TWI814791B - Non-destructive testing methods - Google Patents

Abstract

The object of the present invention is to efficiently confirm the depth position and length of a modified layer formed on a workpiece by laser processing, and to quickly set appropriate laser processing conditions. The solution is to intermittently photograph the inside of the workpiece at predetermined intervals (H) in the Z-axis direction orthogonal to the X-axis and Y-axis planes to obtain multiple X-axis and Y-axis plane images, and calculate The obtained 3D image (101) is a clear 3D image in which blur is removed by deconvolution, and the 3D clear image is cut parallel to the Z-axis. From the 2D image of the cross section of the modified layer, detection is performed Get the Z-axis coordinate value of the modified layer and the length of the modified layer. Laser processing and modified layer status detection can be quickly performed repeatedly, and the most suitable laser processing conditions for forming the modified layer can be quickly found.

Description

Non-destructive testing methods

The present invention relates to a method for detecting the state of a modified layer formed inside a workpiece by laser processing.

There is a method for forming a device on the back of a wafer from an area divided by a predetermined dividing line on the surface, irradiating laser light with a wavelength that is transmissive to the wafer along the predetermined dividing line, and focusing the light on the inside of the wafer. A method in which a modified layer is formed at a light focusing point, and then an external force is applied to the modified layer, and the wafer is divided using the modified layer as a starting point (for example, see Patent Document 1).

In this dividing method, the depth position and length of the modified layer in the thickness direction of the wafer are related to the ease of dividing the wafer. Therefore, the depth, position and length of the modified layer can be observed to determine whether a modified layer that is most suitable for division is formed.

Therefore, a method has also been proposed in which the end portion of the wafer is cut off in advance, a modified layer is formed inside the wafer, and then the cut surface of the wafer is photographed to observe the state of the modified layer (for example, see Patent document 2).

[Prior technical literature] [Patent Document]

[Patent Document 1] Japanese Patent No. 3408805

[Patent Document 2] Japanese Patent Application Publication No. 2017-166961

However, in order to determine whether the depth position and length of the modified layer are optimal for dividing the workpiece, it is necessary to alternately repeat the formation of the modified layer and the observation of the formed modified layer. This is described in Patent Document 2. In this method, it is necessary to cut the wafer to observe the modified layer, so it takes time to set the optimal laser processing conditions.

The present invention was invented in view of the above-mentioned problems. The object is to efficiently confirm the depth, position and length of the modified layer formed on the workpiece by laser processing, and to quickly set appropriate laser processing conditions. .

The present invention is a non-destructive detection method that non-destructively detects laser light with a wavelength that is transmissive to a workpiece having a first surface and a second surface opposite to the first surface. The focus point is a non-destructive inspection method that is positioned inside the object to be processed and irradiates laser light along the X-axis direction to separate the modified layers formed side by side in the X-axis direction. It includes: Preparation process , an inspection device is prepared. The inspection device is equipped with an imaging means equipped with an objective lens and images the inside of the workpiece from the first surface side, and irradiates a wavelength range that is transmissive to the workpiece from the first surface side. The light source of the light, the driving means for making the object lens approach and leave the first surface, and the memory of the image captured by the imaging means The memory method; to obtain the process, when the surface parallel to the first surface is set as the X-axis Y-axis plane, the object is intermittently moved at a predetermined interval H in the Z-axis direction orthogonal to the X-axis Y-axis plane. The lens is close to the first surface, and the focus is positioned at the Z-axis coordinate position in the object to be processed, which is calculated by multiplying the refractive index of the object to be processed by the movement amount of the object lens in the Z-axis direction. Out of the Z-axis coordinate positions at intervals, a plurality of X-axis and Y-axis plane images including the modified layer inside the object are obtained for each Z-axis coordinate value, and stored in the memory means; the memory process is based on the Obtain the 3D image generated by the complex X-axis and Y-axis plane images of each Z-axis coordinate value memorized in the project, calculate the clear 3D image with the blur removed by deconvolution, and store it in the memory method; And the detection process is to obtain the Z-axis coordinate value and Z of the modified layer based on the 2-dimensional image of the cross-section of the modified layer obtained by cutting the 3-dimensional clear image memorized in the memory process parallel to the Z-axis. The length of the modified layer in the axial direction is determined, and cracks extending from the modified layer in the X-axis direction are detected based on a two-dimensional image of a cross-section obtained by cutting the three-dimensional clear image perpendicularly to the Z-axis.

The aforementioned deconvolution is the Fourier transform of the 3-dimensional image generated based on the X-axis and Y-axis plane images obtained from the Z-axis coordinate values memorized in the process, divided by the focus of the imaging means located in the modified layer. It is better to perform the Fourier transform of the 3D PSF of the blur effect caused by the optical system, and then perform the inverse Fourier transform to calculate a clear 3D image.

The aforementioned 3-dimensional PSF is preferably calculated using the Gibson and Lanni model.

The vivid three-dimensional image recorded in the aforementioned memory project is a three-dimensional image that includes the distinct X-axis and Y-axis plane images of the distinct Z-axis coordinate values after the blur has been removed by the aforementioned deconvolution. By based on the two X-axis and Y-axis distinct plane images adjacent to the object pixel in the Z-axis direction Based on the pixel value and distance of the pixel, use the linear interpolation method to calculate the pixel value of the object pixel between the two opposing X-axis and Y-axis plane images, and interpolate between the separated X-axis and Y-axis clear plane images. The pixel value of the plurality of pixels is better.

In the present invention, a three-dimensional clear image is obtained in the memory process, the three-dimensional clear image is cut parallel to the Z-axis in the detection process, and the Z-axis coordinate of the modified layer is detected from the two-dimensional image of the cross section of the modified layer. value and the length of the modified layer, so the position and length of the modified layer can be grasped without damaging the workpiece. Therefore, the laser processing and the state detection of the modified layer can be repeatedly performed quickly, and the most suitable laser processing conditions for forming the modified layer can be quickly found.

The workpiece W shown in FIG. 1 is, for example, a circular plate-shaped substrate, and its surface (the first surface Wa in the example shown) is divided by a plurality of grid-shaped predetermined dividing lines S. Area, formed with plural devices D. The predetermined dividing line S extends in the X-axis direction and the Y-axis direction.

The tape T is adhered to the second side Wb opposite to the first side Wa. The workpiece W is integrated with the annular frame F via the tape T. Hereinafter, with reference to the attached drawings, the non-destructive method of detecting, condensing, and transmitting laser light of a wavelength that is transparent to the workpiece W having the first surface Wa and the second surface Wb opposite to it will be described. A non-destructive detection method of the modified layer formed inside the object W by irradiation.

(1)Preparation project As shown in FIG. 1 , for example, an inspection device 1 capable of forming a modified layer inside the workpiece W and imaging the inside of the workpiece W is prepared. The inspection device 1 includes a device base 10, and a support 11 with a roughly L-shaped cross section is erected on the rear side of the device base 10 in the Y-axis direction. The device base 10 is provided with a holding base 12 for holding the workpiece W integrated with the frame F. The frame holding means 15 for holding the frame F is arranged around the holding base 12 and moves the holding base 12 in the X-axis direction. The X-axis direction moving means 20 and the Y-axis direction moving means 30 for moving the holding base 12 in the Y-axis direction. The front end of the support 11 extends to the upper side of the path in the moving direction (X-axis direction) of the holding table 12 .

The upper surface of the holding base 12 serves as a holding surface 12a for holding the workpiece W. The holding base 12 is fixed on the cover base 13 having the opening 130 , and the rotating means 14 is connected to the lower part of the holding base 12 . The rotation means 14 can rotate the holding platform 12 through a predetermined angle.

The X-axis direction moving means 20 is provided with a ball screw 21 extending in the X-axis direction, a motor 22 connected to one end of the ball screw 21, and a pair of guide rails 23 extending parallel to the ball screw 21 to rotatably support the balls. The bearing portion 24 at the other end of the screw 21 and the moving base 25 support the holding table 12 through the Y-axis direction moving means 30. One surface of the pair of guide rails 23 is slidably contacted with the moving base 25 , and the ball screw 21 is screwed to a nut formed in the center of the moving base 25 . When the motor 22 rotates the ball screw 21, the moving base 25 moves in the X-axis direction along the guide rail 23, thereby moving the holding table 12 in the X-axis direction.

The Y-axis direction moving means 30 is provided with a ball screw 31 extending in the Y-axis direction, a motor 32 connected to one end of the ball screw 31, and a pair of guide rails 33 extending parallel to the ball screw 31 to rotatably support the balls. The bearing portion 34 at the other end of the screw 31 and the moving base 35 support the holding base 12 . One surface of the moving base 35 is slidably contacted with the pair of guide rails 33 , and the ball screw 31 is screwed to a nut formed in the center of the moving base 35 . When the motor 32 rotates the ball screw 31, the moving base 35 moves in the Y-axis direction along the guide rail 33, so that the position of the holding table 12 in the Y-axis direction can be adjusted.

The inspection apparatus 1 is equipped with the laser processing means 40 which applies laser processing to the 1st surface Wa of the workpiece W held by the holding table 12. The laser processing means 40 includes a laser processing head 41 disposed on the lower side of the front end of the support 11 and irradiating downward a laser beam 43 having a wavelength that is transparent to the workpiece W shown in FIG. 2 . The laser processing head 41 is connected to an oscillator that oscillates the laser light 43 and an output regulator that adjusts the output of the laser light 43 . As shown in FIG. 2 , a condenser lens 42 for condensing the laser light 43 oscillated from the oscillator is built into the laser processing head 41 . The laser processing head 41 can move in the vertical direction and adjust the focusing position of the laser light 43.

Here, an example of forming a modified layer inside the workpiece W by the laser processing means 40 will be described. In this embodiment, for example, laser processing conditions described below are set and implemented. Furthermore, the workpiece W is, for example, a silicon wafer.

[Laser processing conditions] Wavelength of laser light: 1064nm Repetition frequency: 50kHz Average output: 1.0W Pulse width: 10nm Focusing point: φ3.0μm Processing feed speed: 500mm/s

As shown in FIG. 2 , if the workpiece W is sucked and held with the holding surface 12 a of the holding table 12 with the tape T side facing downward, the holding table 12 is moved below the laser processing means 40 . Next, while the holding table 12 is processed and fed in, for example, the X-axis direction at the aforementioned processing feed speed (500 mm/s), the laser with a wavelength that is transmissive to the workpiece W is captured by the condenser lens 42 With the focusing point of the light 43 positioned inside the object W, the laser light 43 is irradiated from the first surface Wa side of the object W along the predetermined dividing line S shown in FIG. A modified layer M with reduced strength is formed inside W.

The inspection device 1 shown in FIG. 1 is equipped with an object lens 52 as shown in FIG. 3 in order to detect the modified layer M formed inside the object W in a non-destructive manner. The imaging means 50 for imaging one surface Wa, the light source 60 for irradiating light in a wavelength range that is transmissive to the workpiece W from the first surface Wa side, and the driving means 70 for moving the object lens 52 toward or away from the first surface Wa. In addition, as shown in FIG. 1 , the inspection device 1 is provided with a memory means 80 that memorizes the image captured by the image pickup means 50 , a control means 90 that can perform image processing based on the image memorized by the memory means 80 , and a display device that displays various data (images). , processing conditions, etc.) monitor 100.

The imaging means 50 is disposed close to the laser processing means 40 on the lower side of the front end of the support 11 . As shown in FIG. 3 , the imaging means 50 includes a camera 51 for imaging the workpiece W from above, an objective lens 52 disposed at the bottom of the camera 51 , and an object lens 52 disposed between the camera 51 and the objective lens 52 . The light emitted by the light source 60 is reflected to the half mirror 53 below. The camera 51 is an infrared camera with a built-in imaging element such as a CCD image sensor or a CMOS image sensor. The light source 60 is composed of, for example, an infrared LED, and can emit infrared rays 61 in a wavelength range that is transparent to the workpiece W. In the imaging means 50, the reflected light of the infrared ray 61 emitted from the light source 60 and reflected inside the object W can be captured based on the X-axis coordinate and Y-axis coordinate inside the object W by using the imaging element. X-axis and Y-axis plane image. The X-axis and Y-axis plane images captured by the imaging means 50 are stored in the memory means 80 .

A driving means 70 is connected to the objective lens 52 . The driving means 70 is an actuator capable of moving the objective lens 52 up and down in the Z-axis direction. The driving means 70 is, for example, a piezoelectric motor composed of a piezoelectric element that expands and contracts in the vertical direction with respect to the workpiece W held by the holding table 12 by application of voltage. In the driving means 70, by adjusting the voltage applied to the piezoelectric element to move the objective lens 52 in the up and down direction, the position of the objective lens 52 can be finely adjusted. Therefore, the driving means 70 can move the position of the objective lens 52 for each desired Z-axis coordinate value, and the imaging means 50 can be used to image the X-axis and Y-axis plane inside the workpiece W for each Z-axis coordinate value. Take video. Furthermore, the driving means 70 is not limited to a piezoelectric motor, and may be configured by a linearly movable voice coil motor, for example.

The control means 90 at least includes a CPU that performs calculation processing using a control program, a ROM that stores the control program, etc., a readable and writable RAM that stores calculation processing results, and an input interface and an output interface. The control means 90 controls the rotation means 14, the X-axis direction movement means 20, the Y-axis direction movement means 30, and the driving means 70. Furthermore, the control means 90 is provided with an image processing unit 91 that processes the image formed by the imaging means 50 and the image stored in the memory means 80 .

In addition, the image processing unit 91 may generate a 3D image from the plurality of captured 2D images, or may cut out the 3D image to form a cross-sectional image of the modified layer formed inside the workpiece W (as is the case with Image cut in a direction parallel to the Z axis). By displaying the two-dimensional plane image and the three-dimensional image thus obtained on the monitor 100, the state of the modified layer can be observed.

(2) Obtain project When the inspection device 1 is prepared and the modified layer M is formed inside the workpiece W, as shown in FIG. 3 , the holding table 12 is processed and fed in the The first surface Wa side of W captures an image of the internal state of the workpiece W. In the image acquisition process shown in this embodiment, X-axis and Y-axis plane images parallel to the first surface Wa of the workpiece W are captured multiple times. In this embodiment, a description will be given of a situation in which the modified layer M is formed along a predetermined dividing line S divided into columns in the X-axis direction and then the image acquisition process is performed.

Here, the infrared ray 61 emitted by the light source 60 shown in FIG. 3 is reflected downward in the half mirror 53 and is incident on the first surface Wa through the objective lens 52. According to the refractive index (N) of the object W to be processed, the infrared ray 61 The refraction angle will change. That is, the refractive index (N) differs depending on the type of material of the workpiece W. FIG. 4 shows the relationship between the refractive index (N) of the object W and the focus of the infrared rays 61 condensed by the objective lens 52 . For convenience of explanation, the angle α relative to the optical axis O shown in the illustrated example shows that the infrared ray 61 passing through the objective lens 52 is not refracted but is incident linearly on the first surface Wa of the object W. From this point on, The distance from the first surface Wa to the focus point P is set to distance h1.

Normally, when infrared rays 61 passing through the objective lens 52 are incident into the object W from the first surface Wa, the infrared rays 61 are not refracted from the angle α, for example, only the refraction angle β, and are focused on the focus point P'. The angle β with respect to the optical axis O corresponds to the refraction angle, and the refractive index (N) of the workpiece W at this time is calculated based on Snell's law and the calculation formula (1) described below.

In addition, by substituting the refractive index (N) calculated using the aforementioned calculation equation (1) into the calculation equation (2) described later, the distance h2 from the first surface Wa of the workpiece W to the focus point P' can be calculated .

The distance h2 is longer than the distance h1, which confirms that the focus is far away from the objective lens.

When imaging the inside of the workpiece W, the driving means 70 intermittently moves the object lens 52 at a predetermined interval H in the Z-axis direction orthogonal to the X-axis and Y-axis planes. Intermittently moving the objective lens 52 means setting a certain interval, and moving the position of the objective lens 52 in the Z-axis direction by a movement amount V each time. That is, interval H=movement amount V. The predetermined interval HR shown in the example of FIG. 5 changes according to the refractive index (N) of the workpiece W to be inspected and the movement amount (V) of the objective lens 52 in the Z-axis direction, and can be calculated using the above-mentioned formula. (1) Calculate by multiplying the calculated refractive index (N) by the amount of movement (V) (HR=N×V).

When the workpiece W shown in this embodiment is, for example, a silicon wafer, its refractive index (N) is 3.6. When the movement amount (V) caused by the driving means 70 is set to 1 μm, for example, by multiplying the refractive index (3.6) of the workpiece W by the movement amount (1 μm), the predetermined interval HR can be calculated to be 3.6 μm. That is, the distance HR between the focal points extending inside the workpiece W (the distance between the Z-axis coordinate value z1 and the Z-axis coordinate value z2) is at least 3.6.

The driving means 70 lowers the objective lens 52 in a direction close to the first surface Wa of the object W to position the focus P1 at the Z-axis coordinate value z1. When the camera 51 shown in FIG. 3 is used to image the inside of the workpiece W, for example, the X-axis and Y-axis plane image 2a shown in FIG. 6 can be obtained. Next, the driving means 70 intermittently moves the objective lens 52 toward the first surface Wa side based on the setting of the aforementioned predetermined interval H (1 μm), and positions the focus P2 at the interval of the focus P1 based on the aforementioned refractive index (N). Extended Z-axis coordinate value z2. When the camera 51 images the inside of the workpiece W, for example, the X-axis and Y-axis plane image 2b can be obtained. In this way, the driving means 70 intermittently moves the position of the object lens 52 at the predetermined interval H, and uses the camera 51 to image the inside of the workpiece W for each Z-axis coordinate value z1, z2..., and can sequentially obtain X Axis Y axis plane images 2a, 2b, 2c, 2d, 2e, 2f and 2g. Then, the obtained X-axis and Y-axis plane images 2a to 2g are stored in the storage means 80 shown in FIG. 1 .

(3)Memory engineering In this project, the X-axis and Y-axis plane images 2a-2g of each Z-axis coordinate value obtained in the project are initially obtained by overlapping and converging, and a three-dimensional image 101 as shown in Figure 7 is obtained. The three-dimensional image 101 formed in this way is formed based on the actual observed X-axis and Y-axis plane image, so there are blurred parts. This image is hereinafter referred to as a three-dimensional observation image.

In the acquisition process, the modified layer M shown in FIG. 3 is observed while shifting the focus of the imaging means 50. Therefore, it can be inferred that the modified layer M is a collection of a large number of point light sources. Then, there is a blurred part in the 3D observation image, so it is necessary to remove the blur through deconvolution and calculate a clear 3D image. As shown in Figure 7, this blur can be expressed as a point spread function PSF (x, y, z) that reveals the three-dimensional spread of light from a point light source. PSF(x,y,z) estimates how a point light source looks like. By removing the blur from a 3D observation image, a clear 3D image can be obtained. The obtained three-dimensional vivid image is stored in the memory device 80. In this project, a clear 3D image is obtained through the method shown below. As the method, there are, for example, the asymptotic method and the inverse filtering method.

(A) Asymptotic method In the asymptotic method, as shown in the calculation formula (3), the estimated value O _k (x, y, z) of the brightness distribution of the three-dimensional clear image is moved toward the true brightness distribution O (x, y, z ) asymptotically.

Here, OTF (x, y, z) is an optical transfer function, which is obtained by performing Fourier transformation on the point spread function PSF (x, y, z).

According to the following calculation formula (4), the estimated value O _k+1 (x, y, z) is updated and asymptotically approaches the true brightness distribution O _k (x, y, z).

In the calculation formula (4), first obtain the difference between the value of O _k (x, y, z) × OTF (x, y, z) and the Fourier transform I (x, y, z) of the three-dimensional observation image. This difference is a fuzzy component included in the estimated value O _k (x, y, z). Subtract the value of this difference from O _k (x, y, z) to obtain the next estimated value O _{k + 1} (x, y, z). Substitute the obtained O _k+1 (x, y, z) into O _k (x, y, z) of the calculation formula (4), and then obtain O _k+1 (x, y, z). This calculation is repeated until the difference becomes 0, that is, until the fuzzy component becomes 0. Furthermore, when calculating the initial calculation caused by equation (4), substitute I(x,y,z) into _Ok (x,y,z). By repeating the calculation of equation (4), O _k+1 (x, y, z) when the fuzzy component becomes 0 becomes a three-dimensional clear image.

When repeating the calculation of the aforementioned calculation formula (4), by using the maximum approximation method to reduce the number of calculations of the calculation formula (4), even in a situation where the 3D observation image has a lot of noise, a clear 3D image can be obtained portrait.

In addition, OTF (x, y, z) can also be estimated by using the blind deconvolution method using the most approximate method to obtain a clear 3D image. In the blind deconvolution method, OTF(x,y,z) is also updated every time the calculation is repeated.

(B)Inverse filtering method In the inverse filtering method, first, the Fourier transform O(x, y, z) of the brightness distribution of the three-dimensional observation image is obtained by the following calculation formula (5).

Here, I (x, y, z) is the Fourier transform of the three-dimensional observation image, and OTF (x, y, z) is the Fourier transform of the point spread function PSF (x, y, z). In the aforementioned calculation formula In (5), the Fourier transform of the 3D observation image is divided by the Fourier transform of the point spread function. Then, by performing inverse Fourier transform on the obtained O(x, y, z), the brightness distribution o(x, y, z) of the three-dimensional clear image is obtained.

In addition, in the Wienner method, as shown in the following calculation equation (6), a constant w is added to the denominator of the aforementioned calculation equation (5), and a larger weight is added to the frequency band with a relatively high S/N ratio. Reconstruction of signal components. The constant w added to the denominator functions as a low-pass filter that removes high-frequency components.

Furthermore, the point spread function PSF (x, y, z) of the aforementioned asymptotic method and inverse filtering method follows the Gibson and Lanni model and can be calculated by the following calculation formula (7).

Here, the variables and constants in calculation formula (7) are as follows. k0: wave number (=2π/wavelength) Λ: light path difference x, y: x coordinate and y coordinate of the observation position x0,y0: x-coordinate and y-coordinate of the position of the point light source NA: Numerical aperture of objective lens ρ: The distance from the center of the objective lens when the center of the objective lens is set to ρ=0 and the outermost periphery of the objective lens is set to ρ=1.

In addition, the optical path difference Λ can be calculated using the following calculation formula (8).

Here, the variables and constants in calculation formula (8) are as follows. z: Z coordinate of the observation position z0: z coordinate of the position of the point light source ns: refractive index of the object to be processed NA: Numerical aperture of objective lens ρ: The distance from the center of the objective lens when the center of the objective lens is set to ρ=0 and the outermost periphery of the objective lens is set to ρ=1.

In addition, in the calculation formula (8), the following calculation formula (9) has a difference in optical path from when there is no workpiece.

In addition, the following calculation formula (10) represents the out-of-focus component derived from the point light source.

(4)Testing project Next, the three-dimensional clear image stored in the memory means 80 is cut parallel to the Z-axis, and a plurality of two-dimensional clear images are obtained. For example, the Z-axis and X-axis clear plane image 400 shown in FIG. 9 is one example. Here, the three-dimensional PSF 102 is estimated to expand toward the first surface Wa and the second surface Wb with the light condensing point P0 as the center.

In the Z-axis and X-axis clear plane image 400 shown in FIG. 9, the double scale on the vertical axis is the interval H×refractive index shown in FIG. 5, which can be used to find the length L of the modified layer M in the Z direction. In addition, the Z coordinates of the upper end and the lower end of the modified layer M can also be obtained (first inspection step).

On the other hand, as shown in FIGS. 10 (a) to (c), the three-dimensional clear image stored in the memory means 80 can also be cut perpendicularly to the Z-axis, that is, parallel to the X-axis and Y-axis plane, and obtained Most X-axis and Y-axis clear plane images 501a, 501b, 501c. However, the X-axis and Y-axis clear plane images facing each other in the Z-axis direction are separated by a distance HR. Therefore, it is necessary to use interpolation (linear interpolation) to find the pixel value of the target pixel between the opposing X-axis and Y-axis clear plane images. For example, as shown in Figure 8, when the pixel value of the pixel 300 between the X-axis Y-axis plane image 2a and the X-axis Y-axis plane image 2b is found, the positive value of the pixel 300 of the X-axis Y-axis plane image 2a is found. The pixel value of the upper pixel 201a and the pixel value of the pixel 201b directly below the pixel 300 of the X-axis and Y-axis plane image 2b. Furthermore, the distance Z11 in the Z-axis direction from the pixel 300 to the pixel 201a and the distance Z12 in the Z-axis direction from the pixel 300 to the pixel 201b are respectively determined. Then, weighting is performed according to each distance, and the weighted weight is used to obtain a weighted average of the pixel value of the pixel 201a and the pixel value of the pixel 201b, and the value of the weighted average is set as the pixel value of the pixel 300. Similarly, for example, find the pixel value of the pixel 202a directly above the pixel 301 of the X-axis and Y-axis plane image 2a and the pixel value of the pixel 202b directly below the pixel 301 of the X-axis and Y-axis plane image 2b, and then calculate the pixel values from the pixels respectively. The distance Z21 in the Z-axis direction from 301 to the pixel 202a and the distance Z22 in the Z-axis direction from the pixel 301 to the pixel 202b are weighted according to each distance, and the pixel value of the pixel 202a and the pixel value of the pixel 202b are obtained using this weighting. The weighted average of pixel values is used as the pixel value of pixel 301. In this way, the pixel value of each pixel constituting the XY plane image 2ab between the X-axis Y-axis plane image 2a and the X-axis Y-axis plane image 2b is obtained. Then, by calculating pixel values for all pixels that may exist between adjacent X-axis and Y-axis plane images, the pixel values of all pixels in the three-dimensional space are specified to form a three-dimensional image (pixel value calculation process). Furthermore, in the pixel value calculation process, interpolation methods such as bilinear interpolation, nearest neighbor method, and bicubic interpolation method are used. The X-axis and Y-axis clear plane image 501a shown in FIG. 10(a) is an image of a part of the cross-sectional image on the side farther away from the first surface Wa than the modified layer M of the three-dimensional clear image after linear interpolation. According to this The image shows that the first crack C1 connected in the X-axis direction is formed above the modified layer M.

The X-axis and Y-axis clear plane image 501b shown in FIG. 10(b) is an enlarged part of the cross-sectional image near the center of the modified layer M. From this image, the state of the central part of the modified layer M can be understood. In addition, cracks C extending in the X-axis direction are formed between adjacent modified layers M, and the state of connection of the modified layers M through the cracks C can also be grasped (second detection process).

The X-axis and Y-axis clear plane image 501c shown in FIG. 10(c) is an enlarged image of a part of the cross-sectional image closer to the second surface Wb than the light-condensing point P0. In this image, the modified layer M can be understood Below, a second crack C2 is formed that is continuous in the X-axis direction. In this way, it is detected that the first crack C1 shown in FIG. 10(a) and the second crack C2 shown in FIG. 10(c) extend continuously in the Y-axis direction, and the pattern connecting the modified layer M is confirmed. As shown in 10(b) From the fact that the crack C extends in the Y-axis direction, it can be seen that the processing conditions at this time are processing conditions that can reliably divide the workpiece by applying external force later (the second inspection process).

In this way, by obtaining a three-dimensional clear image, cross-sectional images can be formed that cut the image from various directions, so the positions and states of the modified layer M and the cracks C can be confirmed from various angles. Furthermore, the X-axis and Y-axis cross-sectional image 501ab shown in FIG. 9 is an image obtained by linear interpolation, and this cross-sectional image can also be confirmed.

In addition, the XY clear plane images 501a, 501b, and 501c obtained in this way are cross-sectional images of the three-dimensional clear image with blur removed in the memory process, and therefore can be more vivid than the X-axis and Y-axis plane images 2a to 2g obtained in the acquisition process. By. Therefore, the positions and states of the modified layer M and the cracks C can be grasped more reliably. The depth, position and length of the modified layer M in the thickness direction of the workpiece are related to the ease of division of the workpiece. Therefore, the modified layer can be judged by grasping the depth, position and length of the modified layer M. Is M formed in such a way that the workpiece can be easily divided? In addition, the depth position and length of the modified layer formed on the object to be processed by laser processing can be confirmed efficiently, so appropriate laser processing conditions can be quickly set.

Even if a three-dimensional observation image is formed from the X-axis and Y-axis plane images 2a to 2g and then the pixel values of the pixels between the X-axis and Y-axis plane images are obtained by interpolation, it will be difficult because the three-dimensional observation image contains blurred parts. The appropriate pixel value is found. However, after removing the blur of the 3D observation image through deconvolution, the pixel value of the pixel between each X-axis and Y-axis plane image is found through interpolation, so it can finally be Get vivid 3D images. Furthermore, the line Sexual interpolation means, for example, dividing adjacent X-axis and Y-axis plane images by 100 and interpolating them.

Obtain various images as mentioned above to understand the position and length of the modified layer and the connection of the cracks. For example, if the cracks are not connected in the X-axis direction, they cannot be properly divided even if external force is applied, so the processing conditions are changed and the modification is performed again. layer formation, each image was again acquired for the same analysis. When judging whether the processing conditions are appropriate, there is no need to cut the workpiece, so the processing conditions can be adjusted quickly and repeatedly to confirm the status of the modified layer and cracks.

Furthermore, the inspection device 1 shown in this embodiment is configured to also have the function of a laser processing device that forms the modified layer M inside the workpiece W. However, the inspection device 1 is not limited to this invention. The device structure shown in the embodiment may be a separate device structure from the laser processing device.

W: workpiece

Wa: first side

S: Predetermined dividing line

D:Device

Wb: second side

M: modified layer

C: Crack

T:Tape

F:frame

1: Check the device

2a, 2b, 2c, 2d, 2e, 2f, 2g: X-axis and Y-axis plane images

10:Device base

11:Pillar

12:Keeping platform

12a:Maintenance surface

13: Covering table

130: opening

14:Rotation means

15: Frame maintenance means

20: X-axis direction movement means

21: Ball screw

22: Motor

23:Guide track

24:Bearing Department

25:Mobile base

30: Y-axis direction movement means

31: Ball screw

32: Motor

33:Guide track

34:Bearing Department

35:Mobile base

40‧‧‧Laser processing methods 41‧‧‧Laser processing head 42‧‧‧Condensing lens 43‧‧‧Laser light 50‧‧‧Camera means 51‧‧‧Camera 52‧‧‧objective lens 53‧‧‧half mirror 60‧‧‧Light source 61‧‧‧Infrared 70‧‧‧Driving means 80‧‧‧Memory means 90‧‧‧Control means 91‧‧‧Image Processing Department 100‧‧‧Monitor 101‧‧‧3D portrait 102‧‧‧3D PSF 400‧‧‧ZX vivid flat image 501a, 501b, 501c‧‧‧XY distinct flat image

[Fig. 1] A perspective view showing the structure of an example of the inspection device.

[Fig. 2] A cross-sectional view showing a state in which a modified layer is formed inside the workpiece.

[Figure 3] A cross-sectional view showing the image acquisition process.

[Fig. 4] An explanatory diagram showing the relationship between the refractive index of the object to be processed and the focus of the objective lens.

[Fig. 5] An explanatory diagram illustrating a state in which the object lens is intermittently moved at a predetermined interval H during the acquisition process.

[Fig. 6] An image diagram showing multiple X-axis and Y-axis plane images obtained in the acquisition process. [Fig. 7] An image diagram showing an example of an observed three-dimensional image. [Fig. 8] A perspective view showing an example of obtaining pixel values by linear interpolation. [Fig. 9] An image diagram showing an example of a Z-axis and X-axis clear plane image obtained by cutting a three-dimensional image parallel to the Z-axis. [Fig. 10] An image diagram showing an example of an X-axis and Y-axis clear plane image obtained by cutting a three-dimensional image parallel to the XY plane.

101‧‧‧3D portrait

102‧‧‧3D PSF

M‧‧‧Modification layer

Claims (4)

A non-destructive detection method that non-destructively detects a focus point of a laser light having a wavelength that is transmissive to a workpiece having a first surface and a second surface opposite to the first surface , a non-destructive testing method that is positioned inside the workpiece and irradiates laser light along the X-axis direction to separate the modified layers formed side by side in the X-axis direction. It is characterized by: preparation process, system An inspection device is prepared. The inspection device is equipped with an imaging means equipped with an objective lens and for imaging the inside of the object to be processed from the first surface side, and irradiates light in a wavelength range that is transmissive to the object to be processed from the first surface side. A light source, a driving means for causing the object lens to approach and leave the first surface, and a memory means for memorizing the image captured by the imaging means; the acquisition process is to set the plane parallel to the first surface as the X-axis and Y-axis plane, in the Z-axis direction orthogonal to the X-axis and Y-axis plane, intermittently bring the object lens close to the first surface at a predetermined interval H, and position the focus on the Z-axis of the object to be processed where the modified layer is formed The coordinate positions are the Z-axis coordinate positions at intervals in the Z-axis direction calculated by multiplying the refractive index of the object to be processed by the movement amount of the object lens in the Z-axis direction. For each Z-axis coordinate value, obtain The plurality of X-axis and Y-axis plane images of the modified layer inside the processed object are stored in the memory means; the memory process is a plurality of X-axis and Y-axis planes based on the Z-axis coordinate values memorized in the acquisition process. The 3D image generated from the image is calculated as a clear 3D image with the blur removed by deconvolution, and is stored in the image. The memory means; and the detection process are to obtain the Z-axis coordinate of the modified layer based on a 2-dimensional image of the cross-section of the modified layer obtained by cutting the 3-dimensional clear image memorized in the memory process parallel to the Z-axis. The value is related to the length of the modified layer in the Z-axis direction, and based on the 2-dimensional image of the cross-section obtained by cutting the 3-dimensional clear image perpendicularly to the Z-axis, a turtle extending from the modified layer in the X-axis direction is detected. crack. For example, the non-destructive testing method described in item 1 of the patent application, wherein the deconvolution is a three-dimensional image generated based on the X-axis and Y-axis plane image of each Z-axis coordinate value memorized in the process. The Fourier transform is divided by the Fourier transform of the 3-dimensional PSF representing the blurring effect caused by the optical system of the focal point of the imaging means located in the modified layer, and then the inverse Fourier transform is performed to calculate a clear 3-dimensional image. For example, the non-destructive testing method described in item 2 of the patent application, wherein the aforementioned 3-dimensional PSF is a calculation formula of the Gibson and Lanni model. The non-destructive detection method described in the first item of the patent application, wherein the vivid 3-dimensional vivid image recorded in the aforementioned memory project is for each Z after the blur has been removed by the aforementioned deconvolution. A 3D image of a plane image with clear complex X-axis and Y-axis coordinate values, by Based on the pixel values and distances from the pixels of the two X-axis and Y-axis sharp plane images adjacent to the object pixel in the Z-axis direction, the linear interpolation method is used to calculate the distance between the two opposing X-axis and Y-axis plane images. The pixel value of the object pixel is interpolated from the pixel value of the complex pixels between the separate X-axis and Y-axis clear plane images.

Images