WO2019163136A1

WO2019163136A1 - X-ray inspection method

Info

Publication number: WO2019163136A1
Application number: PCT/JP2018/007019
Authority: WO
Inventors: 昌幸小林; 竜己服部; 渋谷　久恵; 宮本　敦; 一亥水野; 秀明笹澤
Original assignee: 株式会社日立ハイテクノロジーズ
Priority date: 2018-02-26
Filing date: 2018-02-26
Publication date: 2019-08-29

Abstract

The objective of the present invention is to provide an X-ray inspection method for determining an appropriate imaging condition or image processing parameter for a sample having unknown correct defect information. To achieve this objective, this X-ray inspection method for using an X-ray inspection device to obtain an image of a sample and obtain a defect inspection result for the sample by subjecting the image to image processing is provided with: a step for obtaining an image for reference defect data acquisition under reference defect data acquisition conditions, a step for obtaining a plurality of reference defect data candidates using a plurality of different image processing parameter candidates, a step for selecting an image processing parameter from the plurality of image processing parameter candidates on the basis of the plurality of reference defect data candidates, and a step for obtaining reference defect data from the reference defect data candidate corresponding to the selected image processing parameter.

Description

X-ray inspection method

The present invention relates to an X-ray inspection method.

With the progress of high integration of semiconductors, the evolution of 3D stacking technology has been remarkable in recent years, and the development of semiconductor stacking technology such as TSV (Through Si Via) and micro bumps is progressing. Along with this trend, there is a growing need for non-destructive inspection of defects occurring in electrical connection parts such as TSVs and microbumps. For non-destructive inspection, a technique of irradiating a semiconductor wafer with X-rays and analyzing from the transmission image has been widely studied.

Japanese Unexamined Patent Application Publication No. 2016-118445 (Patent Document 1) is a technique for inspecting an internal defect of a semiconductor device using X-rays. In Patent Document 1, X-rays emitted from an X-ray source are irradiated onto a sample to be inspected on which a structure mounted on a rotary stage is formed, and X-rays transmitted through the sample irradiated with X-rays are disclosed. Is detected by an X-ray detector, and an X-ray transmission image is formed by processing an X-ray transmitted through the sample detected by the X-ray detector by an image processing unit to form an X-ray transmission image. An X-ray inspection method is disclosed in which an X-ray transmission image is processed by a defect determination unit to detect a defect in a sample.

Japanese Unexamined Patent Publication No. 2016-118445

In a general appearance inspection, an imaging condition of a sample and an image processing parameter for determining a defect are set on an inspection apparatus before the inspection. Specifically, a defect to be detected is given to an inspection apparatus or a user as a correct defect, and imaging conditions and image processing parameters are set depending on whether or not the correct defect can be detected. However, in the case of X-ray inspection, since defects may exist inside the sample, it is difficult to obtain accurate defect information.

Patent Document 1 describes an X-ray inspection method for a semiconductor device, but does not disclose a specific method for setting imaging conditions such as an X-ray irradiation time and setting image processing parameters such as a threshold. In such an X-ray inspection method, for example, the user visually determines the image quality while changing the irradiation time condition, and obtains an appropriate irradiation time condition, or determines whether the detection result is good or bad while changing the threshold value. It is assumed that a process is generated such that an appropriate threshold value is obtained by visual inspection.

When the above process is executed in a state where correct defect information is not given, the image quality and the detection criteria for the image are different for each user, so the imaging conditions and image processing parameters depend on the user. As a result, there arises a problem that the performance of the inspection apparatus becomes unstable.

Therefore, an object of the present invention is to provide an X-ray inspection method for determining appropriate imaging conditions and image processing parameters for a sample whose correct defect information is unknown.

In view of the above-described background art and problems, the present invention is an X-ray inspection method for obtaining an image of a sample using an X-ray inspection apparatus and performing image processing to obtain a defect detection result of the sample. A step of obtaining an image for obtaining reference defect data under reference defect data obtaining conditions, a step of obtaining a plurality of reference defect data candidates using a plurality of different image processing parameter candidates, and a plurality based on the plurality of reference defect data candidates Selecting an image processing parameter from the image processing parameter candidates, and obtaining reference defect data from the reference defect data candidates corresponding to the selected image processing parameter.

According to the present invention, it is possible to provide an X-ray inspection method for determining appropriate imaging conditions and image processing parameters for a sample whose correct defect information is unknown.

It is a schematic diagram which shows the structure of the X-ray inspection apparatus in Example 1. FIG. 1 is a plan view of a wafer in Example 1. FIG. 1 is a sectional view of a wafer in Example 1. FIG. It is explanatory drawing of the X-ray transmissive image obtained when X-rays are irradiated from the direction inclined with respect to the wafer in Example 1. It is a figure explaining the flow of the image-analysis process which performs defect detection and size measurement from the X-ray transmission image in Example 1. FIG. It is an example of the X-ray transmission image obtained by irradiating X-rays from the diagonal to the wafer in which one layer TSV in Example 1 was formed. 6 is a flowchart illustrating a processing flow for acquiring reference defect data according to the first exemplary embodiment. 6 is a graph illustrating a method for selecting an image processing parameter from a plurality of image processing parameter candidates based on a plurality of reference defect data candidates in the first embodiment. 12 is a flowchart illustrating processing for selecting an imaging condition from a plurality of imaging condition candidates in the second embodiment. 10 is a graph illustrating a method for selecting an imaging condition from a plurality of imaging condition candidates in the second embodiment. 12 is a flowchart illustrating processing for selecting an image processing parameter from a plurality of image processing parameter candidates and processing for selecting an imaging condition from a plurality of imaging condition candidates in the third embodiment. It is a schematic diagram which shows the method of calculating the coincidence number or the mismatch number of the defect detection position of the reference defect data and the detection result candidate in the third embodiment. 10 is a graph showing a method of selecting an image processing parameter based on the number of coincidences or the number of mismatches between reference defect data and defect detection positions of a plurality of detection result candidates in Example 3. 10 is a graph showing a method for selecting an imaging condition based on the number of coincidence or the number of mismatches in the detection positions of detection results obtained with a plurality of imaging condition candidates in Example 3 and reference defect data. FIG. 10 is an example of a GUI screen for setting imaging conditions in Embodiment 4. FIG. FIG. 10 is an example of a GUI screen for creating a recipe in Embodiment 4. FIG.

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

First, a method for obtaining a sample defect detection result by acquiring an image of a sample using an X-ray inspection apparatus, which is a premise of the present embodiment, and performing image processing will be described.

FIG. 1 is a schematic diagram showing the configuration of an X-ray inspection apparatus. In FIG. 1, an X-ray source 1 irradiates a wafer 2 which is a sample to be inspected with X-rays. As the X-ray source 1, for example, a so-called microfocus X-ray source that has a small focal size and is advantageous in terms of high-resolution imaging can be used. The translation stage 3 holds the wafer 2 and moves the wafer 2 in the X-axis, Y-axis, and Z-axis directions. The rotary stage 4 holds the translation stage 3 and rotates the wafer 2 by rotational movement in the XY plane. The central parts of the translation stage 3 and the rotary stage 4 can be made of glass (not shown) that absorbs little X-rays. The X-ray detector 5 is disposed at a position facing the X-ray source 1 with the translation stage 3 and the rotation stage 4 interposed therebetween. Here, an image intensifier, a CCD camera, or the like can be used for the X-ray detector 5. X-rays emitted from the X-ray source 1 are absorbed by the wafer 2 disposed on the translation stage 3, and the transmitted X-rays are detected by the X-ray detector 5. By changing the position of the wafer 2 by the translation stage 3, the magnification and the width of the field of view are changed. The X-ray detector 5 can rotate in the XZ plane around the X-ray generation position of the X-ray source 1, and the translation stage 3 translates the wafer 2 in accordance with the rotation angle, so that the measurement region does not shift. Adjust as follows. The X-ray source 1, the translation stage 3, the rotary stage 4, and the X-ray detector 5 are arranged inside the X-ray shielding wall 6 so that X-rays do not leak outside.

The X-ray source controller 101 controls the tube voltage of the X-ray source 1, the probe current, the applied magnetic field to the electron optical system, the applied voltage, and the ON / OFF of X-ray generation. The stage controller 102 is the translation stage 3 and the rotary stage 4. The X-ray detector controller 103 reads data from the X-ray detector 5 and sets sensitivity, exposure time, average number of sheets, and the like. The X-ray source controller 101, the stage controller 102, and the X-ray detector controller 103 are controlled by the control unit 104. An X-ray transmission image is picked up while moving the wafer 2 based on the inspection conditions input in advance to the control unit 104 through the GUI.

The image analysis unit 105 receives an X-ray transmission image and a pre-input image processing parameter from the control unit 104, discriminates defects such as voids by image analysis, and measures the size and position of an inspection target such as TSV. The result is displayed on the input / output unit 106. When defects such as vacancies exist in the TSV or microbump, the luminance value of the defective portion is higher than that of the peripheral portion in the X-ray transmission image. By performing threshold processing on the image, a defective area can be extracted. Specifically, a region having a luminance value higher than the threshold value can be a defective portion, and a region having a lower luminance value can be a non-defective portion.

2 and 3 show an example of a schematic diagram of the wafer 2. 2 is a plan view, and FIG. 3 is a cross-sectional view taken along the line A-A 'in FIG. A plurality of dies 10 are regularly formed on the wafer 2, and TSVs 11 are formed on a part of the dies 10. The diameter of TSV11 is Φ, and is formed at a pitch of Px in the X-axis direction and Py in the Y-axis direction. In FIG. 3, the first layer 13, the second layer 14, and the third layer 15 are laminated, and the respective layers are connected by the TSV 11 and the microbump 12. The length of TSV11 is h.

Referring to FIG. 4, an explanation will be given when an X-ray transmission image is acquired by tilting in the ψ direction. Here, description will be made with reference to an XZ sectional view in which only one TSV 11 formed on the second layer 14 of the wafer 2 is taken out. TSV is mainly made of Cu. Since Cu has a larger atomic number than Si constituting most of the wafer 2, X-ray absorption is large. That is, when the second layer 14 is irradiated with X-rays 200 and transmitted X-rays are detected by the X-ray detector 5, the region where TSV11 does not exist in the captured image becomes bright because X-ray absorption is small, and TSV11 is The existing region becomes dark because of the large X-ray absorption. Further, if the void 20 is present inside the TSV 11, X-ray absorption is reduced in the void region, and only the void region becomes brighter than the surroundings, and the void can be detected with this brightness difference.

When ψ = 0 degrees, that is, when X-rays 200 are irradiated from the vertical direction of the second layer 14, there are few pixels that detect X-rays passing through the TSV11 region, and the XV TSV11 incident on the corresponding pixels As a result, the TSV existence area becomes very dark like the profile 30. Even if the void 20 exists in the TSV 11, the absorption at the TSV 11 is very large in the first place, so that the contrast is poor and the detection performance is deteriorated.

On the other hand, when X-rays are irradiated while being tilted in the ψ direction (for example, ψ = 60 degrees), the number of pixels that detect X-rays that have passed through the TSV11 region increases and the TSV11 of the X-rays that are incident on the corresponding pixels. Is also smaller than when ψ = 0 °, and the void 20 can be detected with a higher contrast as in the profile 31.

By acquiring an X-ray transmission image by tilting in the ψ direction with respect to the wafer in this way, the brightness contrast of the void portion with respect to the TSV region is increased, and the void detection accuracy inside the TSV can be improved. .

The processing flow in the image analysis unit 105 will be described with reference to FIG. Here, as an example, an X-ray transmission image including a plurality of TSVs 11 is obtained by tilting in the ψ direction with respect to the wafer 2 on which TSVs 11 are formed on only one layer. FIG. 6 shows an example of an X-ray transmission image. In FIG. 6, 60 is an example of an X-ray transmission image, and the right side is an enlarged view thereof. As shown in FIG. 6, when the TSV 11 has a substantially cylindrical shape, the transmission image 65 has a shape obtained by rounding the left and right sides of a horizontally long rectangle. In addition, when the void 20 exists inside the TSV 11, an image 66 brighter than the surroundings is observed.

5, the image analysis unit 105 receives the X-ray image 60 from the control unit 104, performs preprocessing S601, and creates an inspection image 61. The preprocessing is processing for obtaining an appropriate inspection image 61, and includes shading correction, contrast correction, noise removal, and the like. Processing for correcting imaging distortion caused by the X-ray detector may be included.

Next, TSV region extraction S602 is performed using the inspection image 61. Since the transmission image 65 of the TSV is darker than the peripheral portion, binarization is performed with the darker side being 1 and the brighter side being 0. The binarization threshold can be automatically determined by using Otsu's method or the like. Assuming the presence of the void 20, after tracing the contour of one pattern on the binary image, the convex hull is obtained and the inside is filled with one. Labeling identifies each one separately and extracts information such as area, perimeter length, barycentric position, and minimum and maximum X and Y coordinates.

Next, non-defective image creation S603 is performed using the inspection image 61 and the TSV area information. The non-defective image 62 is an image simulating a normal TSV image. In optical inspection, a method of calculating the average and median by superimposing a plurality of identical patterns using the periodicity of the pattern on the wafer is often used. Even if they are regularly arranged, the shape and size change depending on the imaging position, so a non-defective image of the TSV is created from the transmission image 65 of one TSV. For example, an image in which a bright void portion is darkened is created by applying a minimum value filter of a specified size to the inspection image 61 and then applying a maximum value filter of the same size (opening). Alternatively, by dividing the region according to the distance from the outer periphery and center of the TSV, calculating an average or median for each region and replacing the value, a vertically normal and symmetric normal image is created.

Next, defect detection S604 is performed using the inspection image 61 and the non-defective image 62. A difference image obtained by subtracting the non-defective image 62 from the inspection image 61 is created, binarized by threshold processing, and labeling is performed. The labeled region is a defect candidate, but includes false information detected due to a luminance difference due to density variation that cannot be said to be noise or a defect at the time of image acquisition. For this reason, for each defect area, feature quantities such as area, roundness, relative position with TSV, and luminance value are calculated, and only those having a high probability of defect are extracted as defects by comparison with a predetermined value. Alternatively, learning is performed by teaching whether the defect is a defect or a false report on the feature space, and the defect and the false report are identified using a classification criterion obtained by the learning.

Next, the position / size calculation processing S605 of each TSV area extracted in S602 and the defect area extracted in S604 is performed. Specifically, the center of gravity position of the TSV region, the rectangular size surrounding the TSV region, the center of gravity position of the defect region, the size (area, etc.), and the like.

Finally, position / size correction processing S606 is performed using the parameter 63 input in advance, and TSV or defect-corrected position / size information 64 is output. Specifically, the coordinates of the TSV, the relative coordinates of the visual field center reference, the diameter, the height, the depth from the TSV top of the defect, the size (diameter in terms of a perfect circle), the longitudinal size, the lateral direction size, and the like. . These pieces of information are displayed in a list on the file output and GUI screen in the input / output unit 106. The input / output unit 106 can also display an image in which a defect detected in the input image is marked.

The above is an explanation of obtaining a sample defect detection result by acquiring an image of a sample using an X-ray inspection apparatus, which is a premise of the present embodiment, and performing image processing. Next, FIG. 7 shows a method of determining appropriate imaging conditions and image processing parameters for a sample whose correct defect information is unknown in the configuration of the X-ray inspection apparatus described in FIG. This will be described with reference to FIG.

FIG. 7 is a flowchart showing a processing flow for acquiring reference defect data in this embodiment. In FIG. 7, in S101, a reference defect data acquisition image is acquired under preset reference defect data acquisition conditions.

Here, the reference defect data acquisition condition is an imaging condition in which the image quality is better than that of an actual inspection image. For example, a condition in which the X-ray irradiation time condition is longer than that during actual inspection can be set as the reference defect data acquisition condition. This is because in the X-ray transmission image, the longer the irradiation time, the lower the quantum noise and the better the image quality. For obtaining the reference defect data obtaining image, it is preferable to set the irradiation time as long as possible. Note that the irradiation time may be set so that the contrast and the signal / noise ratio (S / N ratio) become a preset value at a specific position of the sample. In addition, even with a method other than such a long exposure, it is possible to obtain a so-called reproducible defect in which the results (number of defects, location, size, etc.) of a plurality of defect detection processes match. For example, high-intensity illumination, detection from a 3D-CT image, detection under conditions with higher resolution than normal inspection, and the like are possible. In this way, ideal reference defect data can be obtained by acquiring a reference defect data acquisition image with good image quality.

In S102, using the processing method described in FIG. 5 for the reference defect data acquisition image, image processing is executed with a plurality of image processing parameter candidates to obtain a plurality of reference defect data candidates. Here, the image processing parameter may be a threshold value of the threshold processing, or another image processing parameter may be applied according to the content of an image processing algorithm for determining the presence or absence of a defect.

In S103, an image processing parameter is selected from a plurality of image processing parameter candidates based on the plurality of reference defect data candidates (details will be described later). When the image processing parameter is determined, the reference defect data candidate corresponding to the image processing parameter becomes the reference defect data. Thus, the determination of the reference defect data is completed (S104).

FIG. 8 is a graph showing a method of selecting an image processing parameter from image processing parameter candidates based on a plurality of reference defect data candidates in the present embodiment. In FIG. 8, a threshold is used as an image processing parameter, and the reference defect data corresponds to the number of defects. In FIG. 8, a plurality of defect numbers (reference defect data candidates) are obtained with a plurality of threshold values (image processing parameter candidates) by the processing flow shown in FIG. Here, the threshold is a threshold for threshold processing in S604 of FIG. In the case of a void, a region having a luminance value higher than the threshold is a defective portion, and a region having a lower luminance value is a non-defective portion. This graph shows that an appropriate threshold value can be selected from the shape of the threshold value dependency 201 of the number of defects. In other words, the image processing parameter is selected based on the comparison between the reference defect data and a plurality of image processing parameter candidates. That is, when the threshold value is made lower than the appropriate threshold value 202, the number of defects rapidly increases, and more defects are detected than the ideal number of defects 203 (overdetection). This overdetection is caused by erroneously discriminating a minute change in luminance value due to quantum noise as a defect. On the other hand, when the threshold value is made higher than the appropriate threshold value 202, the number of defects decreases, and defects are detected with fewer than the ideal number of defects 203 (missing).

As described above, it is possible to estimate an appropriate threshold value 202 by finding a sudden change point of the defect number from the shape of the threshold value dependency 201 of the defect number. In particular, since the threshold value dependency 201 of the defect number has an inflection point, it is possible to set a threshold value independent of the user by calculating the inflection point and selecting a threshold value corresponding to the inflection point. It becomes.

Alternatively, a reference defect size may be set in advance, and the threshold value dependency of the defect number may be calculated according to the set defect size. Thereby, an appropriate threshold value can be set according to the defect size to be inspected.

Also, a threshold value may be set lower or higher than the inflection point depending on the purpose of inspection. A threshold value is set lower than the inflection point when it is desired to suppress an over-detection even if over-detection is included.

In FIG. 8, the threshold value is described as an example of the image processing parameter. However, the image processing parameter may be applied to other image processing parameters (for example, the parameter relating to the preprocessing S601 in FIG. 5, the feature amount of S604, etc.).

As described above, if reference defect data can be acquired by the method described with reference to FIGS. 7 and 8, appropriate imaging conditions and image processing parameters can be determined, and their dependence on users is suppressed, and the inspection apparatus There is an effect that the performance of is stabilized.

This embodiment is an example in which an imaging condition is automatically set by selecting an imaging condition from a plurality of imaging condition candidates in the first embodiment. Since the configuration of the X-ray inspection apparatus and the method for acquiring the reference defect data are the same as those in the first embodiment, the differences will be described with reference to FIGS.

FIG. 9 is a flowchart showing a flow of processing for selecting an imaging condition from a plurality of imaging condition candidates in the present embodiment. Steps S101 to S104 are the same as in the description of FIG. Here, an irradiation time condition will be described as an example of the imaging condition. In S201, an image is acquired with a plurality of irradiation time condition candidates. Here, a plurality of irradiation time conditions shorter than the reference defect data acquisition conditions are set in advance, and an image is acquired under each irradiation time condition. The image may be acquired by repeatedly irradiating the sample with X-rays according to each irradiation time condition, or by irradiating X-rays for a long time only once, storing images of each frame before integration, It may be generated by changing the cumulative number of stored images. In the latter case, since a plurality of images having different irradiation times can be obtained by one X-ray irradiation, the net irradiation time on the sample can be reduced.

In S202, the presence or absence of defects is determined for images with different irradiation time conditions, and detection results are obtained. At this time, the image processing parameter can be set by using the method described with reference to FIG.

In S203, an irradiation time is selected based on the detection result corresponding to each irradiation time condition (details will be described later). The selection of the irradiation time condition (imaging condition) from the plurality of irradiation time condition candidates (imaging time condition candidates) is thus completed.

FIG. 10 is a graph showing a method for selecting an imaging condition from a plurality of imaging condition candidates in the present embodiment. Here, the number of defects (detection result) corresponding to each irradiation time (imaging condition) was obtained by the method described in FIG. 9 and plotted as the irradiation time dependency 301 of the number of defects. Under a long irradiation time condition, the number of defects is close to the number of defects 302 of the reference defect data. As the irradiation time condition becomes shorter, the number of defects decreases. This is because a threshold value is set so as to discriminate a luminance value change due to quantum noise from a non-defective portion. That is, as the irradiation time condition is shortened, the change in the luminance value due to the quantum noise increases and becomes close to the change in the luminance value of the actual defect. Based on the above, the number of missed defects can be estimated from the difference 303 between the number of defects in each irradiation time condition and the number of defects in the reference defect data. It is possible to select an appropriate irradiation time condition from a plurality of irradiation time condition candidates by setting the level of the number of missed defects that can be allowed in advance.

9 and 10, the method of selecting an appropriate condition using the irradiation time condition as an example of the imaging condition has been described. However, the imaging condition may be applied to other imaging conditions that affect the image quality of the X-ray transmission image ( For example, X-ray source output, X-ray focal spot size, magnification, presence / absence or type of X-ray filter, X-ray energy, etc.).

As described above, according to the present embodiment, it is possible to automatically set appropriate imaging conditions by setting a plurality of imaging condition candidates in advance.

In this embodiment, the image processing parameters are automatically set in the second embodiment so that the image processing parameters are automatically selected based on the number of matches or the number of mismatches between the reference defect data and the defect detection positions of the plurality of detection result candidates. I did it. In addition, the imaging condition is automatically set by selecting the imaging condition based on the number of coincidence or the number of mismatches of the defect detection positions between the plurality of detection results obtained from the plurality of imaging condition candidates and the reference defect data. did. Since the configuration of the X-ray inspection apparatus and the method for acquiring the reference defect data are the same as those in the first embodiment, the differences will be described with reference to FIGS.

FIG. 11 is a flowchart showing processing for selecting an image processing parameter from a plurality of image processing parameter candidates and processing for selecting an imaging condition from a plurality of imaging condition candidates in the present embodiment. Here, an example in which the imaging condition is the irradiation time condition will be described. Since S101 to S104 are the same as the description of FIG. S201 is similar to the description of FIG.

In S306, image processing is executed on each of the images obtained in S201 with a plurality of preset image processing parameter candidates to obtain a plurality of detection result candidates. Here, the image processing parameters are the same image processing parameters as those in S102, and are, for example, threshold values for threshold processing (see Example 1).

In S307, an image processing parameter is selected for each irradiation time condition candidate based on the number of matches or the number of mismatches between the reference defect data and the defect detection position of the detection result candidate (details will be described later).

In S308, the defect detection result candidate corresponding to the image processing parameter determined in S307 is set as the defect detection result.

In S309, an irradiation time is selected based on the detection result for each irradiation time condition candidate obtained in S308 (details will be described later). This completes the selection of the image processing parameter and the selection of the imaging condition. Note that the flowchart described in FIG. 11 may be changed in order or performed in parallel without departing from the gist of the present invention.

Next, a method for selecting an image processing parameter based on the number of coincidences or the number of mismatches between the reference defect data and the defect detection positions of a plurality of detection result candidates will be described with reference to FIGS. FIG. 12 is a schematic diagram illustrating a method for calculating the number of matches and the number of mismatches between the reference defect data and the defect detection position of the detection result candidate in this embodiment. Here, it is shown that reference defect data is obtained in S104, and a defect is detected at the position of the defect 402 in the visual field 401. In S306, a detection result candidate is obtained and a defect is detected at the position of the defect 403. The coincidence determination range 404 is set in advance, and whether the position coordinates match or does not match is determined based on whether or not the position coordinates of the defect 403 are inside the match determination range 404 from the position coordinates of the defect 402. The number of matches is counted as 1 when inside, and the number of mismatches is counted as 1 when outside. As described above, position coordinates match or mismatch are determined for all defects included in the detection result candidate, and the number of matches and the number of mismatches are calculated.

FIG. 13 is a graph showing a method of selecting an image processing parameter based on the number of coincidences or the number of inconsistencies between reference defect data and defect detection positions of a plurality of detection result candidates in this embodiment. A threshold value will be described as an example of the image processing parameter. Here, in S307, by using the method described with reference to FIG. 12, the number of coincidences and the number of mismatches of position coordinates are calculated for a plurality of detection result candidates, and plotted as the threshold dependence 501 of the coincidence ratio and the threshold dependence 502 of the mismatch percentage. did. The coincidence rate is calculated for each threshold value by (number of position coordinate matches) / (number of defects in reference defect data), and the mismatch rate is calculated by (number of position coordinate mismatches) / (number of defects of detection result candidates). The higher the matching rate, the less missed, and the higher the mismatch rate, the more overdetection. As the threshold value is lowered, a minute change in luminance value is discriminated as a defective portion, and the coincidence rate increases. Also, the luminance value change due to quantum noise is erroneously determined as a defective part, so the mismatch rate also increases. Conversely, if the threshold value is increased, a minute luminance value change is determined as a non-defective part. For this reason, the coincidence rate and the non-coincidence rate decrease together. From the viewpoint of the performance of the inspection equipment, it is desirable that the match rate is high and the mismatch rate is low. Set the standard match rate and mismatch rate in advance, and select the appropriate threshold value by selecting the threshold value that satisfies the standard. It becomes possible to do.

Next, a method for selecting the imaging condition based on the number of coincidence or the number of mismatches in the defect detection position between the detection results obtained with a plurality of imaging condition candidates and the reference defect data will be described with reference to FIG. An irradiation time condition will be described as an example of the imaging condition. Here, by using the method described with reference to FIGS. 12 and 13, the position coordinate coincidence rate and the disagreement rate are calculated, the disagreement rate at which the disagreement rate is the lowest in the threshold range where the coincidence rate satisfies the standard is calculated, and irradiation is performed. Plotted as time discordance rate dependence 601. The shorter the irradiation time, the closer the brightness value change due to quantum noise approaches the brightness value change due to defects.If the threshold is lowered to obtain a matching rate that satisfies the criteria, the brightness value change due to quantum noise will be identified as a defective part. , The discrepancy rate becomes high. When the irradiation time is shortened, the quantum noise is reduced, so that the mismatch rate is lowered. From the viewpoint of the throughput of the inspection apparatus, it is desirable that the irradiation time is short and the mismatch rate is low. By setting a reference mismatch rate in advance, it is possible to obtain the shortest irradiation time condition that satisfies the reference from the irradiation time mismatch rate dependency 601, and to select an appropriate irradiation time condition.

As described above, according to this embodiment, when a plurality of image processing parameter candidates and a plurality of imaging condition candidates are set in advance, appropriate image processing parameters and imaging conditions can be automatically set.

In this embodiment, a GUI (Graphical User Interface) of a system for implementing the present invention will be described with reference to FIGS.

FIG. 15 shows an imaging condition setting screen of the GUI screen in this embodiment. There are four GUI screens: an inspection mode screen 130 for inspecting a wafer, a recipe creation screen 131 for creating an inspection recipe, an imaging condition setting screen 132, and a load / unload screen 133 for loading and unloading a wafer. Has a sub-screen.

In the imaging condition setting screen 132, the stage parameters such as the tube voltage of the X-ray source, the probe current, the irradiation time, the total number of sheets, the stage parameter reference distance magnification, the stage height, the tilt angle, and the rotation angle are input.

The reference distance is the distance from the X-ray source 1 to the X-ray detector 5, and when the magnification is input based on this, the stage height is calculated and displayed. Conversely, when the stage height is input, the magnification is calculated. Displayed. The inclination angle is set so that the brightness contrast between the TSV and the void defect is sufficiently high in a range where the TSV does not overlap. The rotation angle is set to be the smallest within a range where TSV images do not overlap.

If the layout information of the wafer is known, the wafer is selected from the pull-down menu 134 and the pre-registered layout information is captured. A die map on the wafer as shown in FIG. 2 is displayed, a user selects a die used for imaging condition setting, further displays TSV layout information in the die, and a user uses the position of a pattern used for imaging condition setting. specify. The coordinates of the translation stage having the designated position as the center of the visual field are calculated and displayed on the stage coordinate display window 135, and the stage is moved to that position. Further, when the wafer layout information is not captured, the stage can be moved by inputting coordinates. Alternatively, the stage can be translated and rotated by a control panel directly connected to the stage controller 102, and the display of the stage coordinates is updated with the movement. If the automatic rotation angle calculation button 136 is pressed in this state, the stage rotation angle without overlapping transmission images is automatically calculated, and the rotation angle of the rotary stage 4 is set to that angle.

In the transmission image display window 137, an X-ray transmission image is displayed in real time, an arbitrary straight line 138 can be drawn, and its brightness profile 139 can be displayed. The X-ray source is adjusted while confirming the brightness with the transmission image and the profile, and the stage is adjusted while confirming the size of the visual field and how the TSV is seen. The reference defect data acquisition irradiation time input window 161 allows the user to input a reference defect data acquisition irradiation time, and acquires a reference defect data acquisition image according to the input value. The irradiation time candidate input window 162 allows the user to input the maximum irradiation time, the minimum irradiation time, and the interval, and changes the irradiation time conditions at equal intervals between the minimum irradiation time and the maximum irradiation time according to the input value. Get an image.

FIG. 16 shows the recipe creation screen 131 of the GUI screen in the present embodiment. On the recipe creation screen 131, wafer parameters such as wafer thickness, TSV diameter, TSV height, and defect detection parameters used for image processing and defect detection in defect detection processing are input. Further, a process of selecting a wafer from the pull-down menu 134 and designating an inspection area is performed. First, when the die selection button 140 is pressed, a die map on the wafer as shown in FIG. 2 is displayed, and the user selects an inspection target die. Next, the TSV layout information in the die is displayed by pressing the layout display button 141, and the inspection area is designated by the user. The size and inclination of the inspection area are calculated according to the magnification, detector inclination angle, and stage rotation angle set on the imaging condition setting screen, and are set by arranging the inspection area of the size and inclination on the layout. . The coordinates of the translation stage in which the set area is included in the field of view are calculated and displayed in the stage coordinate display window 135, and the stage is moved to that position. The stage position may be finely adjusted by coordinate input or a control panel directly connected to the stage controller 102.

Since an X-ray transmission image is displayed in the transmission image display window 137, when the stage position is determined, the add button 142 is pushed to the inspection object. With this process, one inspection area can be designated. In order to designate a plurality of inspection areas, the area designation from the layout display, the confirmation of the X-ray transmission image, and the addition of the inspection object are repeated. At this time, the set inspection region is displayed in the second and subsequent layout displays. Various parameters and inspection area information set on the recipe creation screen 131 are saved together with parameters set on the imaging condition setting screen 132 when the recipe save button 143 is pressed.

The threshold candidate input window 151 allows the user to input a maximum threshold value, a minimum threshold value, and an interval, and obtains a plurality of defect detection result candidates according to the input value. The match rate / mismatch rate input window 152 allows the user to input the match rate and mismatch rate standard values, and the image analysis unit 105 calculates threshold values, reference defect data, and the like based on the input value and the result of the defect detection candidate. Note that the number of matches or the number of matches may be used instead of the match rate and the mismatch rate.

In the inspection mode screen 130, the recipe created on the recipe creation screen 131 can be selected, and inspection can be started / stopped. In addition, inspection results can be stored, read out and analyzed (not shown). The inspection results include TSV position information of the world coordinates of the TSV or relative coordinates of the visual field center reference, or both, TSV size information of the diameter and height, the depth of the defect from the TSV top, the diameter of the defect in a true circle, It is an image such as a defect size information such as a longitudinal size and a transverse direction size, an image in which a detected defect is marked in an input image, and an image representing a TSV area.

On the load / unload screen 133, the wafer set in the X-ray inspection apparatus 100 can be unloaded and the wafer stored in the hoop can be loaded (not shown).

1: X-ray source, 2: Wafer, 3: Translation stage, 4: Rotation stage, 5: X-ray detector, 6: X-ray shielding wall, 10: Die, 11: TSV, 12: Micro bump, 13: No. 1 layer, 14: second layer, 15: third layer, 20: void, 30, 31: profile, 60: X-ray image, 65: transmission image, 100: X-ray inspection apparatus, 101: X-ray Source controller, 102: Stage controller, 103: X-ray detector controller, 104: Control unit, 105: Image analysis unit, 106: Input / output unit, 200: X-ray

Claims

An X-ray inspection method for obtaining an image of a sample using an X-ray inspection apparatus, performing image processing, and obtaining a defect detection result of the sample,
Obtaining a reference defect data acquisition image under reference defect data acquisition conditions;
Obtaining a plurality of reference defect data candidates using different image processing parameter candidates;
Selecting image processing parameters from the plurality of image processing parameter candidates based on the plurality of reference defect data candidates;
Obtaining reference defect data from the reference defect data candidates corresponding to the selected image processing parameters;
An X-ray inspection method characterized by comprising:
An X-ray inspection method for obtaining an image of a sample using an X-ray inspection apparatus, performing image processing, and obtaining a defect detection result of the sample,
Obtaining a reference defect data acquisition image under reference defect data acquisition conditions;
Obtaining reference defect data from the reference defect data acquisition image;
Obtaining a plurality of images with a plurality of different imaging condition candidates;
Obtaining a plurality of detection results from the plurality of images;
Selecting an imaging condition from the plurality of imaging condition candidates based on a comparison of the reference defect data and the plurality of detection results;
An X-ray inspection method characterized by comprising:
An X-ray inspection method for obtaining an image of a sample using an X-ray inspection apparatus, performing image processing, and obtaining a defect detection result of the sample,
Obtaining a reference defect data acquisition image under reference defect data acquisition conditions;
Obtaining a plurality of reference defect data candidates using different image processing parameter candidates;
Selecting image processing parameters from the plurality of image processing parameter candidates based on the plurality of reference defect data candidates;
Obtaining reference defect data from the image for acquiring reference defect data corresponding to the selected image processing parameter;
Obtaining a plurality of images with a plurality of different imaging condition candidates;
Obtaining a plurality of detection result candidates from the plurality of images;
Obtaining a plurality of detection results from the plurality of detection result candidates;
Selecting an imaging condition from the plurality of imaging condition candidates based on a comparison of the reference defect data and the plurality of detection results;
An X-ray inspection method characterized by comprising:
The X-ray inspection method according to any one of claims 1 to 3,
The X-ray inspection method, wherein the reference defect data acquisition condition is an X-ray irradiation time, which is longer than an irradiation time at the time of inspection.
The X-ray inspection method according to claim 1 or 3,
An X-ray inspection method, wherein an image processing parameter is selected based on a comparison between the reference defect data and the plurality of image processing parameter candidates.
The X-ray inspection method according to claim 5,
An X-ray inspection method, wherein an image processing parameter is selected based on the number of matches or the number of mismatches between the reference defect data and the defect detection positions of the plurality of image processing parameter candidates.
The X-ray inspection method according to claim 1 or 3,
An X-ray inspection method, wherein an image processing parameter is selected based on an image processing parameter candidate dependency of the number of defects.
The X-ray inspection method according to claim 7,
An X-ray inspection method, wherein an image processing parameter is selected based on an inflection point that is dependent on the number of defects of the image processing parameter candidate.
The X-ray inspection method according to claim 2 or 3,
An X-ray inspection method, wherein an imaging condition is selected based on the number of coincidences or the number of mismatches in defect detection positions between a plurality of detection results obtained under each imaging condition and the reference defect data.
The X-ray inspection method according to claim 2 or 3,
An X-ray inspection method, wherein an imaging condition is selected based on a comparison between the reference defect data and the number of defects of the plurality of detection results.
The X-ray inspection method according to any one of claims 1 to 3,
An X-ray inspection method comprising a step of allowing a user to input a reference defect data acquisition condition.
The X-ray inspection method according to claim 1 or 3,
An X-ray inspection method comprising a step of allowing a user to input a plurality of different image processing parameter candidates.
The X-ray inspection method according to claim 2 or 3,
An X-ray inspection method comprising a step of allowing a user to input a plurality of different imaging condition candidates.
The X-ray inspection method according to claim 6,
An X-ray inspection method comprising a step of allowing a user to input a matching rate or a mismatch rate of a reference defect detection position.
The X-ray inspection method according to any one of claims 1 to 3,
An X-ray inspection method comprising a step of allowing a user to input a reference defect size.