Kerf check method

Abstract

To enable a kerf to be properly checked while reducing burdens on a worker.SOLUTION: A kerf check method, which checks a kerf formed on a wafer, includes: a photographing step of photographing the kerf with a camera; a recognizing step of recognizing the kerf by brightness in an image photographed in the photographing step; a counting step of counting, as pixel number in a width direction of the kerf, pixel exceeding a preset kerf minimum width, for each position in a longitudinal direction of the kerf, from the recognized kerf; a reference value setting step of executing the photographing step, the recognizing step and the counting step repeatedly plural times and setting a reference value, using an average value, a median value and a mode value of the pixel numbers counted in the counting step; a threshold setting step of setting a threshold for determining whether the kerf is acceptable, using standard deviations with the reference value as a reference, on the basis of the pixel numbers counted in the counting step; and a determining step of determining whether the kerf is acceptable, depending on whether the pixel numbers are equal to or less than the threshold or over the threshold.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a kerf check method. [Background technology]

[0002] In the processing equipment, to check whether the wafer streets are being cut properly, the kerf formed by cutting is photographed with a camera and the size of chipping that has occurred on both sides of the kerf is checked. This process is called a kerf check and is described in, for example, Patent Documents 1 to 4.

[0003] If the kerf check detects chipping of a size that exceeds a threshold (allowable width) set by the operator, the processing device will generate an error. Also, if chipping of a size that exceeds the threshold is detected more than a set number of times, the processing device may generate an error. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-197702 [Patent Document 2] Japanese Patent Publication No. 2023-050704 [Patent Document 3] Japanese Patent Application Publication No. 2020-077668 [Patent Document 4] Japanese Patent Publication No. 2022-158458 Summary of the Invention [Problem to be solved by the invention]

[0005] As described above, the kerf check is performed based on a threshold set by the operator. This requires the operator to set an appropriate threshold. Furthermore, in order to set an appropriate threshold, the operator checks the state of the kerf using an image of the kerf taken when setting the threshold. This presents a problem in that it is difficult to increase productivity.

[0006] The present invention has been made in view of the above points, and has an object to provide a technique that enables appropriate kerf checks while reducing the burden on the worker. [Means for solving the problem]

[0007] A kerf check method according to one aspect of the present invention is a method for checking a kerf formed in a wafer, and includes the following steps: an imaging step of imaging the kerf with a camera; a recognition step of recognizing the kerf based on brightness in the image captured in the imaging step; a counting step of counting pixels from the recognized kerf that exceed a predetermined minimum kerf width for each longitudinal position of the kerf as the number of pixels in the width direction of the kerf; a reference value setting step of performing the imaging step through the counting step multiple times and setting a reference value based on the average, median, or mode of the number of pixels counted in the counting step; a threshold setting step of setting a threshold for determining the pass / fail of the kerf based on the standard deviation with respect to the reference value, based on the number of pixels counted in the counting step; and a determination step of determining the pass / fail of the kerf based on whether the number of pixels is equal to or less than the threshold or exceeds the threshold. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a technique that enables appropriate kerf checks while reducing the burden on the worker. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 2 is a diagram schematically illustrating the configuration of a processing device. [Figure 2] FIG. 2 is a diagram showing how a wafer is cut by a processing device. [Figure 3] FIG. 10 is a diagram showing how a processed groove is imaged by the processing device. [Figure 4] 10A and 10B are diagrams illustrating a configuration for capturing an image of a processed groove formed in a wafer. [Figure 5] 10 is a flowchart illustrating an example of processing performed by a processing device. [Figure 6] FIG. 2 is an enlarged view of a processed groove formed in a wafer. [Figure 7] FIG. 10 is a diagram showing the binarization result of a region of interest that exceeds the minimum kerf width. [Figure 8] FIG. 10 illustrates a truncated normal distribution that the histogram of pixel counts is assumed to follow. [Figure 9] This figure shows positions that are one standard deviation away from the mean value of the truncated normal distribution shown in FIG. 8 and positions that are two times the standard deviation away from the mean value in a histogram of the number of pixels. DETAILED DESCRIPTION OF THE INVENTION

[0010] FIG. 1 is a diagram schematically showing the configuration of a processing device 2 according to this embodiment. FIG. 2 shows how a wafer 11 is cut by the processing device 2. FIG. 3 is a diagram showing how a processed groove 21 is imaged by the processing device 2. The X-axis, Y-axis, and Z-axis directions shown in FIGS. 1 to 3 are perpendicular to one another. The X-axis and Y-axis directions are approximately horizontal, and the Z-axis direction is the up-down direction (approximately vertical). The processing device 2 according to this embodiment will be described with reference to FIGS. 1 to 3.

[0011] The processing device 2 is a processing device that processes the wafer 11 and checks the processed grooves 21 (also called kerfs) formed in the wafer 11 by processing using a kerf check method described below. In the following, an example will be described in which the processing device 2 is a cutting device that cuts the wafer 11 with a cutting blade 40, but the processing device that performs the kerf check method described below is not limited to cutting devices that cut the wafer with a cutting blade 40. For example, it may be a laser processing device that processes the wafer 11 by irradiating it with a laser beam.

[0012] The processing device 2 processes the wafer 11 held by the frame 19, but the wafer 11 handled by the processing device that performs the kerf check using the kerf check method described below does not necessarily have to be part of a frame set. The wafer 11 may be transported, processed, and kerf checked on its own. Note that the frame set is a set in which the dicing tape 17, frame 19, and wafer 11 are integrated together by adhering the back surface 112 of the wafer 11 to a dicing tape 17 that is adhered to the opening of the annular frame 19 so as to cover it, as shown in FIGS. 1 to 3 .

[0013] The wafer 11 processed by the processing device 2 is, for example, a circular semiconductor wafer made of silicon, gallium arsenide, or the like, but there are no limitations on the material, shape, etc. of the wafer 11 handled by the processing device that performs the kerf check using the kerf check method described below. The wafer 11 may be a wafer other than a semiconductor wafer, and may also be a wafer other than a circular one, for example, a rectangular wafer. As shown in Figures 2 and 3, the surface 111 of the wafer 11 is partitioned by a plurality of planned division lines 13 formed in a grid pattern, and devices 15 such as ICs and LSIs are formed in each partitioned area.

[0014] As shown in FIG. 1 , the processing apparatus 2 includes a base 4 that supports various components. The base 4 has a rectangular opening 401. A cassette support table 6 is provided in this opening 401 so that it can move up and down. A cassette 8 that stores wafers 11 is placed on the upper surface of the cassette support table 6. The cassette 8 placed on the cassette support table 6 has multiple storage sections for storing wafers 11 arranged in the Z-axis direction. The cassette 8 has an opening for loading and unloading the wafers 11 into and out of the storage sections, and is placed on the cassette support table 6 with the opening facing a transport unit (not shown). In the processing apparatus 2, the cassette support table 6 moves up and down to move the cassette 8 placed on the cassette support table 6 relative to the transport unit, thereby allowing any wafer 11 stored in the cassette 8 to be loaded or unloaded by the transport unit.

[0015] As shown in FIG. 1 , the processing device 2 includes a gate-shaped support structure 20 that supports two cutting units 18. A rectangular opening 402 extending in the X-axis direction is formed in the upper surface of the base 4, and the support structure 20 is provided on the upper surface so as to straddle the opening 402. The opening 402 is covered by a bellows-shaped dustproof and drip-proof cover 12 that moves together with the X-axis moving table 10. An X-axis moving mechanism (not shown) that moves the X-axis moving table 10 in the X-axis direction is provided below the dustproof and drip-proof cover 12. The X-axis moving mechanism includes a ball screw that moves back and forth in the X-axis direction, a pair of guide rails parallel to the ball screw, and a pulse motor (not shown). The X-axis moving table 10 is slidably attached to the pair of guide rails by threading a nut provided on the underside of the X-axis moving table 10 onto the ball screw. The X-axis moving mechanism moves the X-axis moving table 10 in the X-axis direction along the guide rails by rotating the ball screw with the pulse motor.

[0016] The X-axis moving table 10 is provided with a chuck table 14 that holds the wafer 11. The chuck table 14 suction-holds the wafer 11 on a holding surface 141. The chuck table 14 has a disk shape, and the holding surface 141 that holds the wafer 11 is made of porous ceramic or the like. The chuck table 14 is connected to a vacuum suction source (not shown), and is sucked by the vacuum suction source to hold the wafer 11 placed on the holding surface 141 by suction. Furthermore, four clamps 16 are installed around the periphery of the chuck table 14. The four clamps 16 secure, from all four sides, an annular frame 19 that supports the wafer 11 placed on the chuck table 14. The wafer 11 is loaded and unloaded from the chuck table 14 by a transport unit (not shown), and the wafer 11 is placed on the chuck table 14 with the front surface 111 exposed upward.

[0017] The chuck table 14 moves in the X-axis direction together with the wafer 11 as the X-axis moving table 10 moves in the X-axis direction. Specifically, the chuck table 14 moves between a load / unload position where the wafer 11 is loaded / unloaded relative to the chuck table 14 and a cutting position where the wafer 11 is cut by the cutting unit 18 as the X-axis moving table 10 moves. The chuck table 14 is also fed in the X-axis direction during cutting as the X-axis moving table 10 moves. The chuck table 14 is also connected to a table rotation mechanism (not shown). When the table rotation mechanism is driven, the chuck table 14 rotates around a rotation axis parallel to the Z-axis direction (vertical direction), thereby changing the orientation of the wafer 11 placed on the chuck table 14.

[0018] Two cutting unit moving mechanisms 22 are provided on the front surface of the support structure 20. Each cutting unit moving mechanism 22 moves each cutting unit 18 in the Y-axis direction and the Z-axis direction. Each cutting unit moving mechanism 22 is configured to move each cutting unit 18 in the Y-axis direction, and includes a Y-axis moving plate 26 that moves in the Y-axis direction, a Y-axis ball screw 28 that moves back and forth in the Y-axis direction, and a pulse motor 30. The two cutting unit moving mechanisms 22 share a pair of Y-axis guide rails 24 that are parallel to the Y-axis direction. The Y-axis guide rails 24 and the Y-axis ball screw 28 are attached to the support structure 20. Each cutting unit moving mechanism 22 is configured to move each cutting unit 18 in the Z-axis direction, and includes a pair of Z-axis guide rails 32 that are parallel to the Z-axis direction, a Z-axis moving plate 34 that moves in the Z-axis direction, a Z-axis ball screw 36 that moves back and forth in the Z-axis direction, and a Z-axis pulse motor 38. The Z-axis guide rails 32 and the Z-axis ball screw 36 are attached to the Y-axis moving plate 26. Each cutting unit 18 is provided below each Z-axis moving plate 34 .

[0019] The cutting unit 18 is equipped with a cutting blade 40 that cuts the wafer 11. The cutting blade 40 is fixed to a spindle whose rotation axis faces the Y-axis direction. An imaging unit 44 is installed adjacent to the cutting unit 18. The imaging unit 44 is an example of a camera that captures images of the wafer 11, and is fixed to the same Z-axis moving plate 34 as the cutting unit 18, for example, so as to move integrally with the cutting unit 18. The imaging unit 44 is equipped with an imaging element, such as a CCD (Charge-Coupled Device) imaging element or a CMOS (Complementary MOS) imaging element.

[0020] The cutting unit 18 and the imaging unit 44 are indexed and fed in the Y-axis direction by the cutting unit moving mechanism 22 moving the Y-axis moving plate 26 in the Y-axis direction. In addition, the cutting unit 18 and the imaging unit 44 are raised and lowered in the Z-axis direction by the cutting unit moving mechanism 22 moving the Z-axis moving plate 34 in the Z-axis direction. This allows the cutting unit 18 to arbitrarily change its position relative to the wafer 11 held on the chuck table 14, and allows cutting of any position on the wafer 11 along the X-axis direction. In addition, the imaging unit 44 can capture an image of any machined groove 21 formed by cutting.

[0021] A circular opening 403 is further formed in the base 4. A cleaning mechanism 46 for cleaning the wafer 11 after cutting is provided inside the opening 403. The cleaning mechanism 46 has, for example, a spinner table, and cleans the wafer 11 placed on the spinner table with cleaning water, and further dries the wafer 11 with air jetted from a nozzle.

[0022] The processing device 2 includes a control device 48 and an input device 50. The control device 48 is a computer having a processor such as a CPU (Central Processing Unit) and a storage unit 49. The input device 50 is a device for inputting information to the control device 48, such as a mouse or a keyboard. In the processing device 2, the processor of the control device 48 executes a program stored in the storage unit 49, thereby performing various processes such as cutting the wafer 11 and checking the processed grooves 21 formed in the wafer 11 (kerf check).

[0023] Referring to FIG. 2, the cutting process performed by the processing apparatus 2 according to this embodiment will be described. When cutting the wafer 11, the processing apparatus 2 first captures an image of the wafer 11 held on the chuck table 14 using the imaging unit 44, and detects the position, orientation, etc. of the dividing line 13 based on the captured image. Information regarding the position, orientation, etc. of the dividing line 13 is stored in the memory unit 49. Next, the processing apparatus 2 moves and rotates the chuck table 14 and the cutting unit 18 relative to each other, and positions the cutting blade 40 above the dividing line 13 to be processed and parallel to the dividing line 13. Thereafter, the processing apparatus 2 rotates the cutting blade 40 and lowers it to a height at which it can contact the wafer 11, and brings the cutting blade 40 into contact with the dividing line 13 to be processed. Furthermore, the chuck table 14 and the cutting unit 18 move relative to each other in a direction parallel to the dividing line 13 to be processed. 2, the wafer 11 is cut along the intended dividing lines 13 of the processing target, and grooves 21 are formed in the wafer 11 along the intended dividing lines 13. The grooves 21 formed by the cutting process may have a depth that completely cuts the wafer 11, or may have a depth that does not completely cut the wafer 11. In other words, the cutting process may be a full cut or a half cut.

[0024] FIG. 4 is a diagram illustrating a configuration for imaging the kerf grooves 21 formed in the wafer 11. Referring to FIGS. 3 and 4, imaging of the kerf grooves 21 performed by the processing apparatus 2 according to this embodiment will be described. As shown in FIGS. 3 and 4, the imaging unit 44 includes a housing 52, a microscope unit 54, an oblique illumination unit 56, and a light source 60, and is controlled by a control device 48 to image the kerf grooves 21 and obtain images of the kerf grooves 21 used for kerf checking. When imaging the kerf grooves 21, the imaging unit 44 performs coaxial epi-illumination and oblique illumination on the wafer 11 on which the kerf grooves 21 are formed. Coaxial epi-illumination is performed by deflecting light LA ​​emitted from the light source 60 vertically downward by a half mirror 58 provided in the housing 52 and irradiating the light onto the surface 111 of the wafer 11 via an objective lens 62 of the microscope unit 54. Oblique illumination is performed by irradiating the surface 111 of the wafer 11 with light LB emitted from a light source 66 of an oblique illumination unit 56 arranged in a ring shape surrounding the objective lens 62. The light reflected from the wafer 11 subjected to coaxial epi-illumination and oblique illumination is focused onto the image pickup element 64 via a half mirror 58 by the objective lens 62. The image pickup unit 44 converts the light incident on the image pickup element 64 into an electrical signal, thereby capturing an image of the groove 21 formed in the wafer 11 and generating an image of the groove 21, which is then output to the control device 48. The image of the groove 21 output to the control device 48 is stored in a memory unit 49.

[0025] FIG. 5 is a flowchart showing an example of processing performed by the processing device 2. FIG. 6 is an enlarged view of a processed groove 21 formed in a wafer. FIG. 7 is a diagram showing the binarization result of a region of the region of interest R that exceeds the kerf minimum width W. FIG. 8 is a diagram showing a truncated normal distribution B that is assumed to follow the histogram of pixel counts. FIG. 9 is a diagram showing, on the histogram of pixel counts, positions that are spaced a standard deviation σ from the mean value μ of the truncated normal distribution B shown in FIG. 8 and positions that are twice the standard deviation σ from the mean value μ. The kerf check method used in the processing device 2 described above will be described below with reference to FIGS. 5 to 9.

[0026] 5 is started by the control device 48 executing a predetermined program, the processing device 2 cuts the wafer 11 along the planned dividing lines 13 with the cutting unit 18 to form a cutting groove 21 (kerf) (step S1). In step S1, an image of the wafer 11 is taken to recognize the planned dividing lines 13, the cutting blade 40 is aligned on the planned dividing lines 13, and the wafer 11 is cut along the planned dividing lines 13.

[0027] When the groove 21 is formed, the processing device 2 captures an image of the groove 21 (kerf) using the imaging unit 44 (step S2). Step S2 is an example of an imaging step in which the groove 21 is captured by the imaging unit 44 (camera). Thereafter, the processing device 2 recognizes the groove 21 (kerf) based on the image acquired in the imaging step (step S3).

[0028] In step S3, the control device 48 recognizes the kerf 21 by, for example, identifying a low-brightness portion as the kerf 21 from the image capturing the kerf 21 acquired in step S2. That is, step S3 is an example of a recognition step for recognizing the kerf 21 by brightness in the image captured in the imaging step of step S2. More specifically, the control device 48, for example, binarizes the image capturing the kerf 21 based on brightness to distinguish low-brightness regions and high-brightness regions at the pixel level, and identifies the kerf 21 by excluding regions outside the wafer 11 from the low-brightness regions.

[0029] When the kerf 21 is recognized, the processing device 2 counts the number of pixels that exceed a preset minimum width of the kerf (step S4). In step S4, the control device 48 sets a minimum kerf width and counts the number of pixels in the width direction of the kerf 21 that exceeds the preset minimum width from the kerf 21 (kerf) recognized in step S3 for each longitudinal position (position in the X-axis direction) of the kerf 21. The minimum kerf width is specified based on, for example, setting information of the processing device 2 stored in the memory unit 49 (for example, the type of cutting blade 40, the type of wafer 11, etc.).

[0030] For example, in the case where the machined groove 21 shown in Fig. 6 is imaged, the number of pixels in step S4 may be counted for the entire width of the machined groove 21, but in order to reduce the amount of calculation and shorten the processing time, it may be counted for only one side area of ​​the center line 211 of the machined groove 21 (area R in Fig. 6). Fig. 7 is a diagram showing only the results of the binarization process obtained by performing a binarization process based on brightness on an image of the machined groove 21 shown in Fig. 6, the results being for an area within area R shown in Fig. 6 that is at least half the kerf minimum width W away from the center line 211. When the number of pixels in the width direction of the machined groove 21 from the binarization process result 441 is counted for each position in the X direction, the following results are obtained. (0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 5, 0, 1, 0, 0, 0, 0, 3, 6, 3, 0, 1)

[0031] After counting the number of pixels, the processing device 2 images the entire one-line groove 21 formed in step S1, and determines whether the number of pixels has been counted at all positions in the longitudinal direction of the one-line groove 21 (step S5). If the number of pixels has not been counted at all positions in the one-line groove 21 (step S5 NO), the processing device 2 moves the field of view in the longitudinal direction and performs the processes of steps S2 to S5 again. The processing device 2 repeats the above processes, that is, the processes from the imaging step (step S2) to the counting step (step S4), multiple times until the number of pixels has been counted at all positions in the longitudinal direction (X-axis direction) of the one-line groove 21 (step S5 YES). This makes it possible to obtain information from the entire groove 21 even if the entire one-line groove 21 does not fit within the field of view of the imaging unit 44.

[0032] Once the number of pixels for each position in the longitudinal direction (X-axis direction) of the processed groove 21 is obtained, the processing device 2 performs settings for a kerf check (step S6). In step S6, the control device 48 calculates, based on the number of pixels obtained in step S4, what the statistical width of the processed groove 21 will be when the kerf is formed (cut) by the processing device 2. Then, based on the calculation result, the control device 48 sets the number of pixels corresponding to a width that occurs with a statistical probability that it can be determined to be abnormal as a threshold value for determining whether the formed kerf is good or bad.

[0033] More specifically, in step S6, the control device 48 assumes that the histogram of the number of pixels obtained in step S4 follows a predetermined normal distribution, that is, that the occurrence frequency of the width of the machined groove 21 formed by the processing device 2 follows a predetermined normal distribution, and estimates the mean value μ and standard deviation σ, which are parameters that specify the predetermined normal distribution, based on the number of pixels counted in step S4. Thereafter, the control device 48 calculates, from the mean value μ and standard deviation σ, the number of pixels that occur with a probability (e.g., about 4.5%, about 0.3%) that can be statistically determined to be abnormal in the predetermined normal distribution (e.g., μ + 2σ, μ + 3σ), and sets this as a threshold for determining whether the machined groove 21 (kerf) is good or bad.

[0034] In other words, the control device 48 sets a reference value based on the average value μ of the number of pixels obtained in step S4, and further sets a threshold value based on the standard deviation σ relative to the reference value, based on the number of pixels obtained in step S4. The reference value is, for example, the average value μ itself, and the threshold value is, for example, the average value μ plus twice the standard deviation σ (μ + 2σ).

[0035] In this example, the number of pixels is obtained by counting the number of pixels that exceeds the preset minimum kerf width W in step S4. Therefore, the number of pixels is limited to a lower limit value (0). Taking this into consideration, in step S6, a truncated normal distribution B consisting of a portion equal to or greater than the average value μ as shown in FIG. 8 may be assumed as the normal distribution that the histogram of the number of pixels follows. Note that the assumed normal distribution is not limited to the truncated normal distribution B shown in FIG. 8, but may also be, for example, a truncated normal distribution consisting of a portion equal to or greater than the number of pixels corresponding to the minimum kerf width W (i.e., 0).

[0036] Once the threshold value is set, the processing device 2 judges whether the processed groove 21 is good or bad (step S7). In step S7, the control device 48 compares the number of pixels counted in step S4 with the threshold value set in step S6 to judge whether the processed groove 21 formed in step S1 is good or bad. Specifically, the control device 48 judges whether the number of pixels counted at each longitudinal position in step S4 is below or exceeds the threshold value. For example, if the number of pixels at any position in the processed groove 21 is below the threshold value, the formed processed groove 21 is judged to be good, and if the number of pixels at any position in the processed groove 21 exceeds the threshold value, the formed processed groove 21 is judged to be bad.

[0037] After determining whether the processing groove 21 is good or bad, the processing device 2 determines whether or not processing of the wafer 11 should be stopped (step S8). If it is determined in step S7 that the processing groove 21 is bad, the processing device 2 determines in step S8 that processing of the wafer 11 should be stopped (step S8YES), then notifies the user of the error (step S9), ends the processing of Fig. 5, and stops processing of the wafer 11. On the other hand, if it is determined in step S7 that the processing groove 21 is good, the processing device 2 determines in step S8 that processing of the wafer 11 should not be stopped (step S8NO), then determines whether cutting (kerf formation) has been completed (step S10).

[0038] In step S9, if cutting has not been completed for all planned division lines 13, the processing device 2 determines not to end cutting (step S10 NO), and repeats the processing from step S1 onwards for the next planned division line 13. If cutting has been completed for all planned division lines 13, it determines to end cutting (step S10 YES), and ends the processing in Fig. 5.

[0039] According to the kerf check method used in the processing shown in Fig. 5 performed by the processing device 2, the threshold value used to determine the acceptability of the processed portion in the kerf check can be automatically set. This reduces the burden of the setting work required for the kerf check compared to when the threshold value is set manually by an operator. Furthermore, since the processing is performed without waiting for the operator to complete the setting, waiting time is reduced and high throughput can be achieved.

[0040] Furthermore, this kerf check method can avoid the drawbacks that may arise from setting a threshold value manually by an operator, thereby supporting the implementation of an appropriate kerf check. When an operator manually sets a threshold value while viewing an image of the machined groove 21, the threshold value is likely to be set based on the size of chippings that appear in the image that the operator is viewing at the time of setting the threshold value. In other words, the threshold value is set based on the size of chippings in the image, without sufficient consideration of whether the chippings in the image actually occur rarely or frequently in normal machining. A threshold value set in this manner is not necessarily appropriate as a criterion for determining machining abnormalities.

[0041] In contrast, the above-described kerf check method allows a threshold value to be set that can be statistically determined to be abnormal based on the number of pixels counted at each longitudinal position of the machined groove 21. Therefore, compared to manual setting, there is less chance involved, making it possible to set a more appropriate threshold value. This allows for accurate detection of defects in the machined groove 21, resulting in a high-quality processed product. Furthermore, it is possible to avoid stopping the processing due to a good machined groove 21 being mistaken for a bad machined groove 21, thereby reducing unnecessary downtime. Therefore, the above-described kerf check method achieves both high quality and high throughput. Furthermore, the improved accuracy of error reporting also reduces unnecessary work by operators caused by erroneous error reporting.

[0042] Furthermore, even if an operator were to statistically evaluate the size of chipping identified from an image, such a method would result in an insufficient number of samples (population size) in the case of a machined groove 21 with good machining quality and little chipping, making it difficult to obtain sufficiently reliable results.

[0043] In contrast, the above-described kerf check method does not statistically analyze only chipping, but also the entire machined groove 21, including the portion where chipping does not occur. This makes it possible to extract more information to be used for statistical analysis from the image, ensuring a sufficient number of samples (population parameters) for statistical analysis, thereby obtaining analysis results that more appropriately represent the machining quality of the machined groove 21, and ultimately a threshold value. Therefore, the kerf check can be performed more appropriately.

[0044] 5, the setting process in step S5 is performed for each line, and the threshold setting is updated. When updating the threshold in step S5, the number of pixels counted for the previously cut planned division line 13 may be used together with the number of newly counted pixels. This increases the number of samples used in statistical processing as cutting progresses, and therefore, the more cutting progresses, the more reliable the threshold value to determine pass / fail.

[0045] The embodiments of the present invention are not limited to the above-described embodiments, and may be variously modified, substituted, or altered without departing from the spirit and scope of the technical idea of ​​the present invention. Furthermore, if the technical idea of ​​the present invention can be realized in a different way due to technological advances or other derived technologies, it may be implemented using that method. Therefore, the claims cover all embodiments that may fall within the scope of the technical idea of ​​the present invention.

[0046] In the above-described embodiment, an example has been shown in which a kerf check is performed each time the wafer 11 is cut along one planned dividing line 13, but the frequency and timing of the kerf check are not limited to this example. For example, each time cutting is performed along multiple planned dividing lines 13, images of multiple cut grooves 21 may be taken, the number of pixels as described above may be counted, and a threshold value may be set by statistical processing. Also, instead of setting a threshold value each time the wafer 11 is cut, a threshold value may be set for each wafer 11. For example, a threshold value may be set when the wafer 11 is cut for the first time, and the same threshold value may be used for subsequent kerf checks.

[0047] In the above-described embodiment, an example has been shown in which the machining groove 21 is recognized based on brightness from an image of the machining groove 21, but the information used for image recognition of the machining groove 21 is not limited to brightness. Any information that allows the machining groove 21 to be appropriately recognized may be used, and for example, the machining groove 21 may be recognized based on the brightness of the image or any pixel value of the image. Furthermore, the machining groove 21 may be recognized using any image recognition technology.

[0048] In the above-described embodiment, an example has been shown in which the threshold value is set by counting and statistically processing the number of pixels in the width direction that exceed the minimum width of the groove 21 (i.e., corresponding to the chipping size), but the subject of statistical processing is not limited to the number of pixels in the width direction that exceed the minimum width of the groove 21. For example, the threshold value may be set by counting and statistically processing the number of pixels that correspond to the width of the groove 21 itself.

[0049] In the above-described embodiment, an example was shown in which an average value was used as the reference value that serves as the basis for setting a threshold value, but the reference value is not limited to the average value. Other statistical values, such as a median or a mode, may also be used as the reference value. For example, the median may be used as the reference value, and a value obtained by adding twice the standard deviation to the median may be used as the threshold value. Furthermore, in the above-described embodiment, an example was shown in which the threshold value was set to a value obtained by adding twice the standard deviation to the reference value, but the threshold value may be determined from the reference value and the standard deviation. For example, the threshold value may be set to a value obtained by adding three or more times the standard deviation to the reference value.

[0050] In the above-described embodiment, the number of pixels was obtained from the entire processing groove 21 of one line formed on the planned division line 13, but as long as it is possible to obtain the number of pixels at a sufficient number of different positions to statistically determine the threshold value, the number of pixels may be obtained from only a part of the processing groove 21 of one line. [Industrial Applicability]

[0051] As described above, the kerf check method described above can reduce the burden of the setting work required for kerf check, and is therefore useful in processing devices that form processed grooves (kerfs), and is extremely useful in any processing devices that form processed grooves, such as cutting devices and laser processing devices. [Explanation of symbols]

[0052] 2: Processing equipment 11: Wafer 13: Planned division line 14: Chuck table 15: Device 16: Clamp 18: Cutting unit 21: Machining groove 22: Cutting unit movement mechanism 40: Cutting blade 44: Imaging unit 48: Control device 49: Storage section 50: Input device 52: Housing 54: Microscope unit 56: Oblique lighting unit 58: Half mirror 60, 66: Light source 62: Objective lens 64: Image sensor 111 :Surface 112: Back side 141: Holding surface 211: Center line 441: Binarization result 442: Histogram B: Truncated normal distribution L1, L2: light R: area W: Minimum kerf width μ: Average value σ: Standard deviation

Claims

[Claim 1] A kerf check method for a kerf formed in a wafer, comprising: an imaging step of imaging the kerf with a camera; a recognition step of recognizing the kerf based on brightness in the image captured in the imaging step; a counting step of counting pixels that exceed a preset minimum kerf width from the recognized kerf for each longitudinal position of the kerf as the number of pixels in the width direction of the kerf; a reference value setting step of performing the steps from the imaging step to the counting step multiple times and setting a reference value based on the average, median, or mode of the number of pixels counted in the counting step; a threshold value setting step for setting a threshold value for determining whether the kerf is good or bad based on the number of pixels counted in the counting step and a standard deviation with respect to the reference value; a determining step of determining whether the kerf is good or bad based on whether the number of pixels is equal to or less than the threshold value or exceeds the threshold value; A kerf check method including:

Images