Ultrasonic inspection device and ultrasonic inspection method

What is AI technical title?
AI technical title is built by Patsnap AI team. It summarizes the technical point description of the patent document.
The ultrasonic inspection device addresses the challenge of simultaneous high sensitivity and high-speed defect detection in semiconductor wafers by using a dual-probe system for comprehensive defect analysis, achieving efficient and accurate inspection results.

JP2025070728APending Publication Date: 2025-05-02HIATACHI POWER SOLUTIONS CO LTD

View PDF 5 Cites 0 Cited by

Patent Information

Application Number: JP2023181240
Authority / Receiving Office: JP · JP
Patent Type: Applications
Current Assignee / Owner
Filing Date: 2023-10-20
Publication Date: 2025-05-02

AI Technical Summary

Technical Problem

Existing ultrasonic inspection technologies struggle to simultaneously achieve high sensitivity and high-speed defect detection in multi-layered semiconductor wafers, particularly due to limitations in image resolution and processing time.

Method used

The ultrasonic inspection device employs a dual-probe system, where a first array-type probe provides high-speed inspection of a wide area with lower resolution, and a second single-scan probe performs high-sensitivity inspection of a narrower area with higher resolution, integrating defect candidate information from both probes for comprehensive analysis.

Benefits of technology

This approach enables rapid and sensitive detection of defects across the entire surface of multi-layered semiconductor wafers, improving inspection efficiency and accuracy while reducing processing time.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure 2025070728000001_ABST

Patent Text Reader

Abstract

To carry out inspection of an internal state related to an analyte including a multi-layer laminated semiconductor wafer, by using supersonic waves with high sensitivity and at a high speed.SOLUTION: An ultrasonic inspection device 11 comprises: a setting unit 17 that sets a first inspection area and a second inspection area as inspection target ranges related to an analyte 15; a first inspection unit 19 that acquires a first inspection image in the first inspection area, and acquires first defect candidate information on the basis of the first inspection image; a second inspection unit 21 that acquires a second inspection image in the second inspection area, and acquires second defect candidate information on the basis of the second inspection image; and an output unit 29 that outputs a result of inspection of an internal state related to the analyte 15. The output unit 29 integrates the first defect candidate information acquired by the first inspection unit 19 and the second defect candidate information acquired by the second inspection unit 21, and outputs the resultant information as the result of inspection of the internal state related to the analyte 15.SELECTED DRAWING: Figure 1

Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an ultrasonic inspection apparatus and an ultrasonic inspection method for inspecting the internal state of a test object, including a semiconductor wafer stacked in multiple layers, by using ultrasonic waves. [Background technology]

[0002] In advanced semiconductor wafer manufacturing processes, a new technology called "hybrid bonding" has begun to be introduced, in which silicon wafers or silicon wafers and silicon dies are directly bonded together via copper (Cu) electrodes. Therefore, it is becoming increasingly important to inspect the internal state of the wafer for bonding defects and other issues, and provide feedback that will lead to early improvements in process conditions. Ultrasonic testing is an effective way to non-destructively inspect the internal state of a multi-layered test object. In ultrasonic testing, a probe irradiates the test object with ultrasonic waves, receives the reflected waves from the bonded interface, and generates an inspection image based on the intensity of the reflected waves. Appropriate image processing is then performed on the generated inspection image to extract defects.

[0003] To achieve high-sensitivity inspection using ultrasound, the ultrasonic beam aimed at the inspection target is narrowed and a high-resolution inspection image is generated using a single-type ultrasonic probe with a single scan method. However, a single-type ultrasonic probe with a single scan method requires a considerable amount of time to scan the entire surface of the wafer. On the other hand, to realize high-speed inspection using ultrasound, an array-type ultrasonic probe of the electronic scanning type is effective. However, in the case of an array-type ultrasonic probe of the electronic scanning type, a decrease in the resolution of the generated inspection image, that is, a decrease in image quality, is unavoidable compared to the case of using a single-type ultrasonic probe of the single scanning type.

[0004] Non-Patent Document 1 discloses an ultrasonic inspection device that can simultaneously mount both a single-type ultrasonic probe of a single scan type that can realize highly sensitive inspection, and an array-type ultrasonic probe of an electronic scan type that can realize high-speed inspection. Patent Document 1 also discloses an inspection technique for extracting defects from an inspection image taken by ultrasonic waves. In the inspection technique of Patent Document 1, an image of a reference die that is not affected by defects is generated based on an inspection image of the bonding interface of a multilayer laminated wafer obtained by ultrasonic waves, and the image of each die is compared with the reference image, and any location with a large difference in image attributes (brightness) is deemed to be a defect and displayed. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2017-072501 A [Non-patent literature]

[0006] [Non-Patent Document 1] Ultrasonic inspection device FineSAT Hybrid, [online], May 22, 2023, Internet <URL:https: / / www.hitachi-power-solutions.com / product-site / finesat / products / finesat-hybrid / index.html> Summary of the Invention [Problem to be solved by the invention]

[0007] In the inspection technique according to Non-Patent Document 1, a pair of inspection images with different resolutions obtained by an array-type ultrasonic probe and a single-type ultrasonic probe are generally subjected to uniform image processing. Therefore, the type of defect extracted from each inspection image depends on the image resolution. However, in a process using hybrid bonding, the types of defects that occur are diverse. Therefore, in order to provide feedback to process conditions, it is required to quickly and sensitively grasp all kinds of defects that occur on the entire surface of the wafer due to the process. In this regard, the inspection technique according to Non-Patent Document 1 does not meet such a demand.

[0008] In the inspection technique disclosed in Patent Document 1, defect inspection is performed by utilizing the phenomenon that dies having a common circuit pattern are regularly formed at the bonding interface of a wafer. In detail, first, an image of the bonding interface of a wafer is divided into a plurality of images for each die. Next, statistical processing is performed on the divided images for each die, and a reference die image and a dynamic threshold value for suppressing noise are generated based on the results of the statistical processing. Then, minute bonding defects are extracted using these. However, the above-mentioned inspection technique requires a plurality of images of the same pattern structure area for the statistical processing, and therefore it takes a considerable amount of time to complete the inspection.

[0009] The present invention has been made in consideration of the above-mentioned situation, and has an object to provide an ultrasonic inspection apparatus and an ultrasonic inspection method that are capable of using ultrasonic waves to quickly and sensitively inspect the internal state of a specimen, including semiconductor wafers stacked in multiple layers. [Means for solving the problem]

[0010] In order to solve the above problems, an ultrasonic inspection device according to the present invention comprises: An ultrasonic inspection apparatus for inspecting an internal state of a test object including a semiconductor wafer stacked in multiple layers by using ultrasonic waves, comprising: a setting unit that sets a first inspection area and a second inspection area as an inspection target range related to the specimen; a first inspection unit that acquires a first inspection image in the first inspection area using a first resolution and acquires first defect candidate information based on the first inspection image; a second inspection unit that acquires a second inspection image in the second inspection area using a second resolution and acquires second defect candidate information based on the second inspection image; an output unit that outputs an inspection result of an internal condition of the subject, the second inspection area is set within a range of the first inspection area and is smaller than the first inspection area; The second resolution is set to a value higher than the first resolution, The most main feature of the output unit is that it integrates first defect candidate information relating to the first inspection area acquired by the first inspection unit and second defect candidate information acquired by the second inspection unit, and outputs the information as inspection results for the internal condition of the specimen. Effect of the Invention

[0011] According to the present invention, the internal state of a test object including semiconductor wafers stacked in multiple layers can be inspected with high sensitivity and at high speed by using ultrasonic waves. Problems, configurations, and effects other than those described above will be described in detail in the following embodiments. [Brief description of the drawings]

[0012] [Figure 1] 1 is a functional block diagram illustrating a concept of an ultrasonic inspection device according to an embodiment of the present invention. [Figure 2A] FIG. 2 is a functional block diagram showing the concept of an ultrasonic inspection device according to an embodiment of the present invention from a different perspective than that of FIG. [Figure 2B] 1 is a schematic configuration diagram of an ultrasonic inspection device according to an embodiment of the present invention. [Diagram 3] 1 is a flow chart conceptually showing a process flow of an ultrasonic inspection method according to an embodiment of the present invention. [Figure 4]1 is a perspective view of a manufacturing process for a semiconductor wafer as a test object, and a semiconductor wafer having a multilayer structure manufactured by the manufacturing process; [Figure 5A] 1 is an explanatory diagram illustrating a first inspection image of a bonding interface of a semiconductor wafer that is a test object. [Figure 5B] 13 is an explanatory diagram illustrating a second inspection image of a bonding interface of a semiconductor wafer that is a test object. FIG. [Figure 6] 1 is an explanatory diagram conceptually showing a flow of a defect candidate extraction process in a first defect extraction unit provided in an ultrasonic inspection device according to an embodiment of the present invention and a setting process related to a second investigation area. FIG. [Figure 7] 7 is an explanatory diagram illustrating an example of a defect type extracted by a first defect extraction unit illustrated in FIG. 6. [Figure 8] 4 is an explanatory diagram conceptually showing the flow of a second defect extraction process related to a second defect extraction unit included in the ultrasonic inspection device according to the embodiment of the present invention; FIG. [Figure 9] 3 is an explanatory diagram of a defect map showing distribution information of defect candidates extracted by a first defect extraction section and a second defect extraction section included in the ultrasonic inspection apparatus according to the embodiment of the present invention. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] An ultrasonic inspection device and an ultrasonic inspection method according to an embodiment of the present invention will be described in detail with reference to appropriate drawings. In the description of the ultrasonic inspection device according to the embodiment of the present invention, components having common functions are given common reference numerals, and duplicated descriptions thereof will be omitted.

[0014] [Concept of ultrasonic inspection device 11 according to an embodiment of the present invention] First, the concept of an ultrasonic inspection device 11 according to an embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a functional block diagram showing the concept of an ultrasonic inspection device 11 according to an embodiment of the present invention. The ultrasonic inspection device 11 according to the embodiment of the present invention has a function of using ultrasonic waves to inspect the internal state of a test object including a semiconductor wafer 15 formed by stacking multiple silicon wafers on which a plurality of identical circuit patterns are regularly formed, and a MEMS (Micro Electro Mechanical Systems) wafer. The ultrasonic inspection device 11 inspects a semiconductor wafer 15 manufactured by stacking multiple wafers on which dies (synonymous with chips) are regularly formed, and inspects the state of the bonding interface between the stacked wafers using ultrasonic waves.

[0015] In detail, as shown in FIG. 1, an ultrasonic inspection apparatus 11 according to an embodiment of the present invention is configured to include a setting unit 17 that sets a first inspection area and a second inspection area as an inspection target range for a specimen including a semiconductor wafer 15 stacked in multiple layers, a first inspection unit 19 that acquires a first inspection image in the first inspection area using a first resolution and acquires first defect candidate information based on the first inspection image, a second inspection unit 21 that acquires a second inspection image in the second inspection area using a second resolution and acquires second defect candidate information based on the second inspection image, and an output unit 29 that outputs inspection results of the internal condition of the specimen.

[0016] The setting unit 17 sets the second inspection area based on the first defect candidate information related to the first inspection area acquired by the first inspection unit 19. For example, when the first defect candidate is extracted in a predetermined range of the first inspection area, the setting unit 17 sets this predetermined range as the second inspection area. Therefore, the second inspection area is set within the range of the first inspection area but in a narrower range than the first inspection area. The method of setting the second inspection area by the setting unit 17 will be described in detail later.

[0017] The second resolution used in the second inspection area is set to a higher value than the first resolution used in the second inspection area, in order to achieve the intended object of the present invention (to perform inspection of the internal state of a subject with high sensitivity and high speed using ultrasonic waves). That is, in the ultrasonic inspection apparatus 11 according to an embodiment of the present invention, the setting unit 17 sets a wide range, such as the entire surface of the semiconductor wafer 15, as a first inspection area, and the first inspection unit 19 acquires a first inspection image using a first resolution in the first inspection area set over a wide range, and acquires first defect candidate information based on the first inspection image.

[0018] On the other hand, the setting unit 17 sets a range narrower than the first inspection area as a second inspection area based on the first defect candidate information related to the first inspection area, and the second inspection unit 21 acquires a second inspection image using a second resolution in the second inspection area set in a range narrower than the first inspection area, and acquires second defect candidate information based on the second inspection image. Here, in order to realize high speed, the first inspection unit 19 uses the first ultrasonic probe 13a of array type (electronic scanning type) to obtain a first inspection image in a first inspection area using a first resolution, and obtains distribution information related to a first defect candidate present in the first inspection area based on the first inspection image. The first ultrasonic probe 13a of array type (electronic scanning type) performs electronic scanning in the Y-axis direction for the number of transducers. Therefore, high speed scanning (inspection) is possible compared to the case where the second ultrasonic probe 13b of single type (single scanning type) is used.

[0019] In short, the high-speed mode using the array-type first ultrasonic probe 13a is effective for scanning (inspecting) a wide range in a short time, but is inferior in terms of resolution (high speed but low resolution) compared to the case using the single-type (single scan method) second ultrasonic probe 13b. On the other hand, in the second inspection section 21, in order to realize high sensitivity, a second inspection image is obtained using a single type (single scan type) second ultrasonic probe 13b, and distribution information related to second defect candidates existing in the second inspection area is obtained based on the second inspection image. The single type (single scan type) second ultrasonic probe 13 generally uses ultrasonic waves of a higher frequency than the array type (electronic scan type) first ultrasonic probe 13a.

[0020] Therefore, it is possible to obtain a more highly detailed inspection image (high sensitivity). The high sensitivity mode using the single-type second ultrasonic probe 13b is effective for scanning (inspecting) a relatively narrow range with high sensitivity. However, it is inferior to the case where the array-type (electronic scanning type) first ultrasonic probe 13a is used in that it takes a longer inspection time (high resolution but low speed). Therefore, in the ultrasonic inspection apparatus 11 according to an embodiment of the present invention, a primary inspection (high-speed mode) is performed by the first inspection unit 19, which inspects a wide range (first inspection area), to narrow down the area in which the first defect candidates are distributed, and a secondary inspection (high-sensitivity mode) is performed by the second inspection unit 21, which inspects the narrower range (second inspection area), to reveal the location of the second defect candidates and their defect types.

[0021] The output unit 29 integrates the first defect candidate information acquired by the first inspection unit 19 and the second defect candidate information acquired by the second inspection unit 21, and outputs the integrated defect candidate information as an inspection result of the internal state of the semiconductor wafer 15 (test object). The integrated defect candidate information is displayed, for example, on a display screen of a display monitor belonging to the GUI 37. According to the ultrasonic inspection device 11 of the embodiment of the present invention, it is possible to obtain the desired effect of inspecting the internal state of the semiconductor wafer 15 (test object) with high sensitivity and at high speed using ultrasonic waves.

[0022] Furthermore, the ultrasonic inspection apparatus 11 according to an embodiment of the present invention may be configured to further include a defect type database (defect type DB) 23 that stores defect type information corresponding to each of a plurality of types of defect candidates (see, for example, FIG. 7). In this case, the first inspection unit 19 acquires position information related to the first defect candidate based on distribution information related to the first defect candidate acquired by the first inspection unit 19, and also acquires defect type information corresponding to the first defect candidate by referring to the defect type DB 23.

[0023] Furthermore, the second inspection section 21 acquires position information relating to the second defect candidate acquired based on the second inspection image, and also acquires corresponding defect-type information by referring to the defect-type DB 23. The output unit 29 outputs defect candidate information obtained by integrating the position information and defect type information related to the first defect candidate and the position information and defect type information related to the second defect candidate as an inspection result of the internal state related to the semiconductor wafer 15 (test object). The integrated defect candidate information is displayed, for example, on a display screen of a display monitor provided in the GUI 37 (see FIG. 2B).

[0024] With this configuration, defect candidate information that integrates the position information and defect type information related to the first and second defect candidates can be output as the inspection result of the internal state related to the semiconductor wafer 15 (test object). As a result, in addition to the above-mentioned expected effects, detailed defect candidate information regarding what types of defect candidates are present, where and how many of them are present can be provided. Furthermore, in the ultrasonic inspection apparatus 11 according to an embodiment of the present invention, the first inspection unit 19 may be configured to add first identification information for a first defect candidate to the position information and defect type information for the first defect candidate, and the second inspection unit 21 may be configured to add second identification information for a second defect candidate to the position information and defect type information for the second defect candidate.

[0025] The output unit outputs the defect candidate information to which the first identification information and the second identification information are respectively added as an inspection result of the internal state of the semiconductor wafer 15 (test object). The defect candidate information is displayed on the display screen of the display unit 31. With this configuration, defect candidate information that integrates the position information and defect type information related to the first and second defect candidates and to which the first identification information and the second identification information are respectively added can be output as the inspection result of the internal state related to the semiconductor wafer 15 (test subject). As a result, in addition to the above-mentioned expected effects, detailed defect candidate information regarding what types of defect candidates are present, where they are located, and how many of them are present, can be provided together with the origins of the first and second defect candidates.

[0026] Furthermore, the ultrasonic inspection apparatus 11 according to an embodiment of the present invention may be configured to further include a defect occurrence cause database (defect occurrence cause DB) 25 that stores information on defect occurrence causes in the manufacturing process related to the semiconductor wafer 15 (subject) in association with each of a plurality of defect type information, and an estimation unit 27 that refers to the defect type database to estimate the defect occurrence cause corresponding to the defect candidate information. With this configuration, the estimation unit 27 can estimate a defect occurrence factor in the manufacturing process related to the semiconductor wafer 15 (test object) that corresponds to the defect candidate information. Therefore, in addition to the desired effects described above, it is possible to provide information on the defect occurrence factor in the manufacturing process related to the test object.

[0027] Furthermore, in the ultrasonic inspection device 11 according to an embodiment of the present invention, the setting unit 17 may be configured to extract an area in which the first defect candidates are irregularly distributed based on distribution information relating to the first defect candidates acquired by the first inspection unit 19 using the first ultrasonic probe 13a, and set the extracted area as the second inspection area. With this configuration, the setting unit 17 sets, as the second inspection area, an area in which defect candidates extracted based on distribution information relating to the first defect candidates acquired by the first inspection unit 19 using the first ultrasonic probe 13a are irregularly distributed, so that the inspection target area can be appropriately narrowed down in the secondary inspection (high sensitivity mode) by the second inspection unit 21.

[0028] Therefore, it is possible to accurately identify the location of the second defect candidate and its defect type, and as a result, the desired effect of inspecting the internal state of the semiconductor wafer 15 (test object) with high sensitivity and high speed using ultrasound can be achieved at an even higher level. Furthermore, in the ultrasonic inspection device 11 according to an embodiment of the present invention, the setting unit 17 may be configured to extract a normal area from which no defect candidate is extracted based on distribution information relating to the first defect candidate acquired by the first inspection unit 19 using the first ultrasonic probe 13a, and to set the extracted normal area as a non-inspection area.

[0029] With this configuration, the setting unit 17 extracts normal areas from which no defect candidates are extracted based on the distribution information relating to the first defect candidates acquired by the first inspection unit 19 using the first ultrasonic probe 13a, and sets the extracted normal areas as non-inspection areas, so that, similarly to the above, it is possible to appropriately narrow down the inspection target areas in the secondary inspection (high sensitivity mode) relating to the second inspection unit 21. Therefore, it is possible to quickly identify the location of the second defect candidate and its defect type by eliminating inspection of areas that are essentially unnecessary, and as a result, the desired effect of inspecting the internal state of the semiconductor wafer 15 (subject) with high sensitivity and high speed using ultrasound can be achieved at an even higher level.

[0030] [Concept of the ultrasonic inspection device 11 according to the embodiment of the present invention based on a different viewpoint from the above] Next, a concept based on another aspect of the ultrasonic inspection device 11 according to the embodiment of the present invention will be described with reference to Fig. 2A. Fig. 2A is a functional block diagram showing the concept of the ultrasonic inspection device 11 according to the embodiment of the present invention from another aspect based on Fig. 1. As shown in FIG. 2A, the ultrasonic inspection device 11 according to the present invention is configured to include an ultrasonic exploration section 22, an A / D conversion section 18, an image processing section 24, and an overall control section 33.

[0031] The ultrasonic inspection section 22 is configured with a first ultrasonic probe 13a and a first flaw detector 16a belonging to a first inspection system, and a second ultrasonic probe 13b and a second flaw detector 16b belonging to a second inspection system. A characteristic of ultrasonic waves is that as they propagate through the inside of the test object, if there is a boundary (interface) where the material properties (acoustic impedance) change, some of the waves will be reflected. In particular, if there is a gap (air layer) in the direction of ultrasonic wave propagation, most of the waves will be reflected. Therefore, based on the reflected strength of ultrasonic waves, it is possible to extract defects such as voids and peeling that exist at the bonding interface of the wafer.

[0032] The first flaw detector 16a belonging to the first inspection system drives the first ultrasonic probe 13a belonging to the first inspection system by applying a predetermined pulse signal to the first ultrasonic probe 13a. As described above, the first ultrasonic probe 13a is an array-type (electronic scanning type) ultrasonic probe that is preferably used in a high-speed mode in which a wide range is scanned (inspected) in a short time. The first ultrasonic probe 13a driven by the first flaw detector 16a irradiates ultrasonic waves at a predetermined time interval to a first inspection region (whole wafer surface) of the semiconductor wafer 15, which is the test object. When the ultrasonic waves are incident on the semiconductor wafer 15, a first reflected wave 14a is generated from the wafer surface and the wafer bonding interface. The first ultrasonic probe 13a receives the first reflected wave 14a. The first flaw detector 16a performs required signal processing on the first reflected wave 14a, and converts and outputs a reflection intensity signal related to the first reflected wave 14a.

[0033] On the other hand, the second flaw detector 16b belonging to the second inspection system drives the second ultrasonic probe 13b belonging to the second inspection system by applying a predetermined pulse signal to the second ultrasonic probe 13b. As described above, the second ultrasonic probe 13b is a single type (single scan type) ultrasonic probe that is preferably used in a high sensitivity mode in which a narrow range is scanned (inspected) with high sensitivity. The second ultrasonic probe 13b driven by the second flaw detector 16b irradiates ultrasonic waves to the second inspection area (narrower than the first inspection area) at a predetermined time interval. When the ultrasonic waves are incident on the semiconductor wafer 15, a second reflected wave 14b is generated from the wafer surface and the wafer bonding interface. The second ultrasonic probe 13b receives the second reflected wave 14b. The second flaw detector 16b performs required signal processing on the second reflected wave 14b, and converts and outputs a reflection intensity signal related to the second reflected wave 14b.

[0034] In the ultrasonic inspection unit 22, the operations of the first ultrasonic probe 13a and the first flaw detector 16a belonging to the first inspection system, and the operations of the second ultrasonic probe 13b and the second flaw detector 16b belonging to the second inspection system are sequentially executed in time series such that the operation of the second inspection system is performed after the operation of the first inspection system. In other words, the operation of the first inspection system and the operation of the second inspection system are not executed in parallel in time. The A / D converter 18 converts the reflection intensity signal related to the first reflected wave 14a belonging to the first exploration system or the reflection intensity signal related to the second reflected wave 14b belonging to the second exploration system into digital waveform data, and outputs the converted digital waveform data to the image processor 24. Here, the reflection intensity signal related to the first reflected wave 14a belonging to the first exploration system includes information on a first inspection image related to a first inspection area. Similarly, the reflection intensity signal related to the second reflected wave 14b belonging to the second exploration system includes information on a second inspection image related to a second inspection area.

[0035] The image processing unit 24 has a function of generating a first inspection image for the first inspection area and a second inspection image for the second inspection area, and also has a function of performing predetermined image processing on the generated first and second inspection images. In order to realize such a function, the image processing unit 24 is configured with an image generation unit 26, a first defect extraction unit 28a belonging to the first inspection system, a second defect extraction unit 28b belonging to the second inspection system, and an output unit 29.

[0036] The image generating unit 26 generates a first inspection image for the first inspection area based on the reflection intensity signal for the first reflected wave 14a, and generates a second inspection image for the second inspection area based on the reflection intensity signal for the second reflected wave 14b. In more detail, the image generating unit 26 converts the digital waveform data for each of the multiple scanning positions related to the reflection intensity signal after A / D conversion by the A / D converting unit 18 into a grayscale value (for example, when generating an inspection image with 256 gradations, any one of 0 to 255 is assigned as the grayscale value). As a result, the image generating unit 26 generates a cross-sectional image of the wafer bonding surface of the laminated structure (an image of an XY cross section related to a specific depth (Z value) from the surface of the wafer and parallel to the wafer surface).

[0037] The first defect extraction unit 28a belonging to the first inspection system performs a predetermined image processing on the first inspection image to extract a first defect candidate related to the first inspection area and assign a defect type to the extracted first defect candidate. The second defect extraction unit 28b belonging to the second inspection system performs a predetermined image processing on the second inspection image to extract second defect candidates related to the second inspection area and assign a defect type to the extracted second defect candidates.

[0038] Here, the first defect extraction unit 28a and the second defect extraction unit 28b each perform a different defect extraction process (image processing). As described above, the first ultrasonic probe 13a and the second ultrasonic probe 13b each have a different ultrasonic wave irradiation method and a different reflected wave reception method, and the resolution of the inspection image obtained by each is also different. The first defect extraction process by the first defect extraction unit 28a and the second defect extraction process by the second defect extraction unit 28b will be described in detail later.

[0039] The output unit 29 integrates the first defect candidate information extracted by the first defect extraction unit 28a and the second defect candidate information extracted by the second defect extraction unit 28b, and outputs the integrated information to the overall control unit 33 as the inspection result of the internal condition of the semiconductor wafer 15 (subject). The overall control unit 33 plays a role in overall control of the operations of the ultrasonic exploration unit 22, the A / D conversion unit 18, and the image processing unit 24. The operation of the overall control unit 33 will be described in detail later.

[0040] [Overall configuration of ultrasonic inspection device 11 according to an embodiment of the present invention] Next, a schematic configuration of an ultrasonic inspection device 11 according to an embodiment of the present invention will be described with reference to Fig. 2B. Fig. 2B is a schematic configuration diagram of the ultrasonic inspection device 11 according to an embodiment of the present invention. In describing the ultrasonic inspection device 11 shown in Fig. 2B, a coordinate system of three orthogonal axes X, Y, and Z shown in Fig. 2B will be used to represent directions. As shown in FIG. 2B, in the ultrasonic inspection device 11 according to an embodiment of the present invention, the ultrasonic exploration unit 22 (see FIG. 2A) includes a first scanner 41a and a second scanner 41b that drive the first ultrasonic probe 13a and the second ultrasonic probe 13b, respectively, to scan in three axial directions.

[0041] Each of the first scanner 41a and the second scanner 41b is supported by a scanner table 43 so as to be capable of scanning and driving in three axial directions. A water tank 47 is provided horizontally on the scanner table 43. Each of the first scanner 41a and the second scanner 41b is disposed so as to straddle the water tank 47. Water 49, which is a good propagation medium for ultrasonic waves, is stored in the water tank 47 up to, for example, the height indicated by the dotted line. A semiconductor wafer 15, which is a test object, is placed at the bottom (underwater) of the water tank 47.

[0042] The first scanner 41a and the second scanner 41b are each connected to a driving unit 51. The driving unit 51 drives the first scanner 41a and the second scanner 41b to scan in three axial directions relative to the scanner base 43. The first ultrasonic probe 13a and the second ultrasonic probe 13b are supported by the first scanner 41a and the second scanner 41b via the first holder 45a and the second holder 45b, respectively. The ultrasonic inspection device 11 according to the embodiment of the present invention performs a predetermined scan using the first ultrasonic probe 13a and the second ultrasonic probe 13b, respectively, in accordance with the scanning movement of the first scanner 41a and the second scanner 41b in accordance with the scanning drive signal of the driving unit 51. In this way, the ultrasonic inspection device 11 is configured to perform ultrasonic inspection of the first and second inspection regions of the semiconductor wafer 15, which is the test object.

[0043] The first ultrasonic probe 13a and the second ultrasonic probe 13b are connected to first and second flaw detectors 16a and 16b via first and second cables 22a and 22b, respectively. As shown in FIG. 2B, the ultrasonic inspection device 11 according to the embodiment of the present invention is configured to include an A / D conversion unit 18, an image processing unit 24, and an overall control unit 33 (all of which are shown in FIG. 2A).

[0044] As shown in FIG. 2B, the image processing unit 24 is configured to include an image generation unit 26, a first defect extraction unit 28a, a second defect extraction unit 28b, and an output unit 29 (all see FIG. 2A), as well as the setting unit 17 (see FIG. 1). The setting unit 17 basically plays a role of setting a narrower range than the first inspection area as the second inspection area based on the first defect candidate information related to the first inspection area. Also, the setting unit 17 receives condition parameters related to one or more combinations inputted via a Graphical User Interface (GUI) 37 in accordance with a control command from the overall control unit 33, and sets the received condition parameters appropriately for the first and second defect extraction units 28a, 28b and the output unit 29.

[0045] Here, the condition parameters include, for example, an operation mode, an electronic scanning method, a diagnostic depth, an ultrasonic transmission frequency, a transmission power, a reception gain, and a data processing condition. These condition parameters are stored in a condition parameter database (condition parameter DB) 34 and are read out as necessary. Connected to the overall control unit 33 are a memory unit 35 for storing various information regarding each defect candidate extracted by the image processing unit 24, such as images, characteristics such as the number and type of defect, position coordinates, dimensions, etc., and a GUI 37 including a keyboard and a display monitor for inputting user requests.

[0046] The overall control unit 33 is equipped with a CPU that performs various types of control, accepts condition parameters etc. in accordance with the user's requests, and generally controls the operations of the ultrasonic exploration unit 22, A / D conversion unit 18, image processing unit 24, GUI 37, and drive unit 51.

[0047] [Ultrasonic inspection method according to an embodiment of the present invention] Next, an ultrasonic inspection method according to an embodiment of the present invention will be described with reference to Fig. 3. Fig. 3 is a flow chart conceptually showing the flow of processing of the ultrasonic inspection method according to the embodiment of the present invention. As shown in FIG. 3, the ultrasonic inspection method according to the present invention is an ultrasonic inspection method used when using ultrasonic waves to inspect the internal state of a specimen including a semiconductor wafer 15 (see FIGS. 1, 2A, and 2B) stacked in multiple layers, and includes a setting step of setting a first inspection area and a second inspection area as an inspection target range of the semiconductor wafer 15 (specimen), a first inspection step (step S11) of acquiring a first inspection image in the first inspection area using a first resolution (steps S111-S112) and acquiring first defect candidate information based on the first inspection image (step S113), a second inspection step (step S12) of acquiring a second inspection image in the second inspection area using a second resolution (steps S121-S122) and acquiring second defect candidate information based on the second inspection image (step S123), and an output step (step S13) of outputting inspection results of the internal state of the semiconductor wafer 15 (specimen).

[0048] In the setting step, a second inspection area is set (step S114) based on the first defect candidate information acquired in the first inspection step (step S11). Setting information 111 regarding the second inspection area is referred to in the second inspection step (step S12). The second inspection area is set within the range of the first inspection area but in a narrower range than the first inspection area, and the second resolution is set to a value higher than the first resolution. In the output process (step S13), the first defect candidate information (step S113) acquired in the first inspection process (step S11) and the second defect candidate information (step S123) acquired in the second inspection process (step S12) are integrated and output as inspection results of the internal state of the semiconductor wafer 15 (test object).

[0049] Here, in the first inspection process (step S11), in order to achieve high speed, a first ultrasonic probe 13a of an array type (electronic scanning type) is used to obtain a first inspection image in a first inspection area using a first resolution, and distribution information related to first defect candidates present in the first inspection area is obtained based on the first inspection image. As described above, the high-speed mode using the array-type first ultrasonic probe 13a is effective for scanning (inspecting) a wide range in a short time, but is inferior in terms of resolution (high speed but low resolution) compared to the case using the single-type (single scan method) second ultrasonic probe 13b.

[0050] In contrast, in the second inspection process (step S12), in order to achieve high sensitivity, a second inspection image is obtained using a single-type (single-scan type) second ultrasonic probe 13b, and distribution information regarding second defect candidates present in the second inspection area is obtained based on the second inspection image. As described above, the high sensitivity mode using the single-type second ultrasonic probe 13b is effective for scanning (inspecting) a relatively narrow range with high sensitivity. However, it is inferior to the case where the array-type (electronic scanning type) first ultrasonic probe 13a is used in that it takes a long inspection time (high resolution but low speed).

[0051] Therefore, in the ultrasonic inspection method according to an embodiment of the present invention, a primary inspection (high-speed mode) for a first inspection process (step S11) that inspects a wide range (first inspection area) is performed to narrow down the area in which the first defect candidate is distributed, and a secondary inspection (high-sensitivity mode) for a second inspection process (step S12) that inspects the narrower range (second inspection area) that has been narrowed down is performed to reveal the location of the second defect candidate and its defect type. In the output process (step S13), the first defect candidate information (step S113) acquired in the first inspection process (step S11) and the second defect candidate information (step S123) acquired in the second inspection process (step S12) are integrated and output as an inspection result of the internal state of the semiconductor wafer 15 (test object). The integrated defect candidate information is displayed, for example, on a display screen of a display monitor provided in the GUI 37 (see FIG. 2B).

[0052] According to the ultrasonic inspection method of the embodiment of the present invention, it is possible to obtain the desired effect of inspecting the internal state of the semiconductor wafer 15 (test object) with high sensitivity and at high speed using ultrasonic waves. Furthermore, in the ultrasonic inspection method according to the embodiment of the present invention, the first inspection step (step S11) may be configured to acquire a first inspection image in a first inspection area using a first resolution (steps S111-S112), and acquire distribution information (step S113) related to a first defect candidate present in the first inspection area based on the first inspection image. Note that the distribution information related to the first defect candidate corresponds to a subordinate concept of the first defect candidate information. In other words, the first defect candidate information is a concept that includes the distribution information related to the first defect candidate.

[0053] According to the ultrasonic inspection method of the embodiment of the present invention, in addition to the above-mentioned desired effects, defect candidate information regarding where and how many first defect candidates are present can be provided. Furthermore, in the ultrasonic inspection method according to an embodiment of the present invention, a configuration may be adopted in which, in the first inspection process (step S11), position information for the first defect candidate is acquired based on distribution information for the first defect candidate acquired in the first inspection process, and defect type information corresponding to the first defect candidate is acquired by referring to a defect type database 23 (see FIG. 1) that stores defect type information corresponding to each of a plurality of types of defect candidates; in the second inspection process (step S12), position information for the second defect candidate acquired based on the second inspection image is acquired, and corresponding defect type information is acquired by referring to the defect type database 23; and in the output process (step S13), defect candidate information that combines the position information and defect type information for the first defect candidate and the position information and defect type information for the second defect candidate is output as an inspection result for the internal state of the semiconductor wafer 15 (subject).

[0054] With this configuration, defect candidate information that integrates the position information and defect type information related to the first and second defect candidates can be output as the inspection result of the internal state related to the semiconductor wafer 15 (test object). As a result, in addition to the above-mentioned expected effects, detailed defect candidate information regarding what types of defect candidates are present, where and how many of them are present can be provided. Furthermore, in the ultrasonic inspection method according to an embodiment of the present invention, a configuration may be adopted in which, in the first inspection process (step S11), first identification information for a first defect candidate is added to the position information and defect type information for the first defect candidate, in the second inspection process (step S12), second identification information for a second defect candidate is added to the position information and defect type information for the second defect candidate, and in the output process (step S13), the defect candidate information to which the first identification information and the second identification information have been added are output as inspection results for the internal state of the semiconductor wafer 15 (subject).

[0055] With this configuration, defect candidate information that integrates the position information and defect type information related to the first and second defect candidates and to which the first identification information and the second identification information are respectively added can be output as the inspection result of the internal state related to the semiconductor wafer 15 (test subject). As a result, in addition to the above-mentioned expected effects, detailed defect candidate information regarding what types of defect candidates are present, where they are located, and how many of them are present, can be provided together with the origins of the first and second defect candidates.

[0056] Furthermore, in the ultrasonic inspection method according to an embodiment of the present invention, a configuration may be adopted that further includes an estimation step (step S14) of estimating a defect occurrence factor corresponding to the defect candidate information by referring to a defect occurrence factor database 25 (see FIG. 1) that stores information on defect occurrence factors in a manufacturing process related to the semiconductor wafer 15 (subject) in association with each of a plurality of pieces of defect type information. With this configuration, the estimation process (step S14) can estimate a defect occurrence factor in the manufacturing process related to the semiconductor wafer 15 (test object) that corresponds to the defect candidate information. In addition to the desired effects described above, this can provide information on the defect occurrence factor in the manufacturing process related to the test object.

[0057] Furthermore, in the ultrasonic inspection method according to an embodiment of the present invention, a configuration may be adopted in which, in the first inspection process (step S11), a first inspection image is acquired (steps S111-S112) using a first ultrasonic probe 13a of an array type (electronic scanning type), while, in the second inspection process (step S12), a second inspection image is acquired using a second ultrasonic probe 13b of a single type (single scanning type). With this configuration, in the first inspection step (step S11), the area where the first defect candidates are distributed is narrowed down by a primary inspection (high-speed mode) using the first ultrasonic probe 13a of the array type (electronic scanning type) that inspects a wide area (first inspection area), and in the second inspection step (step S12), the location of the second defect candidates and their defect type are made apparent by a secondary inspection (high-sensitivity mode) using the second ultrasonic probe 13b of the single type (single scanning type) that inspects the narrowed down area (second inspection area). As a result, the expected effect of inspecting the internal state of the semiconductor wafer 15 (test object) with high sensitivity and high speed using ultrasonic waves can be fully obtained.

[0058] In addition, in the ultrasonic inspection method according to an embodiment of the present invention, the setting step (step S114) may be configured to extract an area in which the first defect candidates are irregularly distributed based on distribution information related to the first defect candidates acquired using the first ultrasonic probe 13a in the first inspection step (step S11), and set the extracted area as a second inspection area. With this configuration, in the setting process (step S114), an area in which defect candidates extracted based on distribution information on the first defect candidates acquired using the first ultrasonic probe 13a in the first inspection process (step S11) are irregularly distributed is set as the second inspection area, so that the inspection target area can be appropriately narrowed down in the secondary inspection (high sensitivity mode) in the second inspection process (step S12).

[0059] Therefore, it is possible to accurately identify the location of the second defect candidate and its defect type, and as a result, the desired effect of inspecting the internal state of the semiconductor wafer 15 (test object) with high sensitivity and high speed using ultrasound can be achieved at an even higher level.

[0060] [Overview of the semiconductor wafer 15 to be inspected] Next, an overview of the semiconductor wafer 15 to be inspected will be described with reference to Fig. 4. Fig. 4 is a perspective view of the manufacturing process of the semiconductor wafer 15 to be inspected, and the semiconductor wafer 15 having a multi-layer structure manufactured by the manufacturing process. The semiconductor wafer 15 having a multilayer structure, which is the subject of the test, is manufactured, for example, by using a manufacturing process 52 shown in Fig. 4. That is, as shown in Fig. 4, the semiconductor wafer 15 can be manufactured in a first case (see reference numeral 61) by sequentially forming circuit patterns and copper (Cu) electrodes on a plurality of silicon wafers 53 (see reference numerals 55 and 57) and bonding the copper (Cu) electrodes of the first wafer 55 and the second wafer 57 together, and in a second case (see reference numeral 63) by cutting the first wafer 55 into die units and then bonding the cut pieces of the first wafer 55 into die units to the second wafer 57.

[0061] In either manufacturing process, a regular circuit pattern is arranged on the bonding surface of a semiconductor wafer 15 (subject) in die units (see reference numerals 65 and 67). Information necessary for identifying defects in the manufacturing process includes the tendency of defects to occur on the entire surface of the wafer, the performance of each die (OK / NG, etc.), and the presence or absence of minute pattern defects. In the present invention, it is desirable to obtain such necessary information with high sensitivity and high speed. The ultrasonic inspection apparatus 11 according to an embodiment of the present invention inspects a semiconductor wafer 15 (subject) to check whether all dies are correctly bonded to a bonding interface where a plurality of dies 65, 67... (regularly arranged rectangles) are bonded to a disk-shaped silicon wafer 53 without generating any void areas such as voids or peeling, and to check for defects including pattern defects that occur during the circuit pattern formation process.

[0062] Here, the ultrasonic inspection method according to the embodiment of the present invention will be described more specifically with reference to Figs. 1, 2A, 2B, and 3 as appropriate. When ultrasonic inspection is started by the ultrasonic inspection device 11, in step S111 belonging to the first inspection process (step S11), the driving unit 51 drives the array-type (electronic scanning type) first ultrasonic probe 13a via the first scanner 41a (see Figures 2A and 2B) provided in the ultrasonic exploration unit 22.

[0063] In step S112, the first ultrasonic probe 13a included in the ultrasonic inspection unit 22 performs a primary inspection (high-speed scanning according to the high-speed mode) on a wide range (first inspection region: the entire wafer bonding surface) as an inspection target. In step S113, the image processing unit 24 acquires first defect candidate information based on the first inspection image (see reference numeral 15-1 in FIG. 3) relating to the first inspection area.

[0064] The numbers 1, 2, 3, ..., N in the first inspection image shown in Fig. 3 are identification numbers given to each of the multiple dies. The first defect extraction unit 28a belonging to the image processing unit 24 performs a first defect extraction process on the obtained first inspection image to extract first defect candidates. The image processing unit 24 assigns a label (defect type label) according to the defect type to each of the first defect candidates. In step S114, the image processing unit 24 sets a second inspection area according to the distribution state of the first defect candidates each assigned with a defect type label.

[0065] Next, in step S121 belonging to the second inspection process (step S12), the driving unit 51 drives the second ultrasonic probe 13b of the single type (single scan method) via the second scanner 41b (see Figures 2A and 2B) provided in the ultrasonic exploration unit 22. In step S122, the second ultrasonic probe 13b included in the ultrasonic exploration unit 22 performs a secondary inspection (low-speed scanning in accordance with the high sensitivity mode) on the narrow range (second inspection area) set in step S114 as the inspection target.

[0066] In step S123, the image processing unit 24 acquires high-definition second defect candidate information based on the second inspection image (see reference numeral 15-2 in FIG. 3) relating to the second inspection region. The dashed-line regions 1 and 2 in the second inspection image shown in Fig. 3 are second inspection regions from which a high-definition second inspection image can be obtained. The second inspection region may be a specific unit die region like dashed-line region 1, or may be a region spanning multiple dies like dashed-line region 2. The second defect extraction unit 28b belonging to the image processing unit 24 performs a second defect extraction process on the obtained second inspection image to extract second defect candidates. The image processing unit 24 assigns a label (defect type label) according to the defect type to each of the second defect candidates.

[0067] In step S13, the output unit 29 belonging to the image processing unit 24 integrates the first and second defect candidates, each of which has been assigned a defect type label, and outputs the integrated defect candidate information to the overall control unit 33 as the inspection result of the internal condition of the semiconductor wafer 15 (test object). In step S14, the image processing unit 24 (estimation unit 27 shown in FIG. 1) refers to the defect occurrence cause DB 25 (see FIG. 1) and estimates a defect occurrence cause in the manufacturing process that corresponds to the defect candidate information.

[0068] In an ultrasonic inspection method according to an embodiment of the present invention, in a first inspection process (step S11), scanning and inspection is performed at high speed on a first inspection area (the entire surface of the wafer bonding surface), and in a second inspection process (step S12), scanning and inspection is performed finely (with high sensitivity) on a limited narrow area (second inspection area). According to the ultrasonic inspection method of an embodiment of the present invention, in the manufacturing process of semiconductor wafers 15, the need for secondary inspection, which requires a long time to obtain inspection images, is determined based on the results of the primary inspection in the first inspection process (step S11) 1, thereby reducing the inspection time required for each semiconductor wafer 15 and enabling inspection of more mass-produced wafers.

[0069] Fig. 5A is an explanatory diagram showing a first inspection image of a bonding interface of a semiconductor wafer 15 as an inspected object. Fig. 5B is an explanatory diagram showing a second inspection image of a bonding interface of a semiconductor wafer 15 as an inspected object. Reference numeral 501 in Fig. 5A denotes a first inspection image of the wafer bonding interface acquired via the first ultrasonic probe 13a. An image in which a regular pattern (hereinafter referred to as a "shot") is formed, with a total of eight dies, four in the X direction and two in the Y direction, as shown in Fig. 5A, as one set, is defined as the first inspection image.

[0070] Reference numeral 502 in Fig. 5B denotes a second inspection image of a second inspection region of one shot indicated by a dashed line in reference numeral 501 (see Fig. 5A), acquired via the second ultrasonic probe 13b. A detailed pattern can be resolved within the die. The symbols 503 and 504 shown in the second inspection image 502 are examples of defects (large void, entire die defect) extracted by the first defect extraction unit 28a (see Figures 2A and 2B), and the circle in the symbol 505 is an example of a defect (micropattern defect) extracted by the second defect extraction unit 28b (see Figures 2A and 2B).

[0071] In the example shown in FIGS. 5A and 5B, the first and second defect extraction units 28a and 28b extract defects of different types and sizes. FIG. 6 is an explanatory diagram showing an overview of the first defect extraction process performed by the first defect extraction unit 28a. 5, the first defect extraction unit 28a first extracts images for each shot from a first inspection image 601 in which shots (including dies in 2 rows and 4 columns) are regularly arranged on the wafer bonding surface, and aligns a first inspection image group 602. Aligning the inspection images means performing position correction so that the same pattern in the extracted first inspection images 603 to 608 exists at the same coordinate value in each inspection image.

[0072] Next, the first defect extraction unit 28a collects brightness values related to common coordinates among the multiple shot images, and calculates statistical features related to the multiple shot images. Feature calculations 609 and 610 indicate a process of calculating statistical features based on brightness values related to mutually different coordinates. This image processing is performed for all pixels in the multiple shot images, and a reference image 611 and a dynamic threshold value 612 (a threshold value that changes for each coordinate according to the statistical features between the shot images) are generated.

[0073] Here, examples of the statistical feature include the average of the collected brightness values, the most frequent brightness value, and the standard deviation. Next, the first defect extraction unit 28a uses the calculated feature to generate a reference image 611. In an example in which the most frequent brightness value is used as the statistical feature, the statistical feature amount μ_brightness at each coordinate position (x, y) in each shot image after alignment is calculated using the following (Equation 1), and the calculated value is set as the reference image G(x, y).

number

[0074] In formula 1, d1(x,y) through dN(x,y) are the brightness values at the coordinates (x,y) in each shot image, and maxCounter is a filter that finds the most frequent brightness value from the brightness values d1(x,y) through dN(x,y). The reference image 611 is generated based on the most frequent brightness value at the coordinates (x,y) in each shot image. Similarly, the dynamic statistical threshold 612 is generated based on the standard deviation of d1(x,y) through dN(x,y).

[0075] Next, for each of the multiple shot images 603 to 608, an integrated comparison is performed with the generated reference image 611 and a dynamic threshold value 612 to extract first defect candidates.

[0076] The first defect extraction unit 28a assumes that multiple identical images exist in the first inspection image, and estimates a statistically defect-free reference image and noise intensity based on the statistical characteristics between the first inspection images, and extracts defect candidates by separating them from noise. The first defect extraction unit 28a assigns a defect type label to the first defect candidates extracted in this way according to the defect type, using the shape and texture information in the defect candidates. The image processing unit 24 extracts an area where the first defect candidates are irregularly distributed, based on the distribution information and defect type information regarding the first defect candidates, and sets the extracted area as a second inspection area.

[0077] FIG. 7 shows an example of types of defects (defect type labels assigned) extracted by the first defect extraction process associated with the first defect extraction unit 28a. In the hybrid bonding process, the die is attached regularly on the silicon wafer 701 as a rule. The die 702 is a normal die with no gap between it and the silicon wafer 701. The dies 703 to 705 are dies including bonding defects in which minute cracks (gaps) occur between the silicon wafer 701 and the die, or huge voids (gaps) that span between the dies. In addition to these bonding defects that occur randomly in the wafer (hereinafter referred to as random defects), there are bonding defects (hereinafter referred to as die defects) that occur on a die-by-die basis that are specific to hybrid bonding wafers, such as "no die" 706 where the die is not at the position where it should be bonded, "floating die" 707 where the die is floating, "die bending" 708, and "incompletely bonded die" 709 where the die is bonded at an angle.

[0078] The first defect extraction unit 28a extracts bonding defect candidates from both these random defects and die defects occurring on a die-by-die basis, and classifies them according to their defect types (assigns defect type labels). Next, the image processing unit 24 sets the area in which a die (such as 703 in the figure) having random and relatively small defects exists as a second inspection area, and passes the setting information 111 (see FIG. 3) to the second inspection process (step S12).

[0079] In FIG. 7, reference numeral 710 denotes a defect map extracted and classified by the first defect extraction unit 28 a , and reference numeral 720 denotes an example of a second inspection area set based on the defect map 710 . Note that the second inspection area may be set in multiple locations. Also, an area where no defects are extracted may be set as a normal area, and the normal area may be excluded from the second inspection area.

[0080] FIG. 8 is an explanatory diagram showing an overview of the second defect extraction process performed by the second defect extraction unit 28b. Reference numeral 801 denotes a second inspection image relating to a second inspection area. In step S801, the second defect extraction unit 28b sets a plurality of small regions (hereinafter referred to as queries). In the second inspection image 801 relating to the second inspection region shown in Fig. 8, a vertically shaded region represents a query. The second defect extraction unit 28b appropriately sets a plurality of queries in consideration of covering the second inspection region.

[0081] In step S802, the second defect extraction unit 28b sets a reference query of the same size in the second inspection area (horizontal stripe hatched areas 803a, 803b, etc.) for a query of interest (802) set at coordinates of interest (x, y) in the second inspection image 801 related to the second inspection area, and searches for a reference query having a feature amount in the query similar to that of the query of interest while scanning the query. Representative examples of the feature amount in the query are normalized cross-correlation with the query of interest and a sum of brightness differences in the query. However, a feature amount representing brightness or texture information in the query may be adopted as the feature amount in the query. The second defect extraction unit 28b stores N reference query positions having similar features (horizontal stripe query, diagonal stripe query, ...).

[0082] In step S802, the second defect extraction unit 28b sequentially replaces the query of interest with the reference queries, such as arranging the query with the closest feature (horizontal stripe query) at the position of the query of interest, arranging the query with the second closest feature (diagonal stripe query), and so on. In step S803, the second defect extraction unit 28b generates N reference images. In steps S804-S805, the second defect extraction unit 28b calculates the differences between each of the N reference images 804, 805, ... generated in step S803 and the second inspection image 801, determines a representative value of the statistical differences based on the N calculated differences, and extracts locations where the representative value is equal to or greater than a threshold value as defect candidates.

[0083] The concept of the second defect extraction process related to the second defect extraction unit 28b is to assume that the second inspection image has high resolution and is a narrow area, and there is no guarantee that a periodic pattern exists in the second inspection image, but that a similar pattern exists in a very local area (i.e., within the query), generate multiple reference images, and extract what are statistically likely to be defects (i.e., defect candidates). The image processing unit 24 assigns a defect type label to each of the second defect candidates extracted as described above.

[0084] Reference numeral 807 denotes a representative image of the difference at each query position calculated in step S805. Reference numeral 808 (encircled) denotes an example of a second defect candidate extracted within the second inspection area in a threshold determination process set based on the representative image of the difference. For example, a minute crack or a minute pattern defect that occurs depending on a circuit pattern is extracted as the second defect candidate.

[0085] FIG. 9 shows an example of a defect map 901 that is output by integrating the first and second defect candidates that are extracted in the first inspection process (see step S11 in FIG. 3) and the second inspection process (see step S12 in FIG. 3), respectively, and to which defect type labels have been assigned. According to the defect map 901, it is possible to recognize at a glance distribution information of defect candidates, such as on which die various die bonding defects occur and where pattern defects and pattern formation defects are concentrated (distributed).

[0086] The image processing unit 24 shown in FIGS. 2A and 2B (the estimation unit 27 shown in FIG. 1) estimates a defect occurrence factor in the manufacturing process based on the integrated defect candidate information related to the defect map 901, and presents the estimated defect occurrence factor (see step S14 shown in FIG. 3). Examples of causal relationships between defect occurrence factors in manufacturing processes corresponding to defect candidates include the following: when there are many defective labels per die, it may be suggested to review the hybrid bonding process or annealing conditions; when there are many microscopic pattern formation defects, it may be suggested to review the pattern formation process such as etching conditions; and when relatively large defects have occurred in a circular pattern around the edge of the wafer, it may be suggested to review the CMP process (planarization process).

[0087] It should be noted that the present invention is not limited to the above, and a configuration may be adopted in which more detailed modifications of process conditions are presented. Other embodiments The above-described embodiments are merely examples of the realization of the present invention. Therefore, the technical scope of the present invention should not be interpreted as being limited by these embodiments. The present invention can be embodied in various forms without departing from the spirit or main characteristics of the present invention.

[0088] In addition, a part of the configuration of the embodiment described here can be replaced with the configuration of another embodiment, and a configuration of one embodiment can be added to a configuration of another embodiment. Furthermore, a part of the configuration of each embodiment can be added to, deleted from, or replaced with another configuration.

[0089] In the ultrasonic inspection apparatus 11 according to an embodiment of the present invention, the second defect extraction unit 28b belonging to the image processing unit 24 assumes that the second inspection image is a high-resolution, narrow area, and there is no guarantee that a periodic pattern exists in the second inspection image, but assumes that a similar pattern exists in a very local area (i.e., within the query), and adopts a configuration in which multiple reference images are generated and what are statistically likely to be defects are extracted.

[0090] However, if the distribution information of the first defect candidates can be accurately grasped in advance when the second defect candidates are extracted by the second defect extraction unit 28b belonging to the image processing unit 24, it becomes possible to appropriately set an image processing procedure for extracting the second defect candidates based on the accurate distribution information of the first defect candidates. As a result, the accuracy of extracting the second defect candidates can be significantly improved. Therefore, in the ultrasonic inspection device according to the other embodiment, contrary to the ultrasonic inspection device 11 according to the embodiment of the present invention, first, an inspection image of the entire wafer bonding surface is obtained using the single type second ultrasonic probe 13b. However, it takes a huge amount of time to obtain an inspection image of the entire wafer by the single type second ultrasonic probe 13b. Therefore, this process is not performed for the entire wafer to be inspected, but is performed once when the inspection conditions for the wafer manufactured under the common process conditions are determined. By this process, a high-resolution image is obtained over a huge amount of time, and defect candidates are extracted, so that the location and amount of minute defects in the first inspection area can be grasped in advance.

[0091] Next, ultrasonic inspection is performed using the array-type first ultrasonic probe 13a. At this time, a recipe such as image acquisition conditions for the array-type first ultrasonic probe 13a, parameters of the defect extraction process, etc. are optimized so that the distribution of the obtained second defect candidates approaches the inspection results obtained in the first inspection area. Although an example has been described in which ultrasonic inspection apparatus 11 according to an embodiment of the present invention performs ultrasonic defect inspection on the bonding surface of laminated wafers regularly formed on semiconductor wafer 15, the present invention is not limited to this example. Ultrasonic inspection apparatus 11 according to an embodiment of the present invention may also be applied to tray inspection in which IC trays on which IC circuits are regularly arranged are used as test objects. [Explanation of symbols]

[0092] 11 Ultrasonic inspection equipment 13a First ultrasonic probe 13b Second ultrasonic probe 15 Semiconductor wafer (specimen) 17 Setting section 18 A / D conversion section 19 1st Inspection Department 21 Second Inspection Department 22 Ultrasonic Exploration Unit 23 Defect type database (Defect type DB) 24 Image Processing Section 25 Defect Cause Database (Defect Cause DB) 26 Image Generation Unit 27 Estimation part 28a First defect extraction section 28b Second defect extraction section 29 Output section 33 Overall Control Unit 51 Drive unit

Claims

1. An ultrasonic inspection apparatus for inspecting an internal state of a test object including a semiconductor wafer stacked in multiple layers by using ultrasonic waves, comprising: a setting unit that sets a first inspection area and a second inspection area as an inspection target range related to the specimen; a first inspection unit that acquires a first inspection image in the first inspection area using a first resolution and acquires first defect candidate information based on the first inspection image; a second inspection unit that acquires a second inspection image in the second inspection area using a second resolution and acquires second defect candidate information based on the second inspection image; an output unit that outputs an inspection result of an internal condition of the subject, the second inspection area is set within a range of the first inspection area and is smaller than the first inspection area; the second resolution is set to a value higher than the first resolution, The output unit integrates first defect candidate information related to the first inspection area acquired by the first inspection unit and second defect candidate information acquired by the second inspection unit, and outputs the integrated information as an inspection result of an internal state related to the specimen.