JP2019160999A - Defect inspection device, and defect inspection method - Google Patents

Abstract

To provide a defect inspection device capable of detecting crystal defect of a silicon carbide specimen, especially basal surface displacement becoming problem in manufacturing of power device, in a short time with high efficiency.SOLUTION: A defect inspection device 10 for detecting crystal defect of a silicon carbide specimen A includes irradiation means 11 for irradiating the inspection region of the silicon carbide specimen A with exciting light L1, photography means 12 for photographing PL emission images in the inspection region collectively via a color filter 12a, conversion means 13 for converting the RGB value in the RGB color space of the PL emission images into CIE 1931 colorimetric system, and obtaining the x value in the xy color chromaticity coordinate, and determination means 14 for determining crystal defect of the silicon carbide specimen A.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a defect inspection apparatus and a defect inspection method for detecting crystal defects in a silicon carbide sample.

  Silicon carbide is a wide-gap semiconductor and is a material that has attracted attention in recent years in the field of power devices. Normally, silicon carbide produces a thin film substrate by an epitaxial growth method. However, since there is a large difference in atomic radii between carbon atoms and silicon atoms, there are crystal defects in the epitaxial layer compared to gallium nitride, which is the same wide gap semiconductor. Is likely to occur. Among the crystal defects, the basal plane dislocation causes a forward conduction deterioration and a leak point, and thus becomes a problem in the production of a power device. Therefore, an inspection method using photoluminescence (hereinafter referred to as “PL”) has been developed as a method for detecting a crystal defect in a silicon carbide substrate.

  For example, there is an inspection method for detecting crystal defects by scanning a silicon carbide substrate and mapping PL light (see, for example, Patent Document 1). In the inspection method of Patent Document 1, it is said that the two-dimensional distribution of crystal defects can be examined with high spatial resolution.

  In addition, there is an inspection method for irradiating a silicon carbide substrate with excitation light, separating reflected light from the silicon carbide substrate and PL light, and photographing a defect image by the reflected light and a defect image by the PL light (for example, , See Patent Document 2). In the inspection method of Patent Document 2, basal plane dislocations can be classified from other crystal defects.

JP 2006-147848 A JP 2015-1119056 A

  However, the inspection methods of Patent Documents 1 and 2 have a problem that the inspection time is very long because it is necessary to repeat irradiation of excitation light and detection of PL light many times in order to scan the entire surface of the substrate. Silicon carbide substrates have been increased in diameter due to recent improvements in manufacturing technology, and when the inspection methods of Patent Documents 1 and 2 are applied to large-diameter silicon carbide substrates, the problem of inspection time becomes even more prominent. . In addition, silicon carbide substrates have made significant progress in terms of quality improvement, and in particular, basal plane dislocations can be suppressed to several in the substrate for both 4-inch and 6-inch wafers. Is possible. Inspection of Patent Documents 1 and 2 for identifying a high-quality silicon carbide substrate having only a few basal plane dislocations in the substrate and identifying individual crystal defects with high resolution on the assumption that many crystal defects exist Inspection by the method is inefficient in terms of both inspection time and equipment cost. As described above, the conventional inspection technique cannot sufficiently cope with the increase in the diameter and quality of the silicon carbide substrate, and development of a new inspection method is required.

  The present invention has been made in view of the above problems, and is a defect inspection capable of detecting a crystal defect of a silicon carbide sample, particularly a basal plane dislocation which is a problem in the manufacture of a power device, in a short time with high efficiency. An object is to provide an apparatus and a defect inspection method.

The characteristic configuration of the defect inspection apparatus according to the present invention for solving the above problems is as follows.
A defect inspection apparatus for detecting crystal defects in a silicon carbide sample,
Irradiating means for irradiating the inspection region of the silicon carbide sample with excitation light;
Imaging means for collectively capturing PL emission images in the inspection area via a color filter;
Conversion means for converting an RGB value in the RGB color space of the PL emission image into a CIE 1931 color system to obtain an x value in an xy chromaticity coordinate;
Determination means for determining the presence or absence of crystal defects in the silicon carbide sample,
The determination means refers to a first threshold value larger than the x value of the white point in the xy chromaticity coordinates and a second threshold value smaller than the x value of the white point, and the x value obtained by the conversion of the RGB value is the first value. It is determined that there is a basal plane dislocation at the position of the silicon carbide sample corresponding to a pixel exceeding one threshold, and the silicon carbide sample corresponding to a pixel whose x value obtained by conversion of the RGB value is less than the second threshold It is determined that there is a stacking fault at the position.

  In a silicon carbide single crystal substrate having a thickness of several tens of μm used for manufacturing power devices, crystal defects such as basal plane dislocations have a length of about several hundred μm along the off-cut direction of the crystal. When a silicon carbide sample is a 4-inch wafer of silicon single crystal, the size of crystal defects viewed from the normal direction of the substrate surface is about 1/1000 of the diameter of the silicon carbide sample. Therefore, when the entire surface of a silicon carbide sample is photographed in a lump using a popular product class image sensor having a number of pixels of several million to several tens of millions, crystal defects are photographed in a size corresponding to a plurality of pixels. . Even when a 6-inch wafer is used as a silicon carbide sample, it is possible to easily capture an image in which crystal defects have a size corresponding to a plurality of pixels by collectively photographing an inspection area of about ¼ of the entire surface of the silicon carbide sample. Obtainable. The present inventor has taken a PL emission image collectively under various conditions on the entire surface of such a silicon carbide sample or a considerably wide area obtained by dividing the entire surface by a fraction, and positions where crystal defects exist. The characteristics of the pixel corresponding to are examined in detail. As a result, when the PL emission image is photographed in color through the color filter, the present inventor converts the RGB value in the RGB color space to the xy color of the CIE 1931 color system at the pixel corresponding to the position where the crystal defect exists. It has been found that the x value obtained by converting into degree coordinates shows a characteristic value depending on the type of crystal defect. According to the defect inspection apparatus of this configuration, since the basal plane dislocation and the stacking fault can be distinguished based on the x value obtained by converting the RGB value, the PL emission image in which the crystal defect has a size corresponding to a plurality of pixels. Thus, crystal defects, particularly basal plane dislocations that are problematic in the production of power devices, can be easily identified and detected as stacking faults of the same size. As a result, even if several crystal defects are scattered in a wide inspection region by photographing a PL emission image once without requiring mapping of PL light by scanning a silicon carbide sample, the crystal defects are eliminated. Detection can be efficiently performed in a short time, and the inspection time can be greatly shortened.

In the defect inspection apparatus according to the present invention,
Preferably, the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28.

  According to the defect inspection apparatus of this configuration, the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28. Even if the size of the defect is about the same, both can be clearly identified based on the x value, and the basal plane dislocation causing a problem in the production of the power device can be detected with high accuracy.

In the defect inspection apparatus according to the present invention,
The imaging means further captures a scattered light image in the inspection area,
Preferably, the determination means determines the presence or absence of a micropipe based on the scattered light image.

  According to the defect inspection apparatus of this configuration, by determining the presence or absence of the micropipe based on the scattered light image, the basal plane dislocation and stacking fault for determining the presence or absence based on the PL emission image and the micropipe are clarified. Can be identified.

In the defect inspection apparatus according to the present invention,
The irradiating means preferably includes an ultraviolet LED and a homogenizing means for homogenizing the ultraviolet light emitted from the ultraviolet LED.

  According to the defect inspection apparatus of this configuration, since the ultraviolet light emitted from the ultraviolet LED is made uniform, the inspection region of the silicon carbide sample can be excited uniformly and the crystal defect can be detected with high accuracy.

In the defect inspection apparatus according to the present invention,
The imaging unit is configured to capture a correction PL emission image by changing the imaging range of the PL emission image in a state where the inspection area is included,
It is preferable to further include a correcting unit that corrects the PL light emission image based on the correction PL light emission image.

  According to the defect inspection apparatus of this configuration, when the PL emission image includes noise specific to the imaging means not derived from the PL light, the correction coordinate emission light image captured by changing the imaging range also includes noise at the same coordinates. Therefore, it is possible to easily detect and correct noise specific to the photographing means from the PL emission image. As a result, it is possible to detect the crystal defect with high accuracy while suppressing the influence of noise inherent to the imaging means that is not derived from the PL light.

In the defect inspection apparatus according to the present invention,
The silicon carbide sample is preferably an epitaxial thin film of silicon carbide single crystal.

  According to the defect inspection apparatus of this configuration, it is possible to detect basal plane dislocations and stacking dislocations that are common in epitaxial thin films.

In the defect inspection apparatus according to the present invention,
The silicon carbide sample is preferably a 4-inch wafer of silicon carbide single crystal.

  If the silicon carbide sample is a 4-inch wafer of silicon carbide single crystal, when the entire surface of the silicon carbide sample is photographed collectively as the inspection region, crystal defects in the PL emission image have a size corresponding to a plurality of pixels. As a result, according to the defect inspection apparatus of this configuration, it is possible to detect the crystal defects contained in the silicon carbide sample in a short time with high efficiency only by photographing the entire surface of the silicon carbide sample once as an inspection region. it can.

The characteristic configuration of the defect inspection method according to the present invention for solving the above problems is as follows.
A defect inspection method for detecting crystal defects in a silicon carbide sample,
An irradiation step of irradiating the inspection region of the silicon carbide sample with excitation light;
An imaging process for collectively capturing PL emission images in the inspection area via a color filter;
A conversion step of converting an RGB value in the RGB color space of the PL emission image into a CIE 1931 color system to obtain an x value in an xy chromaticity coordinate;
A determination step of determining the presence or absence of crystal defects in the silicon carbide sample,
In the determination step, the first threshold value larger than the x value of the white point in the xy chromaticity coordinates and the second threshold value smaller than the x value of the white point are referred to, and the x value obtained by the conversion of the RGB values is the first value. It is determined that there is a basal plane dislocation at the position of the silicon carbide sample corresponding to a pixel exceeding one threshold, and the silicon carbide sample corresponding to a pixel whose x value obtained by conversion of the RGB value is less than the second threshold It is determined that there is a stacking fault at the position.

  According to the defect inspection method of this configuration, basal plane dislocations and stacking faults can be discriminated based on the x value obtained by converting the RGB values, so that the PL emission image in which the crystal defects have a size corresponding to a plurality of pixels. Thus, crystal defects, particularly basal plane dislocations that are problematic in the production of power devices, can be easily identified and detected as stacking faults of the same size. As a result, even if several crystal defects are scattered in a wide inspection region by photographing a PL emission image once without requiring mapping of PL light by scanning a silicon carbide sample, the crystal defects are eliminated. Detection can be efficiently performed in a short time, and the inspection time can be greatly shortened.

In the defect inspection method according to the present invention,
Preferably, the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28.

  According to the defect inspection method of this configuration, the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28. Even if the size of the defect is about the same, it can be clearly identified based on the x value, and the basal plane dislocation that causes a problem in the production of the power device can be detected with high accuracy.

FIG. 1 is a schematic configuration diagram of a defect inspection apparatus according to the present invention. FIG. 2 is an explanatory diagram of correction processing by the correction means. FIG. 3 is a scattered light image of the region where the micropipes are formed. FIG. 4 is a flowchart of the defect inspection method of the present invention.

  Hereinafter, embodiments relating to a defect inspection apparatus and a defect inspection method of the present invention will be described in detail with reference to the drawings. However, the present invention is not intended to be limited to the configuration described below.

[Defect inspection equipment]
FIG. 1 is an explanatory diagram of a defect inspection apparatus 10 according to the present invention. The defect inspection apparatus 10 is an apparatus for detecting basal plane dislocations and stacking faults that are crystal defects by inspecting the silicon carbide sample A using PL light L2 generated when the semiconductor is irradiated with excitation light L1. It is. In this embodiment, a silicon carbide sample A, which is a silicon carbide single crystal thin film substrate (4-inch wafer) manufactured by an epitaxial growth method, is an inspection object. The defect inspection apparatus 10 includes an irradiation unit 11, an imaging unit 12, a conversion unit 13, and a determination unit 14, and further includes a correction unit 15 and a visible light irradiation unit 16 as optional configurations.

  Irradiation means 11 irradiates the inspection region of silicon carbide sample A with excitation light L 1 for causing silicon carbide sample A to emit PL light L 2. Irradiation means 11 is disposed at a position where the inspection region of silicon carbide sample A can be irradiated with excitation light L1 all at once. In the present embodiment, the entire surface of the silicon carbide sample A is used as the inspection region, and the irradiation means 11 is arranged at a position where the excitation light L1 can be irradiated in a range including the entire silicon carbide sample A as shown in FIG. The The irradiation unit 11 preferably includes a light source 11a and a uniformizing unit 11b. The light source 11a includes, for example, an ultraviolet LED, and emits light in the ultraviolet region having a wavelength of 350 to 400 nm. The uniformizing unit 11b includes, for example, a rod integrator lens and the like, and uniformizes the illuminance distribution of the emitted light from the ultraviolet LED.

  The photographing unit 12 includes a color filter 12a of three primary colors of red (R), blue (B), and green (G), and a solid-state imaging device such as a CCD (Charged-coupled devices) and a CMOS (Complementary metal-oxide-semiconductor). This is a digital camera having an image sensor 12b arranged in a two-dimensional array and detecting light incident on the photographing means 12 through the color filter 12a and detected by the image sensor 12b. The photographing unit 12 captures light through the color filter 12a, thereby generating an image having RGB values indicating R, G, and B gradation levels in the RGB color space for each pixel. Imaging means 12 is arranged at a position where the inspection region of silicon carbide sample A can be imaged collectively. The photographing means 12 has two photographing modes, a PL light photographing mode and a scattered light photographing mode. In the PL light imaging mode, after irradiating the excitation light L1 from the irradiation means 11, the silicon carbide sample A in a state of emitting the PL light L2 is imaged. Hereinafter, an image photographed in the PL light photographing mode is referred to as a “PL light emission image”. The PL light L2 in the near-infrared light region having a wavelength of 750 nm is emitted from the position where the basal plane dislocation of the silicon carbide sample A exists, and in a solid-state imaging device such as a CCD, the PL light L2 is red light (wavelength). (700 nm or more). As a result, in the PL emission image photographed by the photographing means 12, a bright spot having an R value is generated at a position where the basal plane dislocation exists. In the scattered light imaging mode, the scattered light of the visible light L3 on the surface of the silicon carbide sample A is photographed in a state where the visible light L3 is irradiated on the silicon carbide sample A by the visible light irradiation means 16 using a white LED or the like as a light source. To do. Hereinafter, an image photographed in the scattered light photographing mode is referred to as a “scattered light image”. The photographing unit 12 records the photographed image in a storage such as a hard disk.

  The image sensor 12b preferably has 1 million or more pixels. Crystal defects such as basal plane dislocations have a size of about several hundreds μm in length along the off-cut direction of the crystal. If the number of pixels of the image sensor 12b is 1 million or more, the entire surface of a 4-inch wafer is covered. When the images are taken together, crystal defects can be taken with a size corresponding to a plurality of pixels. Therefore, if the silicon carbide sample A is a 4-inch wafer, the entire surface of the silicon carbide sample A can be set as an inspection region, and the presence or absence of crystal defects can be inspected by one imaging. When the number of pixels is less than 1 million, when the entire surface of a 4-inch wafer is imaged collectively, the imaging size of crystal defects becomes small, and it may be difficult to accurately determine the presence or absence of crystal defects.

  The conversion means 13, the determination means 14, and the correction means 15 are realized by a computer having a CPU, a memory, a storage, and the like when the CPU reads and executes a program recorded in the memory. The conversion unit 13 converts the RGB value of the PL emission image captured by the imaging unit 12 and recorded in the storage into the CIE 1931 color system, and generates an xy chromaticity value in the xy chromaticity coordinates. The generation of the xy chromaticity value by the conversion unit 13 is performed for each pixel of the PL emission image or for each of a plurality of pixels. When the generation of the xy chromaticity value is performed for each of a plurality of pixels, for example, a group of pixels having similar RGB values is detected by edge detection processing, and the average value of the RGB values of the group of pixels is used to determine the xy chromaticity. It is preferable to generate a value. The generation of the xy chromaticity value is performed by converting the RGB value into the tristimulus values X, Y, and Z, and then converting them into the xy chromaticity coordinates. For example, when the sRGB color system defined by the International Electrotechnical Commission is used as the RGB color space of the image data, first, R, G, and B in the sRGB color system are expressed by the following equations (1) to (3). , Converted into R ′, G ′, and B ′ in a color space proportional to the original brightness of the PL light (hereinafter referred to as “linear RGB color space”).

Next, R ′, G ′, and B ′ in the linear RGB color space are converted into tristimulus values X, Y, and Z in the CIE 1931 color system by the following equations (4) to (6).
X = 0.4124 × R ′ + 0.3576 × G ′ + 0.1805 × B ′ (4)
Y = 0.2126 × R ′ + 0.7152 × G ′ + 0.0722 × B ′ (5)
Z = 0.0193 x R '+ 0.1192 x G' + 0.9505 x B '(6)

Furthermore, an xy chromaticity value is obtained from the tristimulus values X, Y, and Z by the following equations (7) and (8).
x = X / (X + Y + Z) (7)
y = Y / (X + Y + Z) (8)

  Determination means 14 refers to the xy chromaticity value generated by conversion means 13 to determine the presence or absence of crystal defects in silicon carbide sample A. In the determination by the determination means 14, specifically, referring to the first threshold value that is larger than the x value of the white point, the carbonization corresponding to the pixel in which the x value of the xy chromaticity value exceeds the first threshold value in the PL emission image. It is determined that a basal plane dislocation exists at the position of the silicon sample A. For example, when the CIE standard light source D65 is used, the white value in the xy chromaticity coordinates has an x value of 0.3127. The first threshold value is preferably set to a value existing between 0.35 and 0.50, and more preferably set to 0.40. By setting the first threshold value to the above value, the basal plane dislocation in the silicon carbide sample A can be detected with high accuracy. When the first threshold value is set to a value less than 0.35, there is a possibility that the basal plane dislocation is erroneously detected at a position where no crystal defect occurs and at a position where another type of crystal defect occurs. If the first threshold value is set to a value exceeding 0.50, there is a possibility that detection failure of basal plane dislocation occurs. In the determination by the determination unit 14, the silicon carbide corresponding to the pixel whose x value of the xy chromaticity value is less than the second threshold value in the PL emission image with reference to the second threshold value smaller than the x value of the white point. It is determined that there is a stacking fault at the position of the sample A. The second threshold is preferably set to a value existing between 0.20 and 0.28, more preferably 0.25. By setting the second threshold value to the above value, the stacking fault in the silicon carbide sample A can be detected with high accuracy. If the second threshold value is set to a value less than 0.20, a stacking fault may be missed. If the second threshold is set to a value exceeding 0.28, stacking faults may be erroneously detected at positions where no crystal defects have occurred and positions where other types of crystal defects have occurred.

  The determination unit 14 further determines the presence / absence of a micropipe based on the scattered light image captured by the imaging unit 12 and recorded in the storage. A micropipe is a kind of crystal defect and is a fine through-hole formed in an epitaxial thin film substrate of a silicon carbide single crystal. FIG. 3 is a scattered light image of the region where the micropipes are formed. As a process for determining the presence or absence of a micropipe, the determination means 14 detects a pixel at the center M of a structure in which six lines having lower brightness than the surroundings spread radially as shown in FIG. 3 by analyzing the scattered light image. Then, it is determined that the micropipe exists at the position of the silicon carbide sample corresponding to the pixel at the center M.

  The correction unit 15 performs correction processing on the PL emission image and the scattered light image which are captured by the imaging unit 12 and recorded in the storage, and noise caused by uneven sensitivity of the image sensor 12b and the solid-state imaging device are not cooled. Removes or reduces noise generated when used in The correction unit 15 corrects the PL emission image and the scattered light image in the storage, and the conversion unit 13 and the determination unit 14 use the corrected image for processing, thereby causing noise caused by unevenness of sensitivity of the image sensor 12b. The presence or absence of crystal defects can be determined with high accuracy. FIG. 2 is an explanatory diagram of correction processing by the correction means 15. FIG. 2A is a diagram illustrating a PL emission image shooting range B1 and a correction PL emission image shooting range B2. FIG. 2B is a schematic diagram of the PL emission image C1 captured in the imaging range B1. FIG. 2C is a schematic diagram of the correction PL emission image C2 captured in the imaging range B2. Before the correction unit 15 performs the correction process, the imaging unit 12 first includes a state in which the inspection region of the silicon carbide sample A is included in the imaging range B1 in which the PL emission image is captured as illustrated in FIG. The shooting range B2 is changed as it is, and a correction PL emission image is shot in the shooting range B2. The change of the photographing range is preferably a parallel movement from the photographing range B1 to the photographing range B2. For example, the mounting table on which the silicon carbide sample A is mounted is moved in parallel with the surface of the silicon carbide sample A. As another method, the photographing means 12 may be moved in parallel with the surface of the silicon carbide sample A. When a camera installed above the silicon carbide sample A is used as the imaging means 12, 50 pixels are respectively arranged in the vertical direction and the horizontal direction in which the solid-state imaging devices are arranged in a two-dimensional array in the image sensor 12b after the PL emission image is captured. The photographing range can be moved in parallel by moving the mounting table by a width corresponding to and photographing the correction PL light emission image. The correction means 15 compares the PL emission image C1 of FIG. 2B taken in the shooting range B1 with the correction PL emission image C2 of FIG. 2C taken in the shooting range B2, and PL emission is performed. The bright spot existing at the same coordinates in the image C1 and the correction PL emission image C2 is determined as noise, and the PL emission image C1 is corrected. In the example of FIG. 2, the bright spot D1 in the PL light emission image C1 has a bright spot D2 having the same luminance at the same coordinates of the correction PL light emission image C2, and therefore the correction means 15 determines that the bright spot D1 is noise. judge. The correcting means 15 corrects the bright spot D1 determined as noise by replacing the pixel F of the correction PL emission image C2 obtained by photographing the position of the silicon carbide sample A corresponding to the bright spot D1. On the other hand, the bright spot E1 in the PL light emission image C1 has no bright spot at the same coordinates of the correction PL light emission image C2, and the bright spot E2 is at the coordinates corresponding to the same position of the silicon carbide sample A in the correction PL light emission image C2. Therefore, the correcting unit 15 determines that the bright spot E1 is not a noise but a bright spot derived from the PL light L2. The correction unit 15 may further perform image processing such as shading correction on the PL emission image and the scattered light image in addition to the above correction processing.

  With the above configuration, the defect inspection apparatus 10 can detect basal plane dislocations and stacking faults that are scattered in several large inspection areas by capturing a single PL emission image. Moreover, the micropipe which exists in the test | inspection area | region of the silicon carbide sample A is detectable by imaging | photography of a once scattered light image. As a result, basal plane dislocations in which about several dots are scattered in a wide inspection region can be distinguished from other crystal defects and detected with high accuracy in a short inspection time.

[Defect inspection method]
FIG. 4 is a flowchart of the defect inspection method of the present invention. The defect inspection method is performed by executing an irradiation step (S1), a photographing step (S2), a conversion step (S3), and a determination step (S4 to S9) in the defect inspection apparatus 10 of FIG. The details of each step shown in the flowchart overlap with the description of the defect inspection apparatus 10, and thus the description thereof will be omitted. Here, description will be given with an emphasis on the procedure shown in the flowchart.

  In the defect inspection method, first, an irradiation step (S1) is performed. In the irradiation step (S1), the luminance distribution of the ultraviolet light having a wavelength of 350 to 400 nm emitted from the light source 11a using the ultraviolet LED is homogenized by the homogenizing means 11b using the rod integrator lens, and the silicon carbide sample is used as the excitation light L1. The entire inspection area A is irradiated. By performing the irradiation step (S1), silicon carbide sample A emits PL light L2. The size of the inspection region of the silicon carbide sample A corresponds to a plurality of pixels with a crystal defect whose length along the off-cut direction of the crystal is about several hundred μm depending on the number of pixels of the image sensor 12 b used in the imaging unit 12. It is preferable to set it to a size that can be photographed at a size to be captured.

  The imaging step (S2) is performed in a state where the silicon carbide sample A emits the PL light L2. In the imaging step (S2), the imaging means 12 generates a PL emission image by collectively imaging the inspection region of the silicon carbide sample A via the color filter 12a.

  Next, the conversion means 13 performs a conversion process (S3). In the conversion step (S3), the RGB values in the RGB color space of the PL emission image captured in the imaging step (S2) are converted into xy chromaticity coordinates to obtain xy chromaticity values for each pixel.

  Next, the determination means 14 performs a determination process (S4 to S9). In the determination step, first, an arbitrary pixel of the PL emission image is selected as a determination position (S4), and it is determined whether or not the x value of the xy chromaticity value at the determination position is larger than the first threshold (S5). When the x value is larger than the first threshold value (S5: Yes), the determination unit 14 detects the basal plane dislocation at the position of the silicon carbide sample A corresponding to the pixel at the determination position (S6), and thereafter, the process proceeds to S9. To proceed. When the x value is equal to or smaller than the first threshold value (S5: No), it is determined whether the x value of the xy chromaticity value at the determination position is smaller than the second threshold value (S7). When the x value is smaller than the second threshold (S7: Yes), the determination unit 14 detects a stacking fault at the position of the silicon carbide sample A corresponding to the pixel at the determination position (S8), and then performs the process at S9. Proceed. If the x value is greater than or equal to the second threshold (S7: No), the process proceeds directly to S9.

  In S9, the determination unit 14 determines whether there is an unprocessed pixel that is not selected at the determination position in the PL emission image. When there is an unprocessed pixel (S9: Yes), the process returns to S4, selects the unprocessed pixel as a determination position, and continues the determination process of S5 to S9. When there is no unprocessed pixel (S9: No), the inspection of the silicon carbide sample A is finished.

[Another embodiment]
The defect inspection apparatus and the defect inspection method of the present invention can also change the configuration described in the above embodiment. Such a modification will be described as another embodiment.

<Another embodiment 1>
The defect inspection apparatus and the defect inspection method of the present invention can inspect not only a silicon carbide single crystal 4-inch wafer but also a silicon carbide single crystal wafer having another diameter as a silicon carbide sample. Further, the inspection region of the silicon carbide sample is not necessarily the entire surface of the silicon carbide sample. The size of the inspection region of the silicon carbide sample can be set so that a 100 μm region of the silicon carbide sample can be imaged with a size of at least two pixels when the inspection region is imaged collectively with the imaging means. preferable. For example, when a PL light emission image is taken with a photographing means having an image sensor having a pixel number of 1 million using a 6-inch wafer of silicon carbide single crystal as a silicon carbide sample, the surface of the silicon carbide sample is divided into four parts. An inspection area is provided. When an inspection area having a width of 1/4 of the entire surface of a 6-inch wafer silicon carbide sample is photographed with an image sensor having one million pixels, a 100 μm region of the silicon carbide sample is photographed with two pixels in the vertical and horizontal directions. A crystal defect of a certain size can be detected.

<Another embodiment 2>
The conversion from the RGB value to the xy chromaticity value is not necessarily performed on all the pixels of the PL emission image. For example, the edge detection process is performed on the PL emission image, the RGB value is converted into the xy chromaticity value only in the region where the edge is detected, and the first threshold value and the second threshold value are compared. May be. This configuration has a small amount of data processing for conversion into xy chromaticity values, and can suppress an increase in inspection time and equipment cost. Therefore, particularly, a large image sensor having a number of pixels of tens of millions or more. This is useful when inspecting a large-diameter wafer exceeding 10 inches.

  Next, the defect inspection apparatus and the defect inspection method according to the present invention will be described in detail based on specific examples.

[Silicon carbide sample]
As a silicon carbide sample, a silicon carbide single crystal thin film substrate (4-inch wafer) manufactured by an epitaxial growth method is used, and a crystal defect having a known position of basal plane dislocation, stacking fault, and inclusion is prepared. did.

[Shooting PL flash images]
An ultraviolet LED having a wavelength of excitation light of 365 nm was used as a light source, the luminance distribution of the ultraviolet light was made uniform by a rod integrator lens, and the entire surface of the silicon carbide sample was irradiated. A silicon carbide sample that emits PL light upon irradiation is photographed with a digital camera having three primary color filters and an uncooled CCD image sensor with 4 million pixels, and RGB values in the sRGB color system are taken for each pixel. A PL emission image was obtained. The PL emission image was taken using a 35 mm lens under the conditions of an exposure time of 30 seconds, an aperture 4 and an ISO sensitivity of 800.
[Xy chromaticity value]
In the PL emission image, the RGB value of a pixel that captures the position of the basal plane dislocation, stacking fault, and inclusion of the silicon carbide sample is converted into tristimulus values X, Y, and Z of the CIE 1931 color system, and xy Chromaticity values were calculated. In the pixel corresponding to the position where the basal plane dislocation exists, the x value was 0.5761. The pixel corresponding to the position where the stacking fault exists has an x value of 0.160816. In the pixel corresponding to the position where the inclusion exists, the x value was 0.291956.

  As a result, when 0.4 is used as the first threshold for detecting the basal plane dislocation, stacking faults based on the x value obtained by converting the RGB value of the PL emission image into the xy chromaticity coordinates, It was confirmed that the basal plane dislocation can be detected by clearly distinguishing it from the inclusion. Further, when 0.25 is used as the second threshold value for detecting the stacking fault, the basal plane dislocation is based on the x value obtained by converting the RGB value of the PL emission image into the xy chromaticity coordinates, and It was confirmed that stacking faults can be detected by clearly distinguishing from inclusions.

  The defect inspection apparatus and the defect inspection method of the present invention can be used for inspection of a silicon carbide sample in a manufacturing process of a silicon carbide substrate and a power device element using the silicon carbide substrate.

DESCRIPTION OF SYMBOLS 10 Defect inspection apparatus 11 Irradiation means 11a Light source (ultraviolet LED)
11b Uniform means 12 Image taking means (digital camera)
12a Color filter 13 Conversion means 14 Judgment means 15 Correction means A Silicon carbide sample

Claims (9)

A defect inspection apparatus for detecting crystal defects in a silicon carbide sample,
Irradiating means for irradiating the inspection region of the silicon carbide sample with excitation light;
Imaging means for collectively capturing PL emission images in the inspection area via a color filter;
Conversion means for converting an RGB value in the RGB color space of the PL emission image into a CIE 1931 color system to obtain an x value in an xy chromaticity coordinate;
Determination means for determining the presence or absence of crystal defects in the silicon carbide sample,
The determination means refers to a first threshold value larger than the x value of the white point in the xy chromaticity coordinates and a second threshold value smaller than the x value of the white point, and the x value obtained by the conversion of the RGB value is the first value. It is determined that there is a basal plane dislocation at the position of the silicon carbide sample corresponding to a pixel exceeding one threshold, and the silicon carbide sample corresponding to a pixel whose x value obtained by conversion of the RGB value is less than the second threshold A defect inspection apparatus that determines that there is a stacking fault at the position.   The defect inspection apparatus according to claim 1, wherein the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28. The imaging means further captures a scattered light image in the inspection area,
The defect inspection apparatus according to claim 1, wherein the determination unit determines the presence or absence of a micropipe based on the scattered light image.   The said irradiation means is a defect inspection apparatus as described in any one of Claims 1-3 which has ultraviolet-ray LED and the homogenization means which homogenizes the ultraviolet light emitted from the said ultraviolet LED. The imaging unit is configured to capture a correction PL emission image by changing the imaging range of the PL emission image in a state where the inspection area is included,
The defect inspection apparatus according to claim 1, further comprising a correcting unit that corrects the PL light emission image based on the correction PL light emission image.   The defect inspection apparatus according to claim 1, wherein the silicon carbide sample is a silicon carbide single crystal epitaxial thin film.   The defect inspection apparatus according to any one of claims 1 to 6, wherein the silicon carbide sample is a silicon carbide single crystal 4-inch wafer. A defect inspection method for detecting crystal defects in a silicon carbide sample,
An irradiation step of irradiating the inspection region of the silicon carbide sample with excitation light;
An imaging process for collectively capturing PL emission images in the inspection area via a color filter;
A conversion step of converting an RGB value in the RGB color space of the PL emission image into a CIE 1931 color system to obtain an x value in an xy chromaticity coordinate;
A determination step of determining the presence or absence of crystal defects in the silicon carbide sample,
In the determination step, the first threshold value larger than the x value of the white point in the xy chromaticity coordinates and the second threshold value smaller than the x value of the white point are referred to, and the x value obtained by the conversion of the RGB values is the first value. It is determined that there is a basal plane dislocation at the position of the silicon carbide sample corresponding to a pixel exceeding one threshold, and the silicon carbide sample corresponding to a pixel whose x value obtained by conversion of the RGB value is less than the second threshold A defect inspection method for determining that a stacking fault exists at the position of.   The defect inspection method according to claim 8, wherein the first threshold value exists between 0.35 and 0.50, and the second threshold value exists between 0.20 and 0.28.