraction device 10 .

The storage unit 120 may include a main memory device and an auxiliary memory device, and may store programs, data, application programs, operation parameters, processing results, etc. necessary for the filter extraction device 10 to operate.

The storage unit 120 may include a filter library 122 . The filter library 122 may include a plurality of 2D image filters prepared to distinguish various types of defects. As will be described later, the filter library 122 may include various types of filters, for example, as shown in FIG. 8 .

The user interface 130 provides an environment in which an operator can access the filter extraction device 10, and may include an input device interface and an output device interface. The input device may include, for example, at least one of a keyboard, mouse, microphone, touch pad, and touch screen. The output device may include, for example, at least one of a display and a speaker.

In one embodiment, an operator may control operating parameters for filter extraction via the user interface 130 . Operational parameters may include pixel depth, mask size, filters to be applied, and the like. In one embodiment, the pixel depth parameter may be selected from among 8 bits, 16 bits, and 32 bits, but is not limited thereto, and it is also possible to further extend the pixel depth to improve accuracy. A mask size parameter can be selected when enlarging the size of a mask image. The filter parameter to be applied provides an environment in which an operator can select the type and/or number of filters to be applied to a difference image.

The user interface 130 may also provide image data input to the filter extraction device 10, a processed result image, and the like through a display device.

The image processing unit 100 generates a difference image based on image data provided from the outside, generates a filtered image by applying a plurality of filters to the generated difference image, respectively, and generates a signal-to-noise ratio (SNR) for each filtered image. Noise Ratio (SNR).

In one embodiment, the image processing unit 100 may include a filtering unit 140 and an SNR measurement unit 150.

The filtering unit 140 is configured to receive image data provided from the outside. The image data may include at least one Defect Of Interest (DOI) set and a NUisance Image (NUI) set. In one embodiment, the DOI set includes a reference image used in a semiconductor pattern inspection device, for example, an image-based measurement device that can be linked with the filter extraction device 10 of the present embodiment, and at least one defect generated according to a pattern inspection result. It may include an image and a mask image generated based on the defect area included in the defect image. The NUI set may include at least one noise image generated according to a pattern inspection result in the measurement device.

As image data including DOI and NUI is provided, the filtering unit 140 may generate a difference image between the defect image and the reference image and extract a defect location based on the mask image. In addition, the filtering unit 140 may generate a plurality of filtered images by applying a plurality of filters selected from filters stored in the filter library 122 to difference images, respectively. A filter to be applied to the difference image may be selected by the user or may be automatically selected by the filtering unit 140 . In this case, the filtering unit 140 may generate a filtered image by selecting all or a plurality of filters included in the filter library 122 .

The SNR measurement unit 150 may distinguish a defect area and a noise area from the filtered image based on the defect location extracted by the filtering unit 140 and calculate a signal to noise ratio (SNR). The SNR measurement unit 150 may be configured to calculate an SNR for each filtered image generated by the filtering unit 140 .

The selection unit 160 may be configured to select an optimal filter based on the SNR calculation result of the SNR measurement unit 150. In an embodiment, the selection unit 160 may determine a filter applied to a filtered image having a maximum SNR as an optimal filter as a result of calculation of the SNR measuring unit 150 .

Once the optimal filter is determined, it can be implanted into the instrumentation device. When inspecting a pattern of a semiconductor device, the measurement device can provide inspection results with improved discriminative power between defects and noise for a pattern to be inspected, by using a pre-implanted filter.

2 is a configuration diagram of a filtering unit according to an embodiment.

Referring to FIG. 2 , the filtering unit 140 may include an image data receiving unit 141 , a difference image generating unit 143 , a filtered image generating unit 145 , and a defect location extracting unit 147 .

The image data receiving unit 141 may receive, for example, at least one DOI set and NUI set from a measurement device. As described above, the DOI set may include a reference image, at least one defect image, and a mask image. The NUI set may include at least one noise image generated according to a pattern inspection result in the measurement device.

The difference image generator 143 may be configured to generate a difference image between the defect image received from the image data receiver 141 and the reference image.

The filtered image generation unit 145 may be configured to generate a plurality of filtered images by applying a plurality of filters selected from filters stored in the filter library 122 to difference images, respectively. A filter usable by the difference image generator 145 may be selected by a user or may be automatically selected by the filtering image generator 145 . At this time, the filtered image generation unit 140 may select all or a plurality of filters included in the filter library 122 and apply each of the selected filters to the difference image to generate each filtered image. In addition, each filtered image generated by the filtered image generator 145 may be provided to the SNR measurer 150 .

The defect location extractor 147 may be configured to extract a defect location based on the mask image received from the image data receiver 141 . The defect location of the mask image may be provided to the SNR measuring unit 150 .

3 is a configuration diagram of an SNR measurement unit according to an embodiment.

Referring to FIG. 3 , the SNR measuring unit 150 may include a defect location identifying unit 151 and an SNR calculating unit 153 .

The defect location identification unit 151 receives the filtered image and the defect location extraction result from the filtering unit 140 and identifies the defect location from the filtered image. Accordingly, an area other than the identified defect location is identified as a noise location.

The SNR calculation unit 153 calculates the SNR based on the defect location and the noise location identified by the defect location identification unit 151 . In one embodiment, the SNR may be calculated as a ratio of the maximum value of the defective region to the maximum value of the noise region as in the following [Equation].

[mathematical expression]

4 is a configuration diagram of a selection unit according to an embodiment.

Referring to FIG. 4 , the selection unit 160 may include a comparison unit 161 and a determination unit 163 .

The comparison unit 161 may be configured to compare the SNR calculation result for each filter calculated by the SNR measurement unit 150 and determine the order of the filters according to the comparison result. The ranking of filters may be determined by, for example, sorting the SNR calculation results in descending or ascending order.

The decision unit 163 may be configured to determine a filter having a maximum SNR as an optimal filter for a corresponding defect as a result of the comparison by the comparison unit 161 . The optimal filter selected by the decision unit 163 can be implanted into the measurement device.

As such, in the present technology, a filtered image is generated by applying a plurality of filters to the difference image between the defect image and the reference image, respectively. Then, the SNR is calculated by dividing the defect location and the noise location in each filtered image, and a filter capable of maximally distinguishing the defect image (DOI) and the noise image (NUI) of interest is extracted as an optimal filter using the SNR. The filter extracted in this way is implanted into a side device and can be applied when pattern inspection of a semiconductor device is performed thereafter.

That is, the measuring device applies the filter implanted from the filter extracting device of the present technology to a comparison image generated as a result of comparing the reference image and the image obtained from the semiconductor device. Accordingly, since defects and noise can be distinguished by applying an optimal filter to the comparison image, the accuracy of defect inspection can be improved.

In one embodiment, the function of the filter extraction device 10 may be produced in the form of software or applications through computer programming and executed through a computing device.

5 is a flowchart illustrating a filter extraction method according to an exemplary embodiment.

The filter extraction device 10 may receive image data from an interlocking measuring device (S101). Image data may include at least one DOI set and NUI set. In one embodiment, the DOI set may include a reference image used by the measurement device, at least one defect image generated according to a pattern inspection result of the measurement device, and a mask image generated based on a defect region included in the defect image. there is. The NUI set may include at least one noise image generated according to a pattern inspection result in the measurement device.

As the image data is received, the filter extraction device 10 may generate a plurality of filtered images by applying a plurality of filters to the difference image generated based on the image data (S103). A plurality of filters to be applied to the difference image may be selected in advance as candidate filters by an operator or the filter extraction device 10 . A plurality of candidate filters may be selected from the filter library 122, and as the number of candidate filters increases, an optimum filter may be more accurately selected.

The process of generating the filtered image (S103) will be described in detail as follows.

Referring to FIG. 6 , the filter extraction device 10 may generate a difference image between the defect image received in step S101 and the reference image (S201).

In addition, the filter extraction device 10 may extract a defect location based on the mask image received in step S101 (S203).

The filter extracting device 10 may generate a filtered image by selecting one of the candidate filters (S205) and applying the filter selected in step S205 to the difference image generated in step S201 (S207).

The filtered image generated in the filtering step (S103) is provided to a subsequent step, S105, so that the filter extraction device 10 may be configured to calculate a signal to noise ratio (SNR) for the filtered image (S105). .

The SNR calculation method for the filtered image will be described in detail with reference to FIG. 7 .

The filter extraction device 10 receives the filtered image generated in step S103 and the defect location extraction result, and identifies a defect location from the filtered image (S301). Accordingly, an area other than the identified defect location is identified as a noise location.

Then, the filter extraction device 10 calculates the SNR for the filtered image based on the defect location and the noise location identified in step S301 (S303). In one embodiment, the SNR may be calculated as a ratio of the maximum value of the defective region to the maximum value of the noise region as in the above [Equation].

When the SNR of the filtered image is calculated, it may be determined whether the SNR was measured by applying all candidate filters to the filtered image (S107).

If all candidate filters are not applied, the process returns to the filtering step (S103) and substantially to the filter selection step (S205) to generate filtered images for other candidate filters and measure SNR. That is, the process of generating the filtered image and measuring the SNR may be repeatedly performed for all candidate filters.

If the filtering image generation and SNR measurement processes have been performed for all candidate filters, the filter extraction device 10 may compare SNRs calculated for each filter (S109). Then, an optimal filter may be selected based on the comparison result (S111). In one embodiment, an optimal filter may be selected as a filter capable of maximally discriminating between a defect image of interest (DOI) and a noise image (NUI) using SNR.

The selected optimal filter can now be implanted into the metrology device and applied to discriminate between defects and noise.

8 are exemplary diagrams of a filter according to an embodiment.

A plurality of filters that may be stored in the filter library 122 may have various resolutions, and the depth of each pixel may also be determined in various ways.

In one embodiment, the resolution of the filter may be configured to have a resolution such as 3x3, 5x5, 7x7, and the like. In addition, each pixel may have a full marking (Full), vertical marking (Vertical), horizontal marking (Horizontal), diagonal marking (Diagonal), cross marking (Cross) form.

8 shows a plurality of filters having a resolution of 5x5. In addition, filters having full marking (a), vertical marking (b), horizontal marking (c), diagonal marking (d, e), and cross marking (f, g) according to pixel depth are shown.

Of course, the resolution and pixel depth of the filter can be configured in various sizes and shapes.

9 is a diagram for explaining a filtering process and an SNR measurement process according to an embodiment.

The resolution of each image constituting the image data (defect image, reference image, mask image, noise image) may be selected from 32x32, 64x64, and 128x128, but may have a higher resolution.

In addition, the pixel depth of each image constituting the image data may be selected from among 8 bits, 16 bits, and 32 bits, but is not limited thereto, and it is possible to further expand the pixel depth to improve accuracy.

Referring to FIG. 9 , a difference image between a defective image (Defective) and a reference image (Reference) may be formed by the filtering unit 140 . Also, the filtering unit 140 may generate a filtered image by applying one of the candidate filters to the difference image.

Meanwhile, the filtering unit 140 may extract a defect location from the mask image (Mask).

Accordingly, the SNR measurement unit 150 may identify a defect region and a noise region from the filtered image according to the extracted defect location, and calculate the SNR according to the analogy.

When the filtering and SNR measurement processes are repeated for each of a plurality of candidate filters, the result shown in FIG. 10 can be obtained.

10 is a diagram for explaining a user interface of a filter extraction device according to an exemplary embodiment.

Simulation target image data may be displayed on the user interface, in particular, region A of the display device. As shown in FIG. 9 , simulation target image data may include a defect image (Defect), a reference image (Reference), a mask image (Mask), and a difference image (Difference).

In area B, SNRs of at least one defect image and at least one noise image to be simulated may be displayed before applying the filter.

Area C may display an SNR of a filtered image generated by filtering at least one defect image and at least one noise image by a selected filter.

Area D may display data information being simulated.

Areas E, F, and G may be areas for setting operating parameters or receiving control commands.

As described above, according to the present technology, an optimal filter for each defect of a semiconductor device can be automatically extracted using image data provided from a measurement device and filters constructed from a filter library. In addition, a filter extraction process may be provided to an operator through a user interface, and the operator may change operating parameters in various ways.

11 is a configuration diagram of a defect detection system according to an embodiment.

The defect detection system 20 according to an embodiment may include a measurement device 210 and a filter extraction device 220 .

The metrology device 210 is an image-based metrology device, and may be an optical metrology device, an electron beam metrology device, or a metrology device using both.

The measurement device 210 may generate image data from an image obtained of a semiconductor substrate to be inspected. The image data may include a reference image, at least one defect image generated according to a pattern inspection result, and a mask image generated based on a defect area included in the defect image.

The filter extraction device 220 receives image data from the measurement device 210 . The filter extraction device 220 may include a filter library and extract an optimal filter based on image data and the filter library. The final filter extracted from the filter extraction device 220 may be implanted into the measurement device 210 .

The measurement device 210 can provide inspection results with improved discriminative power between defects and noise for a pattern to be inspected, by using a pre-implanted filter when inspecting a pattern of a semiconductor device.

The filter extraction device 220 may be the filter extraction device 10 shown in FIGS. 1 to 4 . Therefore, the filter extraction device 220 generates a filtered image by applying a plurality of filters to the difference image between the defect image and the reference image, respectively, calculates the SNR by dividing the defect location and the noise location in each filtered image, and calculates the SNR. A filter capable of maximally discriminating between a defect image of interest (DOI) and a noise image (NUI) can be extracted as an optimal filter.

Accordingly, the measurement device 210 applies the filter transplanted from the filter extraction device of the present technology to a comparison image generated as a result of comparing the reference image and the image obtained from the semiconductor device. Accordingly, since defects and noise can be distinguished by applying an optimal filter to the comparison image, the accuracy of defect inspection can be improved.

As such, those skilled in the art to which the present invention pertains will be able to understand that the present invention may be embodied in other specific forms without changing its technical spirit or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. The scope of the present invention is indicated by the claims to be described later rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

10: filter extraction device
100: image processing unit
110: controller
120: storage unit
130: user interface
140: filtering unit
150: SNR measuring unit
160: selection unit

Claims (17)

a filter library including a plurality of filters for classifying defects of a semiconductor device;
Receive image data including a reference image used by an external device, at least one defect image generated according to a pattern inspection result of the external device, and a mask image generated based on a defect area included in the defect image generates a filtered image for each filter by applying the plurality of filters to the difference image, extracts a defect location from the mask image, and for each of the filtered images, a defect area based on the defect location an image processor configured to classify the noise region and the noise region and calculate a signal-to-noise ratio, which is a ratio of the defective region to the noise region;
A filter extraction device configured to include a. ◈Claim 2 was abandoned when the registration fee was paid.◈ According to claim 1,
The external device is an image-based pattern inspection device,
The image data is configured to further include at least one noise image generated according to a pattern inspection result of the external device. ◈Claim 3 was abandoned when the registration fee was paid.◈ According to claim 1,
The image processing unit generates the difference image from the defect image and the reference image, and a filtering unit configured to apply a plurality of filters selected from the filter library to the difference image, respectively, to generate a filtered image for each of the plurality of filters. A filter extraction device configured to include. delete ◈Claim 5 was abandoned when the registration fee was paid.◈ According to claim 1,
and a selection unit configured to select one filter based on a signal-to-noise ratio calculation result of the image processing unit. ◈Claim 6 was abandoned when the registration fee was paid.◈ According to claim 5,
The filter extraction device configured to implant any one of the selected filters into the external device. A filter extraction method of a filter extraction device including a filter library including a plurality of filters for classifying defects of a semiconductor device,
An image comprising a reference image used by the external device, at least one defect image generated according to a pattern inspection result of the external device, and a mask image generated based on a defect area included in the defect image. generating a difference image by receiving data;
generating, by the filter extraction device, a filtered image by applying one of candidate filters selected from the filter library to the difference image;
extracting, by the filter extraction device, a defect position from the mask image;
calculating, by the filter extraction device, a signal-to-noise ratio, which is a ratio of the defective area to the noise area, by dividing the filtered image into a defective area and a noise area based on the defect location; and
repeating the steps of generating the filtered image and calculating the signal-to-noise ratio for all of the candidate filters;
Filter extraction method configured to include. ◈Claim 8 was abandoned when the registration fee was paid.◈ According to claim 7,
The external device is an image-based pattern inspection device,
The image data is configured to further include at least one noise image generated according to a pattern inspection result of the external device. ◈Claim 9 was abandoned when the registration fee was paid.◈ According to claim 8,
The difference image,
A method of extracting a filter based on the defect image and the reference image. delete ◈Claim 11 was abandoned when the registration fee was paid.◈ According to claim 7,
Filter extraction method configured to further include the step of selecting any one filter based on the signal-to-noise ratio calculation result. ◈Claim 12 was abandoned when the registration fee was paid.◈ According to claim 11,
The filter extraction method configured to further include the step of implanting the selected one filter to the external device. an image-based measurement device that inspects a semiconductor device and outputs image data including a reference image, at least one defect image generated according to a result of the inspection, and a mask image generated based on a defect region included in the defect image; and
A difference image is generated based on a filter library including a plurality of filters for classifying defects of the semiconductor device and the image data provided from the measurement device, and each of the plurality of filters is applied to the difference image, respectively. A filtered image is generated for each filter, a defect location is extracted from the mask image, and for each of the filtered images, a defect area and a noise area are divided based on the defect location, and the ratio of the defect area to the noise area is a filter extraction device configured to calculate a signal-to-noise ratio, select a filter based on the signal-to-noise ratio, and transplant the filter to the measuring device;
Fault detection system configured to include. ◈Claim 14 was abandoned when the registration fee was paid.◈ According to claim 13,
wherein the metrology device is selected from an optical metrology device, an electron beam metrology device, or a light and electron beam metrology device. ◈Claim 15 was abandoned when the registration fee was paid.◈ According to claim 13,
The image data is configured to further include at least one noise image generated according to a pattern inspection result of the measurement device. ◈Claim 16 was abandoned when the registration fee was paid.◈ According to claim 15,
wherein the filter extraction device is configured to generate the difference image from the defect image and the reference image.
delete

Images