JP7763646B2 - Inspection device and inspection method - Google Patents

Description

One aspect of the present invention relates to an inspection device and an inspection method.

An inspection device is known that includes a semiconductor substrate and cuts a wafer, the back surface of which is provided with a functional device layer, along multiple lines by irradiating the wafer from the front side of the semiconductor substrate with laser light to form multiple rows of modified regions within the semiconductor substrate along each of the multiple lines. The inspection device described in Patent Document 1 is equipped with an infrared camera, which makes it possible to observe modified regions formed within the semiconductor substrate and processing damage formed in the functional device layer from the front side of the semiconductor substrate. For example, the inspection device estimates the state of cracks in the wafer after processing based on the results of this internal observation, and determines whether the processing is successful (whether the desired processing was achieved under the set processing conditions) based on the estimated crack state.

JP 2017-64746 A

In the internal observation described above, the positions of modified layers, cracks, etc. are estimated based on the feature quantities of feature points shown in the captured image related to the internal observation. Here, various structures, such as TEGs and films in the device pattern of the wafer, may affect (become noise) the detection of the above-mentioned feature quantities. In this case, the positions of modified layers, cracks, etc. may not be estimated properly. Potential noise in feature detection is not limited to the device pattern described above, but also includes, for example, grinding marks on the incident surface (surface) of the laser light, foreign matter adhering to the surface of the wafer, or foreign matter adhering to the observation optical system.

One aspect of the present invention was conceived in consideration of the above-mentioned circumstances, and relates to an inspection device and inspection method that can eliminate the influence of noise in internal observation after laser processing and improve the accuracy of estimating the processing state of a wafer.

An inspection device according to one aspect of the present invention includes an imaging unit that outputs light that is transparent to the wafer and detects the light that has propagated through the wafer to capture images of the interior of the wafer; a drive unit that moves the imaging unit along the vertical Z direction; and a control unit. The control unit is configured to control the drive unit to sequentially move the imaging unit to positions where each imaging region along the Z direction in a predetermined imaging range of the wafer can be imaged; control the imaging unit to capture each imaging region; and determine the suitability of the imaging range for internal observation after laser processing based on the captured images output by the imaging unit that detects the light associated with each imaging region.

In an inspection device according to one aspect of the present invention, an imaging unit that moves along the Z direction captures images of each imaging region along the Z direction in a predetermined imaging range of the wafer. The inspection device then determines the suitability of the imaging region for internal observation after laser processing based on the captured image of each imaging region. In this way, by taking into account the captured images of each imaging region in the Z direction of the wafer, it is possible to identify the influence of potential noise in internal observation after laser processing along the Z direction. Then, by identifying the influence of potential noise in each region along the Z direction of the imaging range and then determining the suitability of the imaging region for internal observation, it is possible to determine with high accuracy whether the imaging region is suitable for internal observation. This configuration makes it possible to perform internal observation in an imaging region suitable for internal observation after laser processing, thereby eliminating the influence of noise in internal observation after laser processing and improving the accuracy of estimating the processed state of the wafer.

The control unit may determine the suitability of the imaging range for internal observation based on the feature values of the feature points shown in the captured image for each imaging area. By taking into account the feature values of the feature points shown in the captured image, the magnitude of the influence on the detection of feature values during internal observation after laser processing can be appropriately determined, making it possible to more accurately determine whether the imaging range is suitable for internal observation.

The control unit may determine the suitability of an imaging range for internal observation based on the feature values of processing range feature points, which are feature points in the range corresponding to the area where internal observation is desired after laser processing, among the feature points shown in the captured image of each imaging range. With this configuration, only the feature values of feature points (processing range feature points) in the range that are likely to affect the detection of feature values during internal observation after laser processing are taken into account, making it possible to more accurately determine whether the imaging range is suitable for internal observation.

The control unit may compare the feature amounts of the processing range feature points with assumed feature amounts that are assumed to be the feature amounts of the feature points resulting from the laser processing for the region where internal observation is desired after laser processing, and determine that the imaging range is more suitable for internal observation the smaller the feature amount of the processing range feature points is compared to the assumed feature amount. In this way, the feature amounts of the processing range feature points are compared with the feature amounts of the feature points resulting from the laser processing (expected feature amounts), and if the feature amounts of the processing range feature points are relatively small and have little effect on the detection of the feature amounts in internal observation after laser processing, the imaging range is determined to be suitable for internal observation, thereby making it possible to more accurately determine whether the imaging range is suitable for internal observation.

The control unit may be further configured to output the results of the assessment of the suitability of internal observation. This allows the user to confirm the results and, based on the results, allows the user to determine the area in which to perform internal observation.

The control unit may determine the suitability of internal observation for each of multiple different imaging ranges and output the determination results for each of the multiple imaging ranges so that the user can select the imaging range for the area where internal observation will be performed after laser processing. In this way, the suitability of internal observation for each of multiple imaging ranges is determined and each determination result is output, allowing the user to select the area where internal observation will be performed while comparing each determination result.

The control unit may store at least one of the captured image and position information for the imaging range selected by the user as information related to internal observation after laser processing. This ensures that the information for the imaging range selected by the user is registered as information related to internal observation after laser processing, allowing internal observation to be performed using an imaging range that takes into account the user's selection (i.e., an imaging range that is less susceptible to noise).

The wafer has a device pattern on the surface opposite to the surface irradiated with the laser, and the control unit may determine the suitability of the imaging range for internal observation based on captured images of each imaging area excluding the device pattern. In areas where the device pattern is formed, the feature values of the feature points in the captured images become larger, but these are not areas where modified layers or cracks will form. Therefore, by excluding captured images of areas where the device pattern is formed and determining the suitability of internal observation from only the other captured images, it is possible to more accurately determine whether the imaging range is suitable for internal observation based on captured images of the area where internal observation will actually be performed.

An inspection method according to one aspect of the present invention involves capturing images of a predetermined imaging range within a wafer while changing the imaging area along the vertical Z direction, and determining the suitability of the imaging range for internal observation after laser processing based on the captured image of each imaging area.

One aspect of the present invention makes it possible to eliminate the effects of noise in internal observation after laser processing, thereby improving the accuracy of estimating the processing state of the wafer.

1 is a configuration diagram of a laser processing apparatus according to an embodiment; FIG. 2 is a plan view of a wafer according to an embodiment. 3 is a cross-sectional view of a portion of the wafer shown in FIG. 2. FIG. 2 is a configuration diagram of a laser irradiation unit shown in FIG. 1 . FIG. 2 is a configuration diagram of an inspection imaging unit shown in FIG. 1 . FIG. 2 is a configuration diagram of an imaging unit for alignment correction shown in FIG. 1 . 6A and 6B are cross-sectional views of a wafer for explaining the imaging principle of the inspection imaging unit shown in FIG. 5, and images at various locations taken by the inspection imaging unit. 6A and 6B are cross-sectional views of a wafer for explaining the imaging principle of the inspection imaging unit shown in FIG. 5, and images at various locations taken by the inspection imaging unit. 1 is a SEM image of modified regions and cracks formed within a semiconductor substrate. 1 is a SEM image of modified regions and cracks formed within a semiconductor substrate. 6 is a light path diagram for explaining the imaging principle of the inspection imaging unit shown in FIG. 5, and a schematic diagram showing an image at a focal point of the inspection imaging unit. FIG. 6 is a light path diagram for explaining the imaging principle of the inspection imaging unit shown in FIG. 5, and a schematic diagram showing an image at a focal point of the inspection imaging unit. FIG. FIG. 10 is a diagram illustrating crack detection. FIG. 10 is a diagram illustrating crack detection. FIG. 10 is a diagram illustrating the detection of a scratch. FIG. 10 is a diagram illustrating the detection of a scratch. FIG. 10 is a diagram illustrating the detection of a scratch. 1A and 1B are diagrams illustrating internal observation of a mirror wafer. FIG. 10 is a diagram illustrating noise factors in feature detection. FIG. 2 is a diagram illustrating a device pattern on a wafer. 10A and 10B are diagrams illustrating feature detection influenced by a device pattern. FIG. 10 is a diagram illustrating the determination of the suitability of internal observation. 10 is a flowchart illustrating an example of an inspection method. FIG. 10 is a diagram illustrating an example of noise determination processing. 10 is a flowchart illustrating an example of a noise determination process. FIG. 10 is a diagram illustrating an example of a noise determination result. FIG. 10 is a diagram illustrating an example of noise determination processing. 10 is a flowchart illustrating an example of a noise determination process. FIG. 10 is a diagram illustrating an example of a noise determination result. 10A and 10B are diagrams illustrating an example of detection of a BG mark according to a comparative example. 10A to 10C are diagrams illustrating an example of detection of a BG mark according to the present embodiment. FIG. 10 is a diagram illustrating an example of noise determination processing. 10 is a flowchart illustrating an example of a noise determination process. FIG. 10 is a diagram illustrating an example of a noise determination result.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In each drawing, the same or corresponding parts are designated by the same reference numerals, and duplicated explanations will be omitted.
[Configuration of laser processing device]

As shown in FIG. 1, the laser processing device 1 includes a stage 2, a laser irradiation unit 3, multiple imaging units 4 (imaging sections), 5, and 6, a drive unit 7 (drive section), a control section 8, and a display 150. The laser processing device 1 is a device that forms a modified region 12 in an object 11 by irradiating the object 11 with laser light L.

The stage 2 is a suction table that supports the object 11, for example, by adsorbing a film attached to the object 11. Although not shown in FIG. 1, a holding member 600 that holds the wafer 20 may be provided between the object 11 (wafer 20) and the stage 2, as shown in FIG. 19. The stage 2 is movable along both the X and Y directions, and is rotatable around an axis parallel to the Z direction. The X and Y directions are first and second horizontal directions that are perpendicular to each other, and the Z direction is the vertical direction.

The laser irradiation unit 3 focuses laser light L, which is transparent to the object 11, and irradiates the object 11. When the laser light L is focused inside the object 11 supported by the stage 2, the laser light L is particularly absorbed in the area corresponding to the focusing point C of the laser light L, forming a modified region 12 inside the object 11.

The modified region 12 is a region whose density, refractive index, mechanical strength, and other physical properties differ from those of the surrounding unmodified region. Examples of modified region 12 include a melt-processed region, a crack region, an insulation breakdown region, and a refractive index change region. The modified region 12 has the property that cracks tend to extend from the modified region 12 to the incident side of the laser light L and to the opposite side. These properties of the modified region 12 are utilized to cut the object 11.

As an example, when the stage 2 is moved along the X direction and the focal point C is moved along the X direction relative to the object 11, multiple modified spots 12s are formed in a row along the X direction. One modified spot 12s is formed by irradiating one pulse of laser light L. A row of modified regions 12 is a collection of multiple modified spots 12s lined up in a row. Adjacent modified spots 12s may connect with each other or may separate from each other depending on the relative movement speed of the focal point C with respect to the object 11 and the repetition frequency of the laser light L.

The imaging unit 4 captures images of the modified area 12 formed in the object 11 and the tip of the crack extending from the modified area 12.

Under the control of the control unit 8, the imaging units 5 and 6 capture an image of the object 11 supported on the stage 2 using light that passes through the object 11. The images captured by the imaging units 5 and 6 are used, for example, to align the irradiation position of the laser light L.

The drive unit 7 supports the laser irradiation unit 3 and the multiple imaging units 4, 5, and 6. The drive unit 7 moves the laser irradiation unit 3 and the multiple imaging units 4, 5, and 6 along the Z direction.

The control unit 8 controls the operation of the stage 2, laser irradiation unit 3, multiple imaging units 4, 5, and 6, and drive unit 7. The control unit 8 is configured as a computer device including a processor, memory, storage, and communication devices. In the control unit 8, the processor executes software (programs) loaded into memory, etc., and controls the reading and writing of data in the memory and storage, as well as communication via the communication devices.

The display 150 functions as an input unit that accepts information input from the user and as a display unit that displays information to the user.

[Object composition]
As shown in FIGS. 2 and 3 , the object 11 in this embodiment is a wafer 20. The wafer 20 includes a semiconductor substrate 21 and a functional device layer 22. While the present embodiment describes the wafer 20 as having the functional device layer 22, the wafer 20 may or may not include the functional device layer 22, and may be a bare wafer. The semiconductor substrate 21 has a back surface 21 a and a front surface 21 b. The semiconductor substrate 21 is, for example, a silicon substrate. The functional device layer 22 is formed on the back surface 21 a of the semiconductor substrate 21. The functional device layer 22 includes a plurality of functional devices 22 a arranged two-dimensionally along the back surface 21 a. The functional devices 22 a may be, for example, light-receiving devices such as photodiodes, light-emitting devices such as laser diodes, or circuit devices such as memories. The functional devices 22 a may be three-dimensionally configured by stacking multiple layers. The semiconductor substrate 21 has a notch 21 c indicating the crystal orientation, but an orientation flat may be provided instead of the notch 21 c.

The wafer 20 is cut into individual functional elements 22a along each of the multiple lines 15. When viewed in the thickness direction of the wafer 20, the multiple lines 15 pass between each of the multiple functional elements 22a. More specifically, when viewed in the thickness direction of the wafer 20, the lines 15 pass through the center of the street region 23 (the center in the width direction). The street region 23 extends in the functional element layer 22 so as to pass between adjacent functional elements 22a. In this embodiment, the multiple functional elements 22a are arranged in a matrix along the back surface 21a, and the multiple lines 15 are set in a lattice pattern. Note that although the lines 15 are virtual lines, they may also be actually drawn lines.

[Configuration of laser irradiation unit]
As shown in Fig. 4, the laser irradiation unit 3 includes a light source 31, a spatial light modulator 32, and a condenser lens 33. The light source 31 outputs laser light L, for example, by pulse oscillation. The spatial light modulator 32 modulates the laser light L output from the light source 31. The spatial light modulator 32 is, for example, a reflective liquid crystal (LCOS: Liquid Crystal on Silicon) spatial light modulator (SLM). The condenser lens 33 condenses the laser light L modulated by the spatial light modulator 32. Note that the condenser lens 33 may be a correction ring lens.

In this embodiment, the laser irradiation unit 3 irradiates the wafer 20 with laser light L from the front surface 21b of the semiconductor substrate 21 along each of the multiple lines 15, thereby forming two rows of modified regions 12a, 12b inside the semiconductor substrate 21 along each of the multiple lines 15. Of the two rows of modified regions 12a, 12b, modified region 12a is the modified region closest to the back surface 21a. Of the two rows of modified regions 12a, 12b, modified region 12b is the modified region closest to modified region 12a and closest to the front surface 21b.

The two rows of modified regions 12a, 12b are adjacent to each other in the thickness direction (Z direction) of the wafer 20. The two rows of modified regions 12a, 12b are formed by moving two focal points C1, C2 along a line 15 relative to the semiconductor substrate 21. The laser light L is modulated by the spatial light modulator 32 so that, for example, focal point C2 is located behind focal point C1 in the traveling direction and on the incident side of the laser light L. The modified regions may be formed using a single focus or multiple focuses, and in one pass or multiple passes.

The laser irradiation unit 3 irradiates the wafer 20 with laser light L from the front surface 21b of the semiconductor substrate 21 along each of the multiple lines 15. As an example, for a semiconductor substrate 21 that is a 400 μm-thick single-crystal silicon <100> substrate, two focusing points C1 and C2 are aligned at positions 54 μm and 128 μm from the back surface 21a, respectively, and the laser light L is irradiated from the front surface 21b of the semiconductor substrate 21 along each of the multiple lines 15 to the wafer 20. In this case, for example, if the conditions are such that the cracks 14 spanning two rows of modified regions 12a and 12b reach the back surface 21a of the semiconductor substrate 21, the wavelength of the laser light L is 1099 nm, the pulse width is 700 ns, and the repetition frequency is 120 kHz. The output of the laser beam L at focal point C1 is 2.7 W, the output of the laser beam L at focal point C2 is 2.7 W, and the relative movement speed of the two focal points C1 and C2 relative to the semiconductor substrate 21 is 800 mm/sec. If the number of processing passes is 5, for example, the processing positions for the wafer 20 described above may be ZH80 (328 μm from the back surface 21 a), ZH69 (283 μm from the back surface 21 a), ZH57 (234 μm from the back surface 21 a), ZH26 (107 μm from the back surface 21 a), or ZH12 (49.2 μm from the back surface 21 a). In this case, for example, the wavelength of the laser beam L may be 1080 nm, the pulse width may be 400 nsec, the repetition frequency may be 100 kHz, and the movement speed may be 490 mm/sec.

[Configuration of inspection imaging unit]
As shown in FIG. 5 , the imaging unit 4 (imaging section) includes a light source 41, a mirror 42, an objective lens 43, and a light detection unit 44. The imaging unit 4 outputs light that is transparent to the wafer 20 and detects the light that has propagated through the wafer 20, thereby capturing an image of the interior of the wafer 20. The light source 41 outputs light 11 that is transparent to the semiconductor substrate 21. The light source 41 is configured, for example, with a halogen lamp and a filter, and outputs light 11 in the near-infrared region. The light 11 output from the light source 41 is reflected by the mirror 42, passes through the objective lens 43, and is irradiated onto the wafer 20 from the surface 21b side of the semiconductor substrate 21. At this time, the stage 2 supports the wafer 20 on which the two rows of modified regions 12a and 12b have been formed as described above.

The objective lens 43 passes light l1 reflected by the back surface 21a of the semiconductor substrate 21. In other words, the objective lens 43 passes light l1 that has propagated through the semiconductor substrate 21. The numerical aperture (NA) of the objective lens 43 is, for example, 0.45 or greater. The objective lens 43 has a correction collar 43a. The correction collar 43a corrects aberrations that occur in the light l1 within the semiconductor substrate 21, for example, by adjusting the distance between the multiple lenses that make up the objective lens 43. Note that the means for correcting aberrations is not limited to the correction collar 43a, and other correction means such as a spatial light modulator may also be used. The light detection unit 44 detects light l1 that has passed through the objective lens 43 and the mirror 42. The light detection unit 44 is, for example, composed of an InGaAs camera, and detects light l1 in the near-infrared region. Note that the means for detecting (imaging) light l1 in the near-infrared region is not limited to an InGaAs camera, and other imaging means that perform transmission imaging, such as a transmission confocal microscope, may also be used.

The imaging unit 4 can capture images of each of the two rows of modified regions 12a and 12b, and the tips of each of the multiple cracks 14a, 14b, 14c, and 14d (details will be described later). Crack 14a is a crack that extends from modified region 12a toward the back surface 21a. Crack 14b is a crack that extends from modified region 12a toward the front surface 21b. Crack 14c is a crack that extends from modified region 12b toward the back surface 21a. Crack 14d is a crack that extends from modified region 12b toward the front surface 21b.

[Configuration of the imaging unit for alignment correction]
6 , the imaging unit 5 has a light source 51, a mirror 52, a lens 53, and a light detection unit 54. The light source 51 outputs light I2 that is transparent to the semiconductor substrate 21. The light source 51 is configured, for example, with a halogen lamp and a filter, and outputs light I2 in the near-infrared region. The light source 51 may be shared with the light source 41 of the imaging unit 4. The light I2 output from the light source 51 is reflected by the mirror 52 and passes through the lens 53, and is irradiated onto the wafer 20 from the front surface 21b side of the semiconductor substrate 21.

The lens 53 passes light I2 reflected by the back surface 21a of the semiconductor substrate 21. In other words, the lens 53 passes light I2 that has propagated through the semiconductor substrate 21. The numerical aperture of the lens 53 is 0.3 or less. That is, the numerical aperture of the objective lens 43 of the imaging unit 4 is greater than the numerical aperture of the lens 53. The light detection unit 54 detects light I2 that has passed through the lens 53 and mirror 52. The light detection unit 54 is composed of, for example, an InGaAs camera, and detects light I2 in the near-infrared region.

Under the control of the control unit 8, the imaging unit 5 irradiates the wafer 20 with light I2 from the front surface 21b and detects light I2 returning from the back surface 21a (functional device layer 22), thereby capturing an image of the functional device layer 22. Similarly, under the control of the control unit 8, the imaging unit 5 irradiates the wafer 20 with light I2 from the front surface 21b and detects light I2 returning from the positions where the modified regions 12a and 12b are formed on the semiconductor substrate 21, thereby capturing an image of the region including the modified regions 12a and 12b. These images are used to align the irradiation position of the laser light L. The imaging unit 6 has a similar configuration to the imaging unit 5, except that the lens 53 has a lower magnification (e.g., 6x in the imaging unit 5 and 1.5x in the imaging unit 6), and is used for alignment in the same way as the imaging unit 5.

[Principle of imaging by inspection imaging unit]
Using the imaging unit 4 shown in Figure 5, the focal point F (the focal point of the objective lens 43) was moved from the front surface 21b toward the back surface 21a of a semiconductor substrate 21 in which cracks 14 spanning two rows of modified regions 12a and 12b reached the back surface 21a, as shown in Figure 7. In this case, when the focal point F was focused on the tip 14e of the crack 14 extending from the modified region 12b toward the front surface 21b from the front surface 21b, the tip 14e could be seen (image on the right in Figure 7). However, when the focal point F was focused on the front surface 21b from the front surface 21b, the crack 14 itself and the tip 14e of the crack 14 reaching the back surface 21a could not be seen (image on the left in Figure 7). Furthermore, when the focal point F was focused on the back surface 21a of the semiconductor substrate 21 from the front surface 21b, the functional element layer 22 could be seen.

Using the imaging unit 4 shown in FIG. 5, the focal point F is moved from the front surface 21b toward the back surface 21a of a semiconductor substrate 21 in which cracks 14 spanning two rows of modified regions 12a, 12b have not yet reached the back surface 21a, as shown in FIG. 8. In this case, even if the focal point F is aligned from the front surface 21b to the tip 14e of the crack 14 extending from the modified region 12a to the back surface 21a, the tip 14e cannot be seen (the image on the left in FIG. 8). However, by aligning the focal point F from the front surface 21b to the region on the opposite side of the back surface 21a from the front surface 21b (i.e., the region on the functional device layer 22 side of the back surface 21a) and positioning a virtual focus Fv symmetrical to the focus F with respect to the back surface 21a at the tip 14e, the tip 14e can be seen (the image on the right in FIG. 8). The virtual focus Fv is a point symmetrical with the focus F, which takes into account the refractive index of the semiconductor substrate 21, with respect to the back surface 21a.

As described above, the reason why the crack 14 itself cannot be seen is presumably because the width of the crack 14 is smaller than the wavelength of the illumination light l1. Figures 9 and 10 are SEM (Scanning Electron Microscope) images of the modified region 12 and crack 14 formed inside a semiconductor substrate 21, which is a silicon substrate. Figure 9(b) is an enlarged image of region A1 shown in Figure 9(a), Figure 10(a) is an enlarged image of region A2 shown in Figure 9(b), and Figure 10(b) is an enlarged image of region A3 shown in Figure 10(a). As such, the width of the crack 14 is approximately 120 nm, which is smaller than the wavelength of light l1 in the near-infrared region (e.g., 1.1 to 1.2 μm).

Based on the above, the imaging principle is assumed to be as follows. As shown in Figure 11(a), when the focus F is positioned in air, light l1 does not return, resulting in a dark image (the image on the right in Figure 11(a)). As shown in Figure 11(b), when the focus F is positioned inside the semiconductor substrate 21, light l1 reflected from the back surface 21a returns, resulting in a whitish image (the image on the right in Figure 11(b)). As shown in Figure 11(c), when the focus F is aligned with the modified region 12 from the front surface 21b side, the modified region 12 absorbs, scatters, etc., a portion of the light l1 reflected from the back surface 21a and returns, resulting in an image in which the modified region 12 appears dark against a whitish background (the image on the right in Figure 11(c)).

As shown in (a) and (b) of Figure 12, when the focal point F is focused on the tip 14e of the crack 14 from the surface 21b side, for example, optical peculiarities (stress concentration, strain, discontinuity in atomic density, etc.) occurring near the tip 14e and light confinement occurring near the tip 14e cause scattering, reflection, interference, absorption, etc. of a portion of the light l1 reflected from the back surface 21a and returned, resulting in an image in which the tip 14e appears dark against a whitish background (the images on the right in (a) and (b) of Figure 12). As shown in (c) of Figure 12, when the focal point F is focused on a portion of the crack 14 other than the vicinity of the tip 14e from the surface 21b side, at least a portion of the light l1 reflected from the back surface 21a returns, resulting in a whitish image (the image on the right in (c) of Figure 12).

[Detection algorithm for internal observation]
With regard to the internal observation of the wafer 20 described above, an algorithm for detecting (identifying) the crack 14 and an algorithm for detecting (identifying) the scars associated with the modified region will be described in detail.

Figures 13 and 14 are diagrams illustrating crack detection. Figure 13 shows the results of internal observation (an image of the inside of the wafer 20). The control unit 8 first detects a group of straight lines 140 from the image of the inside of the wafer 20 as shown in Figure 13(a). To detect the group of straight lines 140, an algorithm such as the Hough transform or LSD (Line Segment Detector) is used. The Hough transform is a method of detecting straight lines by detecting all straight lines passing through a point on an image and weighting the lines that pass through more characteristic points. The LSD is a method of detecting straight lines by calculating the gradient and angle of brightness values in the image to estimate the area that will become a line segment, and then approximating this area to a rectangle.

Next, the control unit 8 detects a crack 14 from the group of lines 140 by calculating the similarity between the group of lines 140 and the crack line, as shown in Figure 14. As shown in the upper diagram of Figure 14, a crack line is characterized by being very bright in the Y direction both before and after the crack line compared to the brightness value on the line. For this reason, the control unit 8, for example, compares the brightness values of all pixels in the detected group of lines 140 with those before and after in the Y direction, and determines the number of pixels where the difference is greater than or equal to a threshold value both before and after as the similarity score. Then, the group of lines 140 detected with the highest similarity score to the crack line is determined to be the representative value for that image. A higher representative value is an indicator that the possibility of the presence of a crack 14 is higher. By comparing the representative values in multiple images, the control unit 8 determines those with relatively high scores as crack image candidates.

Figures 15 to 17 are diagrams explaining scratch detection. Figure 15 shows the results of internal observation (image of the inside of wafer 20). For the image of the inside of wafer 20 as shown in Figure 15(a), the control unit 8 detects corners (concentrations of edges) within the image as key points, and detects their position, size, and direction to detect feature points 250. Known methods for detecting feature points in this way include Eigen, Harris, Fast, SIFT, SURF, STAR, MSER, ORB, and AKAZE.

As shown in Figure 16, the dent 280 has a strong corner characteristic, as it is made up of circles, rectangles, and other shapes arranged at regular intervals. Therefore, by aggregating the feature amounts of the feature points 250 in the image, it is possible to detect the dent 280 with high accuracy. As shown in Figure 17, by comparing the total feature amounts for each image captured by shifting in the depth direction, a change in the peaks can be confirmed, indicating the amount of crack rows for each modified layer. The control unit 8 estimates the peak of this change as the position of the dent 280. By aggregating the feature amounts in this way, it is possible to estimate not only the dent position but also the pulse pitch.

[Processing for determining internal observation position]
The process of determining the internal observation position will be described in detail below. The process of determining the internal observation position is performed on the wafer 20 before the modified region is formed (the wafer 20 before laser processing). Here, when the modified region is formed for the purpose of cutting the wafer 20, the processing conditions for forming the modified region are derived in advance. These processing conditions are derived based on the results of internal observation of the wafer 20 after laser processing. Here, depending on the structure of the wafer 20, if the internal observation is not performed at an appropriate position, it may not be possible to perform the internal observation after laser processing with high accuracy.

Figure 18 is a diagram illustrating internal observation of a mirror wafer 520, an example of a wafer 20. The mirror wafer 520 does not have a functional element layer on the back surface 521a, and does not have a device pattern (details will be described later) in the street area of the back surface 521a. Assume that the imaging range for internal observation of the mirror wafer 520 includes a direct observation area, which is imaged by shifting the focus from the front surface 521b toward the back surface 521a, as shown in Figure 18(a), and a back surface reflection area, which is imaged by focusing from the front surface 521b on the area opposite the back surface 521a from the back surface 521b, and imaged by light reflected by the back surface 521a. Figure 18(b) is a diagram showing the feature amounts for each feature point in multiple images captured by shifting the focus in the depth direction. In Figure 18(b), the horizontal axis represents the feature amount, and the vertical axis represents the imaging depth. The results shown in Figure 18(b) are the results of internal observation when the mirror wafer 520 is laser-processed to create a bottom-side half-cut (BHC) state in which the crack 14 reaches the back surface 521a. As shown in Figure 18(b), in the direct observation area, feature data 901 of the modified area SD1 on the back surface 521a, feature data 902 of the modified area SD2 on the front surface 521b, and feature data 910 of the tip of the upper crack are detected. In addition, in the back surface reflection area, feature data 903 of the modified area SD1 on the back surface 521a and feature data 904 of the modified area SD2 on the front surface 521b are detected. Note that the numerical values of the feature data in Figure 18(b) (e.g., the numerical value "396" for feature data 901) indicate the total value of the aggregated features. In this way, for mirror wafers 520 that do not have device patterns and do not have other noise factors (described below), modified regions and the like can be appropriately identified from the internal observation results without any special adjustment to the internal observation position, and processing conditions can be appropriately derived from the internal observation results.

On the other hand, for wafers 20 that have some kind of noise factor in feature detection, internal observation after laser processing may not be performed with high accuracy unless internal observation is performed at an appropriate position. Figure 19 is a diagram explaining noise factors in feature detection. Noise factors here refer to factors that affect the above-mentioned feature quantities. Figure 19 illustrates, as examples of noise factors, the device pattern F1, noise F2 on the front surface 21b (incident surface), foreign matter F3 adhering to the front surface 21b, foreign matter F4 adhering to the light detection unit 44, noise F5 on the front surface of the stage 2, and noise F6 on the back surface of the holding member 600.

Figure 20 is a diagram illustrating device pattern F1 on wafer 20. Device pattern F1 is arranged on street region 23 and can include various structures such as TEGs and films. In the example shown in Figure 20, structures such as TEGs indicated by "1-A" to "3-F" are shown as device pattern F1 in the region of one channel (CH1) within the pattern shot of wafer 20. Hereinafter, structures such as TEGs indicated by "1-A" to "3-F" may be referred to as device patterns 1-A to 3-F. Note that while Figure 20 only shows device pattern F1 in the region of one direction (CH1) in grid-shaped street region 23, device pattern F1 may also exist in the region of orthogonal CH2.

Figure 21 is a diagram illustrating feature detection affected by device pattern F1. Figure 21(a) shows the internal observation results (feature detection results) at a position where device pattern F1 is not present. Figures 21(b) to 21(d) show the internal observation results (feature detection results) at positions where device patterns 1-A, 1-C, and 1-F (see Figure 20) are present, respectively. In Figures 21(a) to 21(d), the horizontal axis represents feature value and the vertical axis represents imaging depth. The results shown in Figures 21(a) to 21(d) are the results of internal observation when wafer 20 is laser processed to create a bottom-side half-cut (BHC) state in which cracks reach the backside, for example.

As shown in Figure 21(a), when the device pattern F1 is not present, similar to the internal observation results of the mirror wafer 520 described above (see Figure 18(b)), feature amount data 901 of the modified area SD1 on the back surface 21a, feature amount data 902 of the modified area SD2 on the front surface 21b, and feature amount data 910 of the tip of the upper crack are detected in the direct observation area, and feature amount data 903 of the modified area SD1 on the back surface 21a and feature amount data 904 of the modified area SD2 on the front surface 21b are detected in the back surface reflection area. Because the device pattern F1 is not present, only feature amounts related to the formation of the modified area are detected in the internal observation after laser processing.

On the other hand, as shown in Figures 21(b) to 21(d), when device pattern F1 is present, the feature quantities of device pattern F1 affect the feature quantities of the modified region. For example, in the internal observation results at the position where device pattern 1-A is present as shown in Figure 21(b), feature quantity data 950 and 960 from device pattern 1-A are detected near the position of the dent in modified region SD1, and these feature quantity data 950 and 960 cause the dent position in modified region SD1 to be detected as being shifted (shifted from the feature quantity data 901 and 903 shown in Figure 21(a)). This results in an erroneous detection.

Furthermore, in the internal observation results at the position where device pattern 1-C is present, shown in Figure 21(c), feature amount data 970 due to device pattern 1-C is detected. However, since the feature amount of feature amount data 970 is sufficiently small compared to the feature amounts of feature amount data 901, etc. of the modified area, in such a case, even if feature amount data 970 due to device pattern 1-C is detected, it can be distinguished from feature amount data 901, etc. of the modified area, and no false detection occurs.

Furthermore, in the results of internal observation at the position where device pattern 1-F is present, as shown in Figure 21(d), feature amount data 980 from device pattern 1-F is detected near the back surface 21a (at a position deeper than modified area SD1), and this feature amount data 980 detects a dent at a position even deeper than modified area SD1. This is an erroneous detection.

As such, depending on the type of device pattern F1, it may affect the detection of features related to the formation of modified regions, resulting in false detection. Therefore, when performing internal observation after laser processing on a wafer 20 having device pattern F1, it is necessary to determine the internal observation position in advance to prevent false detection. When determining the internal observation position so as not to be affected by device pattern F1, it is possible to consider, for example, an exclusion setting that does not perform internal observation in at least one of the Z direction and the XY direction.

Another noise factor shown in Figure 19, noise F2 on the surface 21b (incident surface), is noise caused by, for example, grinding marks (BG marks) on the surface 21b or unevenness in the film on the surface 21b. Wafers 20 are ground using a grinding device to thin them from their original thickness, and grinding marks may remain during this process. Grinding marks can take on various shapes on the surface 21b of the wafer 20. Depending on the shape of the grinding marks, the detection of feature quantities related to the formation of modified regions may be affected, resulting in false positives. Therefore, when performing internal observation after laser processing on a wafer 20 with noise F2 on the surface 21b, it is necessary to determine the internal observation position in advance to prevent false positives. When determining the internal observation position to avoid the influence of noise F2 on the surface 21b, it is possible to consider, for example, an exclusion setting that does not perform internal observation in at least one of the Z direction and the XY direction.

The foreign matter F3 adhering to surface 21b, which is another noise source shown in Figure 19, is, for example, dust or other foreign matter adhering to surface 21b. Such foreign matter F3 may affect the detection of feature quantities related to the formation of modified regions and result in false detection. Therefore, when performing internal observation after laser processing on a wafer 20 with foreign matter F3 adhering to surface 21b, it is necessary to determine the internal observation position in advance to prevent false detection. When determining the internal observation position so as not to be affected by foreign matter F3 adhering to surface 21b, it is possible to consider, for example, an exclusion setting that does not perform internal observation in at least one of the Z direction and the XY direction.

The foreign matter F4 adhering to the light detection unit 44, which is another noise source shown in Figure 19, is, for example, dust or other foreign matter adhering to the light detection unit 44. Such foreign matter F4 may affect the detection of feature quantities related to the formation of modified regions and result in false detection. Therefore, when performing internal observation after laser processing in an environment where foreign matter F4 is adhering to the light detection unit 44, it is necessary to determine the internal observation position in advance to prevent false detection. When determining the internal observation position so as not to be affected by foreign matter F4 adhering to the light detection unit 44, it is possible to consider, for example, an exclusion setting that does not perform internal observation in the Z direction.

Another noise factor shown in Figure 19, noise F5 on the surface of stage 2, is, for example, noise caused by the porous structure of the surface of stage 2. Such noise F5 can affect the detection of feature quantities related to the formation of modified regions and result in false detections. Therefore, when performing internal observation after laser processing in an environment where noise F5 on the surface of stage 2 is present, it is necessary to determine the internal observation position in advance to prevent false detections. When determining the internal observation position so as not to be affected by noise F5 on the surface of stage 2, it is possible to consider, for example, an exclusion setting that does not perform internal observation in the Z direction.

Another noise factor shown in Figure 19, noise F6 on the back surface of the holding member 600, is, for example, noise caused by embossing the back surface of the holding member 600. Such noise F6 can affect the detection of feature quantities related to the formation of the modified region and result in false detection. Therefore, when performing internal observation after laser processing in an environment where noise F6 on the back surface of the holding member 600 is present, it is necessary to determine the internal observation position in advance to prevent false detection. Note that when determining the internal observation position so as not to be affected by noise F6 on the back surface of the holding member 600, it is possible to consider, for example, an exclusion setting that does not perform internal observation in the Z direction.

Next, the process of the control unit 8, which determines the internal observation position while taking the above-mentioned noise factors into consideration, will be described. The control unit 8 performs the process of determining the internal observation position before laser processing. That is, the control unit 8 determines the internal observation position for the wafer 20 before laser processing is performed. The control unit 8 is configured to execute the following controls: first control, which controls the drive unit 7 to sequentially move the imaging unit 4 to a position where each imaging region along the Z direction in a predetermined imaging range of the wafer 20 can be imaged; second control, which controls the imaging unit 4 to image each imaging region; third control, which determines the suitability of internal observation for the imaging range after laser processing based on the captured image output from the imaging unit 4 that detects light related to each imaging region; and fourth control, which outputs the determination result of the suitability of internal observation.

In the first control, the control unit 8 first controls the drive unit 7 to move the imaging unit 4 to the imaging start position within the imaging range based on information indicating the imaging range set (input) as an inspection condition. More specifically, the control unit 8 controls the drive unit 7 to move the imaging unit 4 to the imaging start position based on the height set positions of the front surface 21b and back surface 21a of the wafer 20, the Z-direction coordinate position on the wafer 20, and information indicating the imaging range. The imaging range here includes a direct observation area (see, for example, FIG. 24 ) that is imaged by moving the focus from the front surface 21b toward the back surface 21a, and a back surface reflection area (see, for example, FIG. 24 ) that is imaged by focusing from the front surface 21b to the area on the opposite side of the back surface 21a from the front surface 21b and capturing light reflected by the back surface 21a. The control unit 8 then controls the drive unit 7 to move the imaging unit 4 along the Z direction from the imaging start position to the imaging end position.

In the second control, the control unit 8 controls the imaging unit 4 so that imaging is continuously performed by the imaging unit 4 moving in the Z direction in the first control, and each imaging area is imaged. Each imaging area is an area included in the imaging range, and is determined, for example, by the movement speed and imaging cycle of the imaging unit 4.

In the third control, the control unit 8 determines the suitability of the imaging range for internal observation based on the captured image of each imaging area. The control unit 8 determines the suitability of the imaging range for internal observation based on, for example, the feature amounts of the feature points shown in the captured image of each imaging area. The control unit 8 may determine the suitability of the imaging range for internal observation based on the feature amounts of the processing range feature points, which are feature points of the range corresponding to the area where internal observation is desired after laser processing, among the feature points shown in the captured image of each imaging area. The control unit 8 may also compare the feature amounts of the processing range feature points with the assumed feature amounts assumed to be the feature amounts of the feature points resulting from laser processing for the area where internal observation is desired after laser processing, and determine that the imaging range is more suitable for internal observation when the feature amounts of the processing range feature points are smaller than the assumed feature amounts.

Figure 22 is a diagram explaining the suitability assessment of internal observation at the position where device pattern F1 is present. Figure 22(a) shows the results of internal observation before laser processing at the position where device pattern 1-A is present, Figure 22(b) shows the results of internal observation before laser processing at the position where device pattern 1-C is present, and Figure 22(c) shows the results of internal observation before laser processing at the position where device pattern 1-F is present. In Figures 22(a) to 22(c), the horizontal axis shows feature amount, and the vertical axis shows imaging depth.

In the example shown in FIG. 22(a), the presence of device pattern 1-A results in the detection of feature quantity data 950, 960 corresponding to device pattern 1-A. When such feature quantity data 950, 960 are detected, the control unit 8 may first determine whether these feature points are processing range feature points, which are feature points of a range corresponding to an area where internal observation is desired after laser processing. Assume now that the feature points associated with feature quantity data 950, 960 are processing range feature points. In this case, the control unit 8 compares the feature quantity data 950, 960 of the processing range feature points with the assumed feature quantities assumed as feature quantities of feature points resulting from laser processing. As shown in FIG. 21(b), the feature quantity data 950, 960 of the processing range feature points are significantly larger than the assumed feature quantities assumed as feature quantities of feature points resulting from laser processing (e.g., feature quantity data 902 of modified region SD2 on the surface 21b side and feature quantity data 904 of modified region SD2 on the surface 521b side). In this case, the control unit 8 determines that the imaging range is not suitable for internal observation. In the example shown in Figure 22(a), the control unit 8 marks the imaging range as "X" and assigns a score of "25 points" indicating suitability for internal observation.

In the example shown in FIG. 22(b), the presence of device pattern 1-C results in the detection of feature quantity data 970 corresponding to device pattern 1-C. When such feature quantity data 970 is detected, the control unit 8 may first determine whether these feature points are processing range feature points, which are feature points of a range corresponding to the area where internal observation is desired after laser processing. Assume now that the feature points associated with feature quantity data 970 are processing range feature points. In this case, the control unit 8 compares the feature quantity data 970 of the processing range feature point with the assumed feature quantity assumed as the feature quantity of the feature point resulting from laser processing. As shown in FIG. 21(c), the feature quantity data 970 of the processing range feature point is extremely small compared to the assumed feature quantity assumed as the feature quantity of the feature point resulting from laser processing (e.g., feature quantity data 901 of modified area SD1 on the surface 21b side and feature quantity data 903 of modified area SD1 on the surface 521b side). In this case, the control unit 8 determines that the imaging range is suitable for internal observation. In the example shown in FIG. 22(b), the control unit 8 determines the imaging range as "OK" and scores the internal observation as "90 points." When comparing feature amount data of the processing range feature points with assumed feature amounts assumed to be feature amounts of feature points resulting from laser processing, comparisons are made between data of equivalent imaging depths. For example, the feature amount data 970 shown in FIG. 22(c) is compared with feature amount data 901 of the modified region SD1 on the front surface 21b side and feature amount data 903 of the modified region SD1 on the front surface 521b side (see FIG. 21(c)), which have equivalent detected imaging depths.

In the example shown in FIG. 22(c), the presence of device pattern 1-F allows feature data 980 corresponding to device pattern 1-F to be detected. Here, as shown in FIG. 22(d), the Z-direction position of feature data 980 corresponding to device pattern 1-F does not overlap with the position of feature data 901 for modified region SD1 on the back surface 21a, the position of feature data 902 for modified region SD2 on the front surface 21b, or the position of feature data 910 for the tip of the upper crack in the direct observation area. Furthermore, the Z-direction position of feature data 980 does not overlap with the position of feature data 903 for modified region SD1 on the back surface 21a, or the position of feature data 904 for modified region SD2 on the front surface 21b in the back surface reflection area. In other words, the feature points of feature data 980 are not processing range feature points. In this case, because the feature points of feature data 980 do not affect internal observation, the control unit 8 determines that the imaging range is suitable for internal observation. In the example shown in FIG. 22(c), the control unit 8 sets the result of the imaging range as "◯" and the score indicating the suitability of internal observation as "85 points." In this way, the control unit 8 may set the area in which the device pattern F1 is formed in the Z direction as the judgment exclusion area. The control unit 8 may then determine the suitability of internal observation for the imaging range from only the captured image relating to the imaging area excluding the judgment exclusion area (here, the area in which the device pattern F1 is formed).

The control unit 8 may determine the suitability of internal observation for each of multiple different imaging ranges. That is, for example, the control unit 8 may sequentially determine the suitability of internal observation for each of the imaging ranges shown in Figures 22(a) to 22(c) (the imaging ranges corresponding to device patterns 1-A to 1-F).

In the fourth control, the control unit 8 outputs the judgment results for the suitability of the internal observation to the display 150 in a displayable format. For example, when the control unit 8 has obtained judgment results for each of multiple imaging ranges, it outputs the judgment results for each of the multiple imaging ranges so that the user can select the imaging range for the area in which internal observation will be performed after laser processing. For example, when the control unit 8 has obtained judgment results for three device patterns 1-A to 1-F as shown in Figures 22(a) to 22(c), it may output this information so that a diagram showing the respective feature data, the judgment results, and a score indicating suitability are displayed side by side.

The control unit 8 stores at least one of the captured image and position information for the imaging range selected by the user as information related to internal observation after laser processing. As a result, when internal observation after laser processing is performed on the same wafer 20, internal observation will be performed in the imaging range selected by the user.

Figure 23 is a flowchart of an example of an inspection method for determining the position for internal observation. As shown in Figure 23, in this inspection method, the laser processing device 1 is first started (step S1) and warmed up (step S2).

Next, the imaging units 5 and 6 are controlled to perform wafer alignment regarding the irradiation position of the laser light L (step S3), and a height setting process is performed to set the Z-height, which is the processing depth (height) during laser processing (step S4). In the height setting process, the height setting position of the front surface 21b of the wafer 20 is set, and the height setting position of the back surface 21a is also set. Note that information from the wafer alignment and height setting process is used not only for the noise determination process described below, but also for internal observation of the stealth diced wafer 20, so the information may be carried over for the internal observation process. Furthermore, if the wafer alignment and height setting processes have already been completed during processing of the stealth diced wafer 20, the completed information may be carried over, and steps S3 and S4 may be omitted. Furthermore, steps S3 and S4 may be incorporated into the noise determination process (step S6) described below, as necessary.

Next, a judgment position (position in the X and Y directions) for judging the suitability of the internal observation is specified (step S5). The position in the X and Y directions may be linked to the planned processing line for stealth dicing, for example. In this case, for example, index movement is performed in the Y direction for each planned processing line, and the X direction may be determined by the user while viewing the IR camera image. Furthermore, the positions in the X and Y directions may be checked at equal intervals by automatic processing.

Next, a noise determination process (a process for determining the suitability of internal observation) is performed (step S6). The noise determination process is performed on the wafer 20 before any modified regions are formed inside it. Then, it is determined whether it is necessary to change the designated location to change the imaging range (step S7), and if so, the determination position (position in the X and Y directions) is re-specified (step S5). If not, the process is completed.

Next, the details of the noise determination process (step S6) described above will be described with reference to FIGS. 24 to 34. FIGS. 24 to 26 are diagrams illustrating a first example of the noise determination process. FIGS. 27 to 29 are diagrams illustrating a second example of the noise determination process. FIGS. 30 to 34 are diagrams illustrating a third example of the noise determination process.

A first example of the noise determination process will be described. FIG. 24 is a diagram illustrating the first example of the noise determination process. FIG. 25 is a flowchart related to the first example. FIG. 26 is a diagram illustrating an example of the noise determination results of the first example. In the first example, as shown in FIG. 24, when a recipe (processing conditions) for SDBG (Stealth Dicing Before Grinding) processing of a wafer with a device pattern and a thickness of 775 μm is derived through internal observation, the position for internal observation is determined as a preliminary step. In the first example, information indicating the imaging range and the noise determination exclusion area is first set. Now, as shown in FIG. 24, assume that the imaging range is set to ±400 μm from the height set position of the back surface 21a (i.e., 400 μm toward the direct observation area and 400 μm toward the back surface reflection area), and the noise determination exclusion area is set to ±20 μm from the height set position of the back surface 21a. Such noise exclusion areas are set to include, for example, areas where device patterns are formed, but not areas where internal observation is desired (for example, areas near the positions Va and Vb where modified areas are planned to be formed).

The imaging range may be set by specifying a numerical value, or may be set automatically based on a processing method or processing recipe, such as SDBG. The noise exclusion area may be set by specifying a numerical value, or may be automatically assigned by specifying the "device area" (area of the device pattern).

As shown in FIG. 25, in a first example of noise determination processing, first, an imaging start position is calculated based on the height set positions on the front surface 21b and back surface 21a of the wafer 20, the Z-direction coordinate position on the wafer 20, and information indicating the imaging range, and the drive unit 7 is controlled to move the imaging unit 4 to the imaging start position (step S101). Then, imaging is repeatedly performed while the imaging unit 4 moves in the Z direction from the imaging start position to the imaging end position (step S102). In the first example, imaging is performed from the imaging start position to the imaging end position, including the noise determination exclusion area. Imaging is repeatedly performed, for example, by focusing on the device pattern area.

Next, factors that could cause noise during internal observation are detected based on the image captured by the imaging unit 4 (step S103), and the suitability of internal observation within this imaging range is determined (noise determination) based on the detected information (step S104). The noise determination is performed excluding a range of ±20μ from the height set position on the back surface 21a, which is the noise determination exclusion area.

The noise determination result is then displayed on the display 150 (step S105). In Figure 26, the feature amount is shown on the horizontal axis and the imaging depth on the vertical axis, indicating the feature amount for each imaging region. In the example of the noise determination result shown in Figure 26, the feature amount is large only in the noise determination exclusion region, and not in the region to be determined. For this reason, as shown in Figure 26, the noise determination result is determined to be "Good" (suitable for internal observation). Note that, as described above, in the first example, although the noise determination exclusion region is excluded from the noise determination target, imaging is still performed. For this reason, the determination exclusion region can be changed as necessary based on the feature amount detected from the captured image of the noise determination exclusion region.

Finally, at least one of the captured image and the position information is registered as information related to internal observation after laser processing. Registration of internal observation positions may be performed at the user's discretion or automatically. Multiple internal observation positions may also be registered. Information on internal observation positions may also be linked to auto-alignment data, and the system may move automatically (or by button operation, etc.) to the registered position during internal observation.

A second example of the noise determination process will be described. FIG. 27 is a diagram illustrating the second example of the noise determination process. FIG. 28 is a flowchart related to the second example. FIG. 29 is a diagram illustrating an example of the noise determination results of the second example. In the second example, as shown in FIG. 27, when a recipe (processing conditions) for SDBG (Stealth Dicing Before Grinding) processing of a wafer having a device pattern and a thickness of 775 μm is derived through internal observation, the position for internal observation is determined as a preliminary step. In the second example, information indicating the imaging range and the noise determination exclusion area is first set. Now, as shown in FIG. 27, assume that the imaging range is set to ±400 μm from the height set position of the back surface 21a (i.e., 400 μm toward the direct observation area and 400 μm toward the back surface reflection area), and the noise determination exclusion area is set to ±20 μm from the height set position of the back surface 21a.

In the first example described above, the noise determination exclusion area was included in the imaging range, whereas in the second example, the noise determination exclusion area is excluded from the imaging range. For example, in cases where the device pattern area is known (known noise), not imaging the noise determination exclusion area in this way can shorten the imaging time.

As shown in FIG. 28, in the second example of noise determination processing, first, the imaging start position is calculated based on the height set positions on the front surface 21b and back surface 21a of the wafer 20, the Z-direction coordinate position on the wafer 20, and information indicating the imaging range, and the drive unit 7 is controlled to move the imaging unit 4 to the imaging start position (step S201). Then, imaging is repeatedly performed while the imaging unit 4 moves along the Z direction from the imaging start position to the imaging end position (step S202). In the second example, imaging is performed from the imaging start position to the imaging end position for each area excluding the noise determination exclusion area.

Next, factors that could cause noise during internal observation are detected based on the image captured by the imaging unit 4 (step S203), and the suitability of internal observation within the imaging range is determined (noise determination) based on the detected information (step S204). Noise determination is performed for all imaging areas.

The noise determination result is then displayed on the display 150 (step S205). In Figure 29, the horizontal axis represents the feature amount and the vertical axis represents the imaging depth, showing the feature amount for each imaging region. In the second example, the determination exclusion region is not imaged, so the feature amount for the determination exclusion region is not shown in Figure 29. In the example of the noise determination result shown in Figure 29, the feature amount is not large in the region to be determined, so the noise determination result is determined to be "Good" (suitable for internal observation). Note that the determination exclusion region can be changed (mainly by expanding the range) as necessary based on the noise determination result. Finally, at least one of the captured image and position information is registered as information related to internal observation after laser processing.

A third example of noise determination processing will be described. In this third example, noise determination will be described for a case where grinding marks (BG marks) on surface 21b, in addition to the device pattern, can be a problem as a noise factor. For example, when performing FC (full cut) processing in which cracks reach surface 21b after laser processing, surface 21b is also observed, and grinding marks (BG marks) can be a noise factor. First, an example of BG mark detection in a comparative example will be described with reference to Figure 30.

Figure 30(a) is a diagram showing an example of a BG mark formed on surface 21b. The BG mark can be formed in various directions relative to the planned processing line. For example, in areas where the BG mark is formed perpendicular or diagonally relative to the planned processing line, the BG mark is less likely to be mistakenly detected as a crack, etc., but in areas where the BG mark is formed horizontally relative to the planned processing line, the BG mark may be mistakenly detected as a crack, etc. In a BG mark detection method according to a comparative example, the crack detection position is determined based on, for example, the reflectance (brightness value) in the captured image. Figure 30(b) shows the brightness value at each position on the XY plane. However, such a detection method based on reflectance (brightness value) may not be able to properly distinguish between BG marks that are less likely to be mistakenly detected and BG marks that are more likely to be mistakenly detected.

Figure 31 is a diagram illustrating an example of BG mark detection according to this embodiment. The detection method according to this embodiment makes it possible to appropriately distinguish which positions on the wafer surface (within the XY plane) of BG marks will become noise. That is, BG marks that are nearly perpendicular to the processing direction, as shown in Figure 31(a), are not detected as line segments (noise that may be erroneously detected as cracks), as shown in Figure 31(b), and BG marks that are nearly horizontal to the processing direction, as shown in Figure 31(c), can be detected as line segment 700, as shown in Figure 31(d). This makes it possible to determine that imaging ranges containing BG marks that may be detected as line segments and therefore erroneously detected, are not suitable for internal observation.

Figure 32 is a diagram illustrating a third example of noise determination processing. Figure 33 is a flowchart related to this third example. Figure 34 is a diagram illustrating an example of noise determination results for this third example. In this third example, as shown in Figure 32, when deriving a recipe (processing conditions) for FC (full cut) processing of a wafer with a thickness t = 400 μm and a device pattern through internal observation, the position for internal observation is determined as a preliminary step. In this third example, information indicating the imaging range and the noise determination exclusion area is first set. Now, as shown in Figure 32, assume that the imaging range is set to ±420 μm from the height set position of the back surface 21a (i.e., 420 μm toward the direct observation area and 420 μm toward the back surface reflection area), and the noise determination exclusion area is set to ±20 μm from the height set position of the back surface 21a. In this third example, as in the second example, the noise determination exclusion area is excluded from the imaging range. The device pattern area in the Z direction (area on the back surface 21a) is included in the judgment exclusion area. On the other hand, the BG mark area in the Z direction (area on the front surface 21b) is also an area where cracks are detected after processing, so it is not included in the judgment exclusion area in the Z direction and is included in the imaging range.

As shown in FIG. 33, in the third example of noise determination processing, first, the imaging start position is calculated based on the height set positions on the front surface 21b and back surface 21a of the wafer 20, the Z-direction coordinate position on the wafer 20, and information indicating the imaging range, and the drive unit 7 is controlled to move the imaging unit 4 to the imaging start position (step S301). Then, imaging is repeatedly performed while the imaging unit 4 moves along the Z direction from the imaging start position to the imaging end position (step S302). In the third example, imaging is performed from the imaging start position to the imaging end position for each area excluding the noise determination exclusion area.

Next, factors that could cause noise during internal observation are detected based on the image captured by the imaging unit 4 (step S303), and the suitability of internal observation within this imaging range is determined (noise determination) based on the detected information (step S304). Noise determination is performed for all imaging areas.

In Figures 34(a) and 34(b), the horizontal axis represents feature amount and the vertical axis represents imaging depth, showing feature amounts for each imaging region. As shown in Figure 34(a), assume that feature amount data 990 related to a BG mark is detected near surface 21b. In this case, there is a risk that the line segment related to the BG mark may be erroneously detected as a crack during internal observation, so the noise judgment result is determined to be "x" (not suitable for internal observation). In the third example, noise judgment is performed at multiple locations on the XY plane. That is, after noise judgment is completed, the XY position of the imaging range is moved as shown in Figure 33 (step S305), and the processes of steps S301 to S304 are performed again. As shown in Figure 34(b), assume that in a noise judgment performed after changing the XY position, no feature amount data related to the BG mark is detected, and only feature amount data 971 related to the device pattern is detected. In the example of the noise determination result shown in Figure 34 (b), the feature amount is not large in the area to be determined, so the noise determination result is determined to be "Good" (suitable for internal observation). Note that, based on the noise determination result, the determination exclusion area can be changed (mainly by widening the range) as necessary. Finally, at least one of the captured image and position information is registered as information related to internal observation after laser processing. As described above, in the third example, a position where no false detection occurs is searched for while changing the XY position.

Here, for each of the noise factors described above, it may be possible to distinguish between those that can be noise and those that cannot be noise based on the device type, etc. For example, noise factors include noise caused by device patterns, noise caused by BG marks, and noise inside the silicon of the wafer 20 caused by foreign matter adhering to the light detection unit 44. In this case, depending on the device type, the user may know in advance whether noise caused by device patterns and noise caused by BG marks are present. Furthermore, the user may know in advance from past performance that there is no noise inside the silicon. In such cases, the GUI (Graphical User Interface) of the display 150 may be configured so that the user can specify to exclude judgment areas for noise factors that do not require noise judgment. In other words, the user may specify a combination of judgment areas for noise judgment using the GUI. For example, for a wafer with a mirror-finished front surface and a device pattern on the back surface, it is known in advance that judgment of noise caused by BG marks is unnecessary, so the user may specify in the GUI to perform a combination of judgment of noise caused by Device patterns and judgment of noise inside the silicon. Furthermore, if the areas inside the silicon where noise occurs are known in advance, the user may specify in the GUI that a combination of noise determination due to Debye patterns and noise determination due to BG marks be performed. This can increase the speed of noise determination. In addition to specifying a combination of the three noise determination methods described above, a GUI may also be set that allows specification of performing noise determination only in the vicinity of modified areas, for example.

Next, we will explain the effects of the laser processing device 1 according to this embodiment.

The laser processing device 1 according to this embodiment includes an imaging unit 4 that outputs light that is transparent to the wafer 20 and captures images of the interior of the wafer 20 by detecting the light that has propagated through the wafer 20; a drive unit 7 that moves the imaging unit 4 along the vertical Z direction; and a control unit 8. The control unit 8 is configured to control the drive unit 7 so that the imaging unit 4 sequentially moves to a position where each imaging region along the Z direction in a predetermined imaging range of the wafer 20 can be imaged; control the imaging unit 4 so that each imaging region is imaged; and determine the suitability of the imaging range for internal observation after laser processing based on the captured images output by the imaging unit 4 that detects the light associated with each imaging region.

In the laser processing apparatus 1 according to this embodiment, the imaging unit 4, which moves along the Z direction, captures images of each imaging region along the Z direction in a predetermined imaging range of the wafer 20. The laser processing apparatus 1 then determines the suitability of the imaging region for internal observation after laser processing based on the captured image of each imaging region. In this way, by considering the captured images of each imaging region in the Z direction of the wafer 20, it is possible to identify the influence of potential noise in internal observation after laser processing along the Z direction. Then, by identifying the influence of potential noise in each region along the Z direction of the imaging range and then determining the suitability of the imaging region for internal observation, it is possible to determine with high accuracy whether the imaging region is suitable for internal observation. This configuration makes it possible to perform internal observation in an imaging region suitable for internal observation after laser processing, thereby eliminating the influence of noise in internal observation after laser processing and improving the estimation accuracy of the processed state of the wafer.

The control unit 8 may determine the suitability of the imaging range for internal observation based on the feature values of the feature points shown in the captured image for each imaging area. By taking into account the feature values of the feature points shown in the captured image, the magnitude of the influence on the detection of feature values during internal observation after laser processing can be appropriately determined, making it possible to more accurately determine whether the imaging range is suitable for internal observation.

The control unit 8 may determine the suitability of an imaging range for internal observation based on the feature values of processing range feature points, which are feature points in the range corresponding to the area where internal observation is desired after laser processing, among the feature points shown in the captured image of each imaging range. With this configuration, only the feature values of feature points (processing range feature points) in the range that are likely to affect the detection of feature values during internal observation after laser processing are taken into account, making it possible to more accurately determine whether the imaging range is suitable for internal observation.

The control unit 8 may compare the feature amounts of the processing area feature points with assumed feature amounts that are assumed to be the feature amounts of the feature points resulting from the laser processing for the area where internal observation is desired after laser processing, and determine that the imaging area is more suitable for internal observation the smaller the feature amount of the processing area feature points is compared to the assumed feature amount. In this way, the feature amounts of the processing area feature points are compared with the feature amounts of the feature points resulting from the laser processing (expected feature amounts), and if the feature amounts of the processing area feature points are relatively small and have little effect on the detection of the feature amounts during internal observation after laser processing, the imaging area is determined to be suitable for internal observation, thereby making it possible to more accurately determine whether the imaging area is suitable for internal observation.

The control unit 8 may further be configured to output the judgment result of the suitability of internal observation. This allows the user to confirm the judgment result and, based on the judgment result, the user can decide the area in which to perform internal observation.

The control unit 8 may determine the suitability of internal observation for each of multiple different imaging ranges and output the determination results for each of the multiple imaging ranges so that the user can select the imaging range for the area where internal observation will be performed after laser processing. In this way, the suitability of internal observation for each of multiple imaging ranges is determined and each determination result is output, allowing the user to select the area where internal observation will be performed while comparing each determination result.

The control unit 8 may store at least one of the captured image and position information for the imaging range selected by the user as information related to internal observation after laser processing. This ensures that the information for the imaging range selected by the user is registered as information related to internal observation after laser processing, allowing internal observation to be performed using an imaging range that takes into account the user's selection (i.e., an imaging range that is less susceptible to noise).

The wafer 20 has a device pattern on the surface opposite to the surface irradiated with the laser, and the control unit 8 may determine the suitability of the imaging range for internal observation based on captured images of each imaging area excluding the device pattern. In areas where the device pattern is formed, the feature values of the feature points in the captured images become larger, but these are not areas where modified layers or cracks will form. Therefore, by excluding captured images of areas where the device pattern is formed and determining the suitability of internal observation from only the other captured images, it is possible to more accurately determine whether the imaging range is suitable for internal observation based on captured images of the area where internal observation will actually be performed.

1...laser processing device (inspection device), 4...imaging unit (imaging section), 7...driving unit (driving section), 8...control section, 20...wafer.

Claims (9)

an imaging unit that outputs light that is transmissive to the wafer and detects the light that has propagated through the wafer, thereby imaging the inside of the wafer;
a drive unit that moves the imaging unit along a Z direction that is a vertical direction;
a control unit,
The control unit
controlling the driving unit so that the imaging unit sequentially moves to a position where each imaging region along the Z direction in a predetermined imaging range of the wafer can be imaged;
controlling the imaging unit so that each imaging area is imaged;
determining whether the imaging range is suitable for internal observation after laser processing based on the captured image output from the imaging unit that detected the light related to each imaging region;
is configured to run
the control unit determines the suitability of the internal observation for the imaging range based on a feature amount of a feature point shown in the captured image for each of the imaging regions;
The control unit determines the suitability of the internal observation for the imaging range based on the feature amounts of processing range feature points, which are feature points in a range corresponding to an area where internal observation is desired after laser processing, among the feature points shown in the captured image for each imaging range . 2. The inspection device according to claim 1, wherein the control unit compares, for an area where internal observation is desired after the laser processing, a feature amount of the processing range feature point with an assumed feature amount that is assumed to be a feature amount of a feature point caused by the laser processing, and determines that the imaging range is more suitable for the internal observation as the feature amount of the processing range feature point is smaller than the assumed feature amount. 3. The inspection device according to claim 1 , wherein the control unit is further configured to output a determination result of whether the internal observation is appropriate. The control unit
determining the suitability of the internal observation for each of the plurality of imaging ranges that are different from one another;
4. The inspection device according to claim 3 , wherein the judgment results for each of the plurality of imaging ranges are output so that a user can select the imaging range relating to an area for internal observation after laser processing. 5. The inspection device according to claim 4 , wherein the control unit stores at least one of the captured image and position information for the imaging range selected by a user as information related to internal observation after laser processing. the wafer has a device pattern on a surface opposite to the surface irradiated with the laser;
The inspection device according to claim 1 , wherein the control unit determines the suitability of the internal observation for the imaging range based on the captured image of an imaging area excluding the device pattern among the imaging areas. an imaging unit that outputs light that is transmissive to the wafer and detects the light that has propagated through the wafer, thereby imaging the inside of the wafer;
a drive unit that moves the imaging unit along a Z direction that is a vertical direction;
a control unit,
The control unit
controlling the driving unit so that the imaging unit sequentially moves to a position where each imaging region along the Z direction in a predetermined imaging range of the wafer can be imaged;
controlling the imaging unit so that each imaging area is imaged;
determining whether the imaging range is suitable for internal observation after laser processing based on the captured image output from the imaging unit that detected the light related to each imaging region;
is configured to run
The control unit is further configured to output a determination result of whether the internal observation is appropriate,
The control unit
determining the suitability of the internal observation for each of the plurality of imaging ranges that are different from one another;
An inspection device that outputs a determination result for each of the plurality of imaging ranges so that a user can select the imaging range relating to an area for internal observation after laser processing. taking an image of a predetermined imaging range inside the wafer while changing an imaging region along a Z direction which is a vertical direction;
and determining suitability of the imaging range for internal observation after laser processing based on the captured image of each imaging region ,
determining the suitability of the internal observation for the imaging range based on feature amounts of feature points shown in the captured images for each of the imaging regions;
An inspection method for determining the suitability of the internal observation for each imaging area based on the feature amounts of processing area feature points, which are feature points in a range corresponding to an area where internal observation is desired after laser processing, among the feature points shown in the captured image for each imaging area . taking an image of a predetermined imaging range inside the wafer while changing an imaging region along a Z direction which is a vertical direction;
and determining suitability of the imaging range for internal observation after laser processing based on the captured image of each imaging region,
and outputting a determination result of the suitability of the internal observation.
determining the suitability of the internal observation for each of the plurality of imaging ranges that are different from one another;
An inspection method, comprising: outputting a judgment result for each of the plurality of imaging ranges so that a user can select the imaging range relating to an area for internal observation after laser processing.