Workpiece inspection method and device

Abstract

To provide a workpiece inspection method and device capable of detecting a defect, generated by scattering etc., of laser light, precisely at low cost.SOLUTION: A workpiece inspection method comprises: an imaging process of using an imaging device (306) to capture an image of a workpiece (CW) within an inspection range after laser processing while obliquely lighting the workpiece; an image processing process of subjecting the image of the workpiece within the inspection range after the laser processing to fast Fourier transformation to extract high-frequency components, and then performing binarization processing; and a detecting process of detecting a processing-time defect, formed with scattered light or leakage light in the laser processing, from the image of the workpiece within the inspection range after the laser processing which has been processed through the image processing process.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a workpiece inspection method and apparatus and a workpiece processing method, and more particularly to a technique for cleaving a workpiece having a laser processing region inside the workpiece. [Background technology]

[0002] Conventionally, laser dicing devices are known that align a focal point inside a workpiece such as a silicon wafer and irradiate the workpiece with a laser beam along a processing line, forming a laser processing area within the workpiece along the processing line that will serve as the starting point for cutting. The workpiece with the laser processing area formed therein is then cleaved along the planned cutting line by a cleaving process such as expanding or breaking, and divided into individual chips. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-064049 Summary of the Invention [Problem to be solved by the invention]

[0004] When forming a laser processing area inside a workpiece, some of the laser light may be scattered in a spatter-like manner by structures inside the workpiece (for example, defects (e.g., scratches, voids, etc.), particles such as impurities, processing marks, cracks inside the workpiece caused by laser processing, etc.), or may leak out to the side opposite the incident surface of the laser light. Such scattered light can cause defects (e.g., point-like defects, hereinafter referred to as splash damage) inside the workpiece (see, for example, Patent Document 1). Splash damage can damage the workpiece and devices formed on the surface of the workpiece, reducing the quality of the devices, so there is a need to suppress its occurrence.

[0005] The present invention has been made in consideration of the above circumstances, and aims to provide a workpiece inspection method and apparatus, as well as a workpiece processing method, that are capable of accurately detecting defects caused by scattering of laser light, etc. [Means for solving the problem]

[0006] In order to solve the above problems, a workpiece inspection method according to a first aspect of the present invention comprises an imaging step of using an imaging device to capture an image of an inspection area of ​​a workpiece after laser processing while illuminating the workpiece obliquely; an image processing step of applying a fast Fourier transform to the image of the inspection area of ​​the workpiece after laser processing and performing binarization processing; a detection step of detecting processing defects formed by scattered light or leaked light generated during laser processing from the image of the inspection area of ​​the workpiece after laser processing that has been image-processed in the image processing step; and a setting step of setting processing conditions in accordance with the detection results of the processing defects.

[0007] A workpiece inspection method according to a second aspect of the present invention is the same as the first aspect, but uses a workpiece consisting of only a workpiece layer.

[0008] A workpiece inspection method according to a third aspect of the present invention is the first or second aspect, and comprises the steps of: capturing an image of an inspection area of ​​the workpiece before laser processing using an imaging device while illuminating the workpiece obliquely; and performing a fast Fourier transform on the image of the inspection area of ​​the workpiece before laser processing and performing a binarization process; and the detection step comprises the steps of detecting a defect in the inspection area of ​​the workpiece from the image of the inspection area of ​​the workpiece before laser processing; detecting defect candidates from the defects detected in the image of the inspection area of ​​the workpiece after laser processing; and identifying processing defects by excluding the defect, the laser processed area formed in the workpiece by laser processing, and the shadow of the laser processed area from the defect candidates.

[0009] A workpiece processing method according to a fourth aspect of the present invention performs laser processing of a workpiece to be processed based on processing conditions set by the workpiece inspection method according to any one of the first to third aspects.

[0010] A workpiece inspection device according to a fifth aspect of the present invention comprises an imaging device that captures an image of an inspection area of ​​a workpiece after laser processing while illuminating the workpiece obliquely; an image processing unit that performs a fast Fourier transform on the image of the inspection area of ​​the workpiece after laser processing and performs binarization processing; a detection unit that detects processing defects formed by scattered light or leaked light generated during laser processing from the image of the inspection area of ​​the workpiece after laser processing that has been image-processed by the image processing unit; and a setting unit that sets processing conditions according to the detection results of the processing defects. [Effects of the Invention]

[0011] According to the present invention, by applying a fast Fourier transform to images of the inspection area of ​​the workpiece before and after laser processing and performing binarization processing, it is possible to accurately detect minute defects caused by scattering of laser light, etc. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram showing the general configuration of a laser dicing device. [Figure 2] FIG. 2 is a plan view and a cross-sectional view of a workpiece to be inspected when oblique illumination (dark field illumination system) is used. [Figure 3] FIG. 3 is a plan view and a cross-sectional view of a workpiece to be inspected when using coaxial epi-illumination (bright field illumination system). [Figure 4] Figure 4 shows an image of the workpiece under inspection captured using oblique illumination (dark-field illumination system). [Figure 5] Figure 5 shows an image of the workpiece for inspection captured using coaxial epi-illumination (bright-field illumination system). [Figure 6] FIG. 6 is an image obtained by capturing an image of the workpiece CW for inspection. [Figure 7] FIG. 7 is an image showing the result of binarization processing performed on the original image of FIG. [Figure 8] FIG. 8 is an image showing the result of performing dynamic thresholding on the original image of FIG. [Figure 9]FIG. 9 is an image showing the result of dynamic thresholding of the original image of FIG. [Figure 10] FIG. 10 shows a captured image of the workpiece under inspection after FFT processing. [Figure 11] FIG. 11 is an image showing the extraction result of splash damage. [Figure 12] FIG. 12 is a flowchart showing a method for setting processing conditions for laser dicing. [Figure 13] FIG. 13 is a flowchart showing the process of detecting defects during processing. [Figure 14] FIG. 14 is a flowchart showing the process of evaluating defects during processing. [Figure 15] FIG. 15 is a flowchart showing the process of setting processing conditions. DETAILED DESCRIPTION OF THE INVENTION

[0013] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of a workpiece inspection method and apparatus, and a workpiece machining method according to the present invention will be described with reference to the accompanying drawings.

[0014] In this embodiment, laser processing of an inspection workpiece CW is performed using a laser dicing device 10. Then, images of the inspection range of the inspection workpiece CW before and after processing (for example, the inside, surface (top and bottom surfaces) of the inspection workpiece CW, etc.) are used to distinguish and detect missing parts that existed before the laser processing from processing defects including splash damage that occur due to the laser processing.

[0015] Next, the machining conditions are evaluated based on the detection results of machining defects to determine whether or not a change in the machining conditions is necessary. If it is determined that a change in the machining conditions is necessary, the machining conditions are reset. This makes it possible to appropriately set the machining conditions for the workpiece W to be produced.

[0016] (Laser dicing equipment) First, the laser dicing apparatus 10 will be described with reference to Fig. 1. Fig. 1 is a diagram showing the overall configuration of the laser dicing apparatus 10.

[0017] As shown in the figure, the laser dicing apparatus 10 of this embodiment includes a workpiece moving section 100, an optical system unit (laser engine) 400, a control section 12, etc. The optical system unit 400 includes a laser optical section 200 and an observation optical section 300.

[0018] The control unit 12 includes a CPU (Central Processing Unit), a memory, an input / output circuit unit, etc., and controls the operation of each unit of the laser dicing apparatus 10.

[0019] The workpiece moving unit 100 includes a suction stage 102 that suction-holds the inspection workpiece CW and the workpiece W to be produced (hereinafter, these may be collectively referred to as workpiece W), and an XYZθ table 104 that is provided on a main body base 106 of the laser dicing device 10 and that precisely moves the suction stage 102 in the XYZθ directions. The workpiece W is precisely moved in the XYZθ directions in the drawing by this workpiece moving unit 100.

[0020] The workpiece W has a backgrind (BG) tape B having an adhesive attached to one surface thereof, and is placed on the suction stage 102 with the back surface facing upward.

[0021] The workpiece W may have a dicing sheet having an adhesive attached to one surface thereof, and may be placed on the suction stage 102 in a state where it is integrated with the frame via this dicing sheet. In this case, the workpiece W is placed on the suction stage 102 with its front surface facing upward. In the following description, the surface of the workpiece W exposed to the optical system unit 400 side is referred to as the upper surface, and the surface to which the BG tape B is attached is referred to as the lower surface. The workpiece W may also be directly suctioned to the suction stage 102.

[0022] The laser optical unit 200 includes a laser oscillator 202, a collimating lens 204, a half mirror 206, a condensing lens (light-collecting lens) 208, and a driving means 210 that moves the laser light slightly in the Z direction relative to the workpiece W. As a light source for the laser oscillator 202, for example, a semiconductor laser pumped Nd:YAG (Yttrium Aluminum Garnet) laser is used. The laser light emitted from the laser oscillator 202 passes through an optical system including the collimating lens 204, the half mirror 206, and the condensing lens 208, and is then focused onto the inspection range of the workpiece W. The Z-direction position of the focal point is adjusted by slightly moving the condensing lens 208 in the Z direction.

[0023] The observation optical unit 300 includes an observation light source 302, a condenser lens 304, an imaging device (camera, such as an IR (infrared) sensor or a CCD (Charge Coupled Device) camera) 306, an image processing unit 308, and a monitor 310. The control unit 12 and the observation optical unit 300 are an example of a work inspection device.

[0024] The observation light source 302 irradiates the workpiece W with approximately annular (ring-shaped) oblique illumination light (oblique illumination light) L1 from a position offset from the optical axis AX of the condensing lens 208. The observation light source 302 is a ring illumination unit having multiple lighting units arranged in a circular ring around the condensing lens 208. Here, the multiple lighting units may be, for example, LEDs (Light Emitting Diodes) or directional bullet-shaped LEDs.

[0025] In this embodiment, a ring illumination unit is used as the observation light source 302 for oblique illumination, but the present invention is not limited to this. For example, a ring illumination unit may be arranged around the workpiece W to irradiate the workpiece W with illumination light that is incident obliquely, or the workpiece W may be obliquely illuminated with light converted into a ring shape using a conical lens or a light shielding plate having an annular opening.

[0026] When the workpiece W is obliquely illuminated using the observation light source 302, the light scattered by the workpiece W passes through the condenser lenses 208 and 304 and enters the camera 306, which serves as observation means, and an image of the workpiece W is captured. This captured image data is input to the image processing unit 308 and used for aligning the workpiece W. In addition, this captured image data is displayed on the monitor 310 via the control unit 12.

[0027] The observation optical unit 300 is capable of capturing images of defects, including splash damage, within the inspection range of the workpiece W. In this case, oblique illumination light L1 has a wavelength longer than that of visible light. If the workpiece W is a silicon wafer, for example, infrared light is used. The oblique illumination light L1 irradiated onto the workpiece W passes through the inspection range of the workpiece W and is scattered by splash damage (see FIG. 2) on the underside of the workpiece W. This scattered light passes through condenser lenses 208 and 304 and enters camera 306, which captures an image including defects within the inspection range of the workpiece W. In this captured image, areas other than the area where defects, including splash damage, are formed (hereinafter referred to as background areas) are dark, and defects including splash damage appear as white areas (dark-field illumination system). The dark-field illumination system will be described later with reference to FIGS. 2 to 5.

[0028] This image data is input to an image processing unit 308 and used to detect defects in the inspection range of the workpiece W. In addition, the image of the inspection range of the workpiece W is displayed on a monitor 310 together with the defect detection results. This allows the operator to check and correct the defect detection results while looking at the monitor 310.

[0029] In this embodiment, images including defects in the inspection area of ​​the workpiece W are captured before and after processing, which uses laser light to form a laser processing area. This makes it possible to distinguish and detect missing parts that existed before laser processing from processing defects, including splash damage caused by laser processing.

[0030] In this embodiment, the image of the workpiece W is captured using the laser optical unit 200 and the observation optical unit 300, but the present invention is not limited to this. For example, an image of the inspection range of the workpiece W may be captured using the optical system and camera of a microscope installed for aligning the workpiece W.

[0031] When forming a laser processing area on the workpiece W, a laser beam L is emitted from a laser oscillator 202, and the laser beam L passes through an optical system including a collimator lens 204, a half mirror 206, and a condenser lens 208, and is then irradiated onto an inspection area of ​​the workpiece W. The Z-direction position of the focal point of the irradiated laser beam L is accurately set to a predetermined position within the inspection area of ​​the workpiece W by adjusting the Z-direction position of the workpiece W using the XYZθ table 104 and by controlling the position of the condenser lens 208.

[0032] In this state, the XYZθ table 104 is fed for processing in the X direction, which is the dicing direction. As a result, one line of laser processing area R1 is formed in the inspection range of the workpiece W along the processing line CL1 of the workpiece W (see FIG. 2). Then, once one line of laser processing area R1 has been formed along the processing line CL1, the XYZθ table 104 is indexed and fed one pitch in the Y direction, and a laser processing area R1 is also formed on the next processing line. Next, once the laser processing area R1 has been formed along all of the X-direction processing lines CL1, the XYZθ table 104 is rotated 90° around the Z axis, and laser processing areas R1 are similarly formed on the X-direction processing lines CL1 after the rotation.

[0033] Here, the laser processing region R1 refers to a region where the physical properties of the workpiece W, such as density, refractive index, and mechanical strength, become different from those of the surrounding area due to irradiation with laser light, resulting in a lower strength than the surrounding area. The laser processing region R1 includes, for example, a crack region. The workpiece W with the laser processing region formed therein is transported to a grinding device (not shown), where the back surface of the workpiece W is ground to remove the laser processing region R1. Then, when expanding tape is attached to the back surface of the workpiece W and expanded, the workpiece W is fractured by the cracks that extend from the laser processing region R1 to the front surface of the workpiece W. This separates the workpiece W into individual chips.

[0034] (Oblique illumination (dark field illumination system)) Next, the workpiece CW for inspection will be described with reference to Fig. 2. Fig. 2 shows a plan view and a cross-sectional view of the workpiece for inspection when oblique illumination (dark-field illumination system) is used. Fig. 2(A) is a plan view (top view) of the workpiece for inspection, and Fig. 2(B) is a cross-sectional view taken along line BB of Fig. 2(A).

[0035] As shown in FIG. 2(B), the inspection workpiece CW according to this embodiment includes a workpiece layer WL made of the same material (for example, silicon) as the workpiece W to be produced.

[0036] The inspection workpiece CW is adsorbed and held with its front surface facing downward in the laser dicing device 10. The inspection workpiece CW shown in FIG. 2 includes a defect portion P1 containing impurities (particles). Note that, to simplify the drawing, only one defect portion P1 is shown in FIG. 2. In this embodiment, an observation light source 302 is used to obliquely illuminate the inspection workpiece CW with oblique illumination light L1 that passes through the inspection workpiece CW, and the camera 306 is used to capture an image of the inspection workpiece CW before processing. The control unit 12 then detects the defect portion P1 from this image.

[0037] When laser processing is performed on the inspection workpiece CW, a laser processing region R1 (see FIG. 2(B)) is formed along the processing line CL1 (see FIG. 2(A)) in the inspection range of the workpiece layer WL.

[0038] During laser processing, some of the laser light is scattered in the inspection area of ​​the work layer WL or leaks out to the underside of the work layer WL. When this scattered light and leaked light reach the underside of the work layer WL, the detection layer DL is partially melted. This changes the reflectance of the interface between the work layer WL and the detection layer DL when illumination light (oblique illumination light (polarized illumination light), for example, infrared light) is irradiated onto the inspection workpiece CW. Based on this change in reflectance, defects can be detected from the image of the inspection workpiece CW after processing.

[0039] The example shown in Figure 2 shows splash damages SD1 to SD3 formed by scattered light and damage BD (hereinafter referred to as direct damage) formed by leaked light leaking to the underside of the laser processing area R1. Splash damages SD1 to SD3 are point-like defects formed when part of the laser light is scattered in a spatter-like manner by structures in the inspection range of the work layer WL. Direct damage BD is a defect formed when part of the laser light leaks to the underside of the laser processing area R1, and is formed along the processing line CL1.

[0040] In this embodiment, the inspection workpiece CW is obliquely illuminated with oblique illumination light L1, and the interface between the workpiece layer WL and the detection layer DL is used as the focus plane FP (in-focus plane) using a camera 306, and the inspection workpiece CW after processing is imaged. As shown in Figure 2, the image of the inspection workpiece CW after processing includes the missing part P1 and the direct damage BD in addition to the splash damages SD1 to SD3.

[0041] In these captured images, most of the illumination light L1 that passes through the background region does not return to the camera 306, and the illumination light L1 that is irradiated onto the defect including splash damage is scattered, with part of the scattered light L2 reaching the camera 306. For this reason, the background region is darker than the defect including splash damage, and the defect including splash damage appears as a white region.

[0042] As will be described later, when oblique illumination is performed, a portion of the illumination light L1 that passes through the underside of the inspection workpiece CW is reflected by the porous (e.g., granular ceramic dense) surface of the suction stage (chuck) 102 and returns to the camera 306. This causes uneven brightness in the background area of ​​the captured image (see FIG. 6, etc.). In this embodiment, while the light reflected from the porous surface of the suction stage 102 is blurred, minute defects such as splash damage SD1 to SD3, missing portion P1, and direct damage BD have high-frequency components, which are utilized to extract the defects. Specifically, defects are extracted by using a fast Fourier transform (FFT) to extract only high-frequency components from the captured image of the inspection workpiece CW.

[0043] That is, the control unit 12 uses the image processing unit 308 to apply FFT to the image of the inspection workpiece CW before processing to detect the defect P1 and save information about its position and shape (e.g., coordinates, etc.). Next, the control unit 12 uses the image processing unit 308 to apply FFT to the image of the inspection workpiece CW after processing to detect all defects as candidates for processing defects (hereinafter referred to as defect candidates), including splash damage SD1 to SD3. These defect candidates may include the defect P1 and direct damage BD in addition to the splash damage SD1 to SD3. Then, the control unit 12 excludes the previously detected defect P1 from the defect candidates. Furthermore, the control unit 12 extracts direct damage BD along the processing line CL1 from the image of the inspection workpiece CW after processing and excludes it from the defect candidates. As a result, the splash damage SD1 to SD3 are extracted as processing defects.

[0044] According to this embodiment, by applying FFT to images of the inspection range of the workpiece before and after laser processing, minute defects caused by scattering of laser light or the like can be detected with high accuracy.

[0045] It should be noted that with regard to the defect P1, the area including the previously detected defect P1 may be masked to exclude it from the detection targets for defect candidates. Here, the mask area is an area defined by the outline of the defect P1 to which a margin for repeatability accuracy such as transport error of the inspection workpiece CW is added. By defining the mask area in this way, the mask area can be minimized, thereby maximizing the area that is the target for detection of machining defects and preventing oversight of machining defects.

[0046] Furthermore, it is also possible to generate a pattern on the detection layer DL of the workpiece CW for inspection and detect this pattern using an alignment camera or the like to align the workpiece CW for inspection, thereby correcting the transport error. In this case, it is possible to further reduce the transport error of the workpiece CW for inspection, and therefore the mask area can be further reduced. This allows the area to be subject to detection of machining defects to be larger, making it possible to more reliably prevent undetected machining defects.

[0047] In this embodiment, the inspection workpiece CW made of the workpiece layer WL is used, but the present invention is not limited to this. It is also possible to use the inspection workpiece CW1 having the detection layer DL shown in FIG. 3 formed thereon.

[0048] (Coaxial epi-illumination (bright-field illumination system)) In this embodiment, oblique illumination (dark-field illumination system) is used, but the present invention can also be applied when coaxial epi-illumination (bright-field illumination system) is used. When coaxial epi-illumination is used, an inspection workpiece CW1 having a detection layer DL shown in Figure 3 is used.

[0049] FIG. 3 is a plan view and a cross-sectional view of a workpiece to be inspected when using coaxial epi-illumination (bright field illumination system).

[0050] As shown in FIG. 3, the inspection workpiece CW1 includes a workpiece layer WL and a detection layer DL formed on the lower surface of the workpiece layer WL in the figure (the surface (front surface) side to be inspected). The workpiece layer WL is formed of, for example, silicon or the like. The detection layer DL is formed by depositing a low-melting-point metal (for example, tin) or a low-melting-point alloy on the workpiece layer WL. The detection layer DL may also be formed of a coating of resin or the like. Furthermore, the inspection workpiece CW1 is not limited to one formed with a detection layer DL, and for example, one in which the lower surface of the workpiece WL is mirror-finished or the like to increase reflectivity (for example, a mirror workpiece) may be used.

[0051] The inspection workpiece CW1 is adsorbed and held with the detection layer DL facing downward in a laser dicing device (not shown). Laser processing is then performed on the inspection workpiece CW1, and a laser processing region R1 is formed inside the workpiece layer WL. Note that the symbol R2 in FIG. 3 is a shadow created when part of the illumination light L10 is blocked by the laser processing region R1. This shadow R2 is imaged along the processing line CL1.

[0052] In the region where splash damages SD1 to SD3 are formed, the reflectance of the interface FP between the work layer WL and the detection layer DL changes when the inspection workpiece CW1 is irradiated with highly transmittant illumination light L1 (infrared light if the inspection workpiece CW1 is a silicon wafer). Therefore, splash damages SD1 to SD3 can be detected by performing coaxial epi-illumination using illumination light L10 and capturing an image of reflected light L12 from the interface FP using camera 306.

[0053] In the example shown in Figure 3, a process for forming a low-melting-point metal vapor deposition film is required, but the sensitivity to scattered light and leaked light during laser processing can be increased, making it easy to detect splash damage SD1 to SD3. Based on the detection results of splash damage on the inspection workpiece CW1, laser processing conditions that can suppress the occurrence of splash damage can be obtained. This makes it possible to set processing conditions so that devices formed on the surface of the workpiece are not affected by splash damage when laser processing is performed on the workpiece to be produced.

[0054] (Images captured with dark-field and bright-field illumination systems) Next, we will explain the images of the workpiece for inspection captured using the dark-field illumination system and the bright-field illumination system. Figure 4 shows an image of the workpiece for inspection captured using oblique illumination (dark-field illumination system), and Figure 5 shows an image of the workpiece for inspection captured using coaxial epi-illumination (bright-field illumination system).

[0055] The example shown in Figure 5 is an image obtained by coaxially illuminating an inspection workpiece CW1 (see Figure 3) with a detection layer DL. In this case, in the background area except for the area where the defect, including splash damage (circled area), is formed, the illumination light L1 is reflected by the detection layer DL, which is made of a metal-evaporated film. On the other hand, in the area where the defect, including splash damage (circled area), is formed, the illumination light L10 is diffused by the splash damage SD1 to SD3, so the reflected light L12 does not return to the incident side of the illumination light L10. Therefore, in Figure 5, the background area emits white light, and the defect, including splash damage (circled area), is a dark area (bright-field illumination system). Because splash damage is a minute defect, if the amount of reflected light around the defect is strong, the contrast between the background area and the defect will be reduced due to charge leakage from pixels surrounding the pixel corresponding to the defect on the sensor or lens aberration. For this reason, splash damage is likely to be missed when using a bright-field illumination system.

[0056] On the other hand, in the example shown in Figure 4, the background area is dark, and the defect (circled area), including splash damage, is a white area. The white light emitted from the defect indicates that the area where the defect formed has a locally uneven shape on the underside of the inspection workpiece CW. When dark-field illumination is used, the uneven shape generates reflected light, causing the area where the defect formed to actively shine. As a result, the area where the defect formed is captured enlarged without being affected by leakage charge or lens aberration. Therefore, when a dark-field illumination system is used, the effect of highlighting the defect is obtained, resulting in improved robustness.

[0057] (Splash damage extraction) Next, a procedure for extracting splash damage from a captured image obtained by capturing an image of the inspection workpiece CW will be specifically described.

[0058] Fig. 6 is an image (hereinafter also referred to as the original image) obtained by imaging the inspection workpiece CW, and Fig. 7 is an image showing the result of performing binarization processing on the original image of Fig. 6. The original image of Fig. 6 includes splash damage SD1 to SD4.

[0059] When oblique illumination is performed, part of the illumination light L1 that passes through the underside of the workpiece CW for inspection is reflected by the porous surface of the suction stage (chuck) 102 and returns to the camera 306. This causes uneven brightness in the background area of ​​the captured image, as shown in FIG.

[0060] When the original image in Figure 6 is binarized using a constant threshold, the image shown in Figure 7 is obtained. As shown in Figure 7, splash damages SD2 and SD3 cannot be extracted because they are included in the high-brightness area N1 in the background area.

[0061] Furthermore, noise occurs in the peripheral area N2 of the high-brightness area N1 due to variations in the gray levels of the pixels. The influence of this noise makes it difficult to separate splash damage SD4 located in the peripheral area N2 of the high-brightness area N1 from the background area.

[0062] Therefore, it is conceivable to extract defects by performing dynamic threshold processing on each local region of the captured image. In dynamic threshold processing, for example, a local averaging filter or a local median filter is first used to generate an average value image (= threshold image). Next, this average value image is used as the threshold to perform binarization processing (dynamic threshold processing) that takes brightness unevenness into account. This adjusts the threshold to account for noise caused by minute brightness fluctuations in pixels.

[0063] 8 and 9 are images showing the results of dynamic thresholding of the original image of FIG. FIG. 8 shows an example (Comparative Example 1) in which an appropriate threshold value could be set by dynamic threshold processing, and FIG. 9 shows an example (Comparative Example 2) in which an appropriate threshold value could not be set by dynamic threshold processing. As shown in FIG. 8, when an appropriate threshold value can be set, it is possible to extract splash damage SD1 to SD4 from the background. On the other hand, as shown in FIG. 9, when an appropriate threshold value cannot be set, a large number of noises caused by uneven brightness are extracted as defect candidates. These noises are only slightly different from the splash damages SD1 to SD4 in terms of characteristics such as shape and size, making it difficult to separate the noises from the splash damages SD1 to SD4.

[0064] As shown in Figures 8 and 9, dynamic thresholding is not robust to changes in visibility, and the processing results in production equipment are not stable. As a result, errors may occur frequently depending on the results of dynamic thresholding, causing the production process to repeatedly stop temporarily (so-called short stops). Furthermore, when noise is removed using a general spatial filter, splash damage SD1 to SD4 may also be removed at the same time.

[0065] Therefore, in this embodiment, the following method is used to improve robustness against changes in visibility. That is, while the reflected light from the porous surface of the suction stage 102, which causes noise, is blurred, the defects to be extracted (including the missing part P1 and splash damage SD1 to SD4) have the characteristics of high-frequency components, and this is utilized to extract the defects. Specifically, defects are extracted by using FFT to extract only high-frequency components from the captured image of the inspection workpiece CW.

[0066] FIG. 10 is a captured image of the workpiece for inspection after FFT processing, and FIG. 11 is an image showing the extraction result of splash damage.

[0067] Applying FFT to the original image in Figure 6 yields a two-dimensional spatial frequency characteristic map of the original image. Here, spatial frequency refers to the degree of change in brightness per unit length, expressed as a frequency. The minute defects to be extracted have a short period of change in brightness per unit length, resulting in a high spatial frequency. As a result, as shown in Figure 10, only the high-frequency region containing the defect to be extracted is extracted, and noise in the background region caused by uneven brightness is removed. This has the effect of highlighting only the defect to be extracted in the original image.

[0068] Next, the position, shape, and size of the defect to be extracted are identified by performing binarization on the image after FFT processing in Figure 10. The position, shape, and size of this defect are overlaid on the original image to obtain the image shown in Figure 11. This makes it possible to extract only splash damage SD1 to SD4.

[0069] (How to set processing conditions) Next, a method for setting processing conditions according to this embodiment will be described with reference to Figures 12 to 15. Figure 12 is a flowchart showing a method for setting processing conditions for laser dicing.

[0070] First, the inspection workpiece CW is carried into the laser dicing apparatus 10 (processing section) and is adsorbed and held on the suction stage 102 (step S10). Then, the observation optical section 300 irradiates the inspection workpiece CW with oblique illumination light L1, and the camera 306 captures an image of the inspection workpiece CW before processing (step S12: imaging process). The inspection workpiece CW may be adsorbed and held on the suction stage 102 via the BG tape B, or may be adsorbed directly to the suction stage 102.

[0071] Next, the inspection workpiece CW is processed by the laser dicing device 10, and a laser processing region R1 is formed in the inspection range of the inspection workpiece CW (step S14). Then, the observation optical unit 300 irradiates the inspection workpiece CW with oblique illumination light L1, and the processed inspection workpiece CW is imaged by the camera 306 (step S16: imaging step).

[0072] Next, the control unit 12 compares the images of the inspection workpiece CW before and after processing to detect machining defects, including splash damage formed by laser processing (step S18). Here, the control unit 12 and the image processing unit 308 function as the detection unit of the present invention. Then, the control unit 12 evaluates the detected machining defects in accordance with the detection results of the machining defects (step S20). The control unit 12 determines whether or not to change the machining conditions based on the evaluation results of the machining defects.

[0073] If the processing conditions are to be reset (Yes in step S22), the control unit 12 resets the processing conditions (step S24). On the other hand, if the processing conditions are not to be reset (No in step S22), the control unit 12 saves processing defect information including the evaluation results of the processing defects and images of the inspection workpiece CW before and after processing (step S26). Here, the control unit 12 functions as the setting unit of the present invention. The processing defect information saved in step S26 can be viewed by the operator as a quality record.

[0074] Next, the process of detecting defects during machining will be described with reference to a flowchart shown in FIG.

[0075] First, the control unit 12 applies FFT to the images of the inspection workpiece CW taken before and after processing in steps S12 and S16, respectively, to extract high-frequency components, and then performs binarization processing (step S180: image processing step).The control unit 12 then detects a missing portion (P1) from the image of the inspection workpiece CW before processing that has been image-processed in step S180 (step S182).Next, the control unit 12 detects defect candidates (BD, R2, P1, and SD1 to SD3) and the processing line CL1 from the image of the inspection workpiece CW after processing that has been image-processed in step S180 (step S184).The control unit 12 then excludes the missing portion (P1), direct damage BD along the processing line CL1, and the shadow R2 of the laser processing region R1 from the defect candidates in the image of the inspection workpiece CW after processing, and identifies processing defects SD1 to SD3 (step S186).

[0076] In this embodiment, images of the inspection area of ​​the workpiece CW for inspection before and after laser processing are compared to identify machining defects, but the present invention is not limited to this. For example, map data including information on the position and size of defective portions in the workpiece CW for inspection may be acquired in advance and stored in the control unit 12 (map data acquisition unit). Then, defective portions in the map data may be excluded from the image of the workpiece CW for inspection after processing to identify machining defects. In this case, the imaging of the workpiece CW for inspection before laser processing (step S12) and the detection of defective portions (step S182) are omitted.

[0077] Here, the map data may be created by imaging one inspection workpiece CW (e.g., a silicon wafer) using the same method as above to detect missing portions. This map data may then be used in common for inspecting silicon wafers cut from the same silicon ingot as the inspection workpiece CW used to create the map data.

[0078] In the above embodiment, images or map data before and after laser processing are used to distinguish between missing portions and processing defects, but the present invention is not limited to this. For example, missing portions and processing defects may be distinguished from characteristics such as the shape, size, or position of defect candidates detected from images after laser processing. In other words, processing defects may be detected only from images after laser processing.

[0079] Next, the process of evaluating defects during processing will be described with reference to a flowchart shown in FIG.

[0080] First, the control unit 12 calculates an evaluation value of the machining defects (step S200). The evaluation value of the machining defects is the number of machining defects, their size, and their distance from the machining line CL1. The number of machining defects may be calculated, for example, by calculating the number of machining defects per unit area in the region surrounded by the machining line CL1 (the region for one chip). The evaluation value of the machining defects may also be calculated, for example, as a numerical value normalized by the average value in the inspection workpiece CW.

[0081] Next, the control unit 12 compares the evaluation value of the processing defect with the error limit value (step S202). In steps S202 to S204, the control unit 12 compares a representative value (for example, a maximum value, an average value, a median value, etc.) of the evaluation value of various processing defects with the error limit value, and determines whether or not to reset the processing conditions.

[0082] If any of the evaluation values ​​of the machining defects is equal to or greater than the error limit value (Yes in step S204), the control unit 12 outputs an instruction to reset the machining conditions (step S206). On the other hand, if any of the evaluation values ​​of the machining defects is less than the error limit value (No in step S204), the control unit 12 outputs an instruction not to reset the machining conditions (step S208).

[0083] Next, the process of setting the processing conditions will be described with reference to a flowchart shown in Fig. 15.

[0084] First, the control unit 12 identifies the type of evaluation value of the processing defect that is equal to or greater than the error limit value (step S240). Next, the control unit 12 displays candidates for changing the processing conditions on the monitor 310 according to the type of evaluation value that is equal to or greater than the error limit value (step S242).

[0085] Next, the control unit 12 receives an instruction input from the operator to select a candidate for the processing condition (step S244), and resets the processing condition (step S246). Note that the resetting of the processing condition may be arbitrarily set by the operator.

[0086] Here, the processing conditions include, for example, the wavelength, spot diameter, output, repetition frequency, pulse width, numerical aperture (NA) of the condenser lens 208, position of the focal point, polarization characteristics, processing feed speed, etc. of the laser beam. The spot diameter and position of the focal point of the laser beam can be adjusted, for example, by moving the condenser lens 208 in the Z direction. The numerical aperture of the condenser lens 208 can be adjusted, for example, by an aperture. The polarization characteristics can be adjusted, for example, by using a wave plate. By adjusting the processing conditions exemplified above, evaluation values ​​of processing defects, such as the number, size, and distance of processing defects from the processing line CL1, are adjusted so that they are less than the error limit value.

[0087] In the workpiece processing method according to this embodiment, laser processing is performed on the workpiece W to be processed based on the processing conditions reset by the above steps, and the workpiece W is ground and cleaved. As a result, the workpiece W is divided into individual chips.

[0088] According to this embodiment, by extracting high-frequency components from an image of the workpiece CW for inspection captured using a dark-field illumination system, it becomes possible to detect processing defects that occur due to scattering of laser light, etc., at low cost and with high accuracy. Furthermore, according to this embodiment, it becomes possible to accurately evaluate the effect of laser processing on the quality of the device based on the detection results of processing defects, and to appropriately set the processing conditions for the laser processing.

[0089] Furthermore, in this embodiment, the control unit 12 may determine whether the laser processing area is formed at the target position based on the detection results of the position of the shadow R2 of the laser processing area R1 and the direct damage BD along the processing line CL1 (laser processing area). If the control unit 12 determines that the laser processing area is not formed at the target position, it may calculate the amount of correction for the laser processing position and reset the laser processing conditions. [Explanation of symbols]

[0090] 10...laser dicing device, 12...control unit, 100...work moving unit, 102...suction stage, 104...XYZθ table, 106...main body base, 200...laser optical unit, 202...laser oscillator, 204...collimating lens, 206...half mirror, 208...condensing lens (light-collecting lens), 210...driving means, 300...observation optical unit, 302...observation light source, 304...condensing lens, 306...camera, 308...image processing unit, 310...monitor, 400...optical system unit (laser engine)

Claims

1. an imaging step of capturing an image of an inspection range of the workpiece after laser processing of the workpiece using an imaging device while obliquely illuminating the workpiece; an image processing step of applying a fast Fourier transform to an image of the inspection range of the workpiece after the laser processing to extract high-frequency components and performing a binarization process; a detection step of detecting defects during processing formed by scattered light or leaked light generated during the laser processing from the image of the inspection range of the workpiece after the laser processing, which has been image-processed in the image processing step; A workpiece inspection method comprising:

2. 2. The workpiece inspection method according to claim 1, wherein the workpiece consists of only a workpiece layer.

3. a step of capturing an image of an inspection range of the workpiece before the laser processing using the imaging device while obliquely illuminating the workpiece; a step of applying a fast Fourier transform to an image of the inspection range of the work before the laser processing to extract high frequency components and performing a binarization process; The detecting step detecting a missing portion in the inspection range of the workpiece from an image of the inspection range of the workpiece before the laser processing; detecting defect candidates from defects detected in an image of the inspection range of the workpiece after the laser processing; A step of identifying the processing defect by excluding the missing portion, the laser processed area formed in the workpiece by the laser processing, and a shadow of the laser processed area from the defect candidate; The workpiece inspection method according to claim 1 or 2, further comprising:

4. an imaging device that captures an image of an inspection range of a workpiece after laser processing of the workpiece while obliquely illuminating the workpiece; an image processing unit that performs a fast Fourier transform on an image of the inspection range of the workpiece after the laser processing to extract high frequency components and perform binarization processing; a detection unit that detects defects during processing formed by scattered light or leaked light generated during the laser processing from an image of the inspection range of the workpiece after the laser processing that has been image-processed by the image processing unit; A workpiece inspection device comprising:

Images