JP2007258555A - Method and device for inspecting crystal defect in silicon wafer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for inspecting the crystal defect of a silicon wafer capable of discriminating a fine void from a particle, and to provide a device therefor. <P>SOLUTION: The silicon wafer is irradiated with infrared light, and the crystal defect is inspected from an image which is obtained by performing imaging with the use of the transmission infrared light. In a low-magnification inspecting step S1, a low-magnification inspecting image having the fixed visual field region of the silicon wafer is acquired through the use of a low-power objective lens 14a. The low-magnification inspecting image is image-processed so as to confirm and store the position of the crystal defect in a position confirming step S3. In a high-magnification inspecting step S4, the high-magnification inspecting image of the visual field region is acquired by positioning the position of the stored crystal defect in the visual field region so as to change an objective distance through the use of a high-power objective lens 14b. In a determining step S5, the shape of the crystal defect is obtained based on the high-magnification inspecting image so as to determine whether the crystal defect is the void or the particle by whether the shape is a circular shape or an unspecified shape other than the circular shape. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a crystal defect inspection method and apparatus for a silicon wafer. In particular, the present invention relates to a method and apparatus for inspecting a crystal defect of a silicon wafer using infrared rays.

  An ingot that becomes a single crystal grown from a silicon melt may generate spherical voids (voids) that become crystal defects in the process of lowering the temperature. And in the silicon wafer which sliced this ingot, the said void exists in a bulk or the said void appears on the surface. On the other hand, irregularly shaped particles (Particles) may adhere to the surface of the silicon wafer.

  As a method for inspecting voids generated on a silicon wafer, an infrared interference method using infrared rays (so-called IR method), an ultrasonic flaw detection method using an ultrasonic microscope (SAM), and an X-ray topography method using X-rays. and so on. Table 1 compares the void detection limit values according to the inspection methods and the current state of throughput (processing time per wafer).

  As shown in Table 1, the ultrasonic flaw detection method is inferior in throughput as compared with the infrared interference method, but is widely used for crystal defect inspection of silicon wafers because it can detect even fine voids. Infrared interferometry with excellent throughput is used for defect inspection of bonded wafers.

  For example, as a defect inspection apparatus related to a bonded wafer, a bonded wafer defect inspection apparatus including an image input unit, a defect detection unit, and a defect inspection quality determination unit has been invented (for example, see Patent Document 1).

  In the wafer defect inspection apparatus according to Patent Document 1, the image input means irradiates the bonded wafer with infrared light, detects the transmitted infrared light from the wafer, and inputs it as a quantized image. The defect detection means detects the defect of the wafer from the wafer image input obtained by the image input means. The defect inspection pass / fail determination means compares the detection result of the defect detection means with preset inspection conditions, and extracts regular wafer defects.

  Further, as a manufacturing method and a defect inspection apparatus related to a bonded wafer, a bonded wafer manufacturing method and a bonding defect inspection apparatus capable of observing defects in a silicon wafer bonding operation have been invented (for example, see Patent Document 2).

The method for manufacturing a bonded wafer according to Patent Document 2 is performed while reflecting two infrared mirrors and observing the overlapping surface when two mirror-polished wafers are bonded to each other. In addition, it is possible to easily detect a defect in the bonding work, and it is possible to reuse a defective product. Further, it is possible to confirm a bonding failure caused by the surface state of the wafer. Moreover, the optimum conditions for achieving normal bonding can be obtained. The bonding defect inspection apparatus disclosed in Patent Document 2 irradiates infrared rays on a bonding jig, and captures and detects infrared rays reflected on the bonded wafer by a camera, and can display a screen in real time.
JP-A-63-139237 JP-A-7-161596

  The inventions of Patent Documents 1 and 2 can also detect voids, which are crystal defects of a silicon wafer, due to a decrease in transmittance or reflectance accompanying infrared absorption. However, since particles adhering to the surface (front surface and back surface) of the silicon wafer also absorb infrared rays, there is a problem that the particles are erroneously detected as voids. In addition, in order to distinguish voids from particles, it is necessary that the detected voids have a diameter of a millimeter order or more, and therefore, the detection limit value of the inspection apparatus is substantially limited. There is. Furthermore, the analog image capturing according to Patent Documents 1 and 2 has a problem that the background noise of infrared images is large, and voids and particles of the order of several tens of microns cannot be distinguished.

  Here, by reducing the background noise, if a void and a particle of the order of several tens of microns can be identified, the false detection rate of the void can be reduced. If the voids and particles can be accurately identified, the silicon wafer can be sent to the next process as a non-defective product by removing the particles by cleaning. In addition, if infrared rays are used to improve throughput and detect voids finer than the current detection limit values shown in Table 1 above, productivity in the inspection process will increase, and silicon will increase. Wafer inspection accuracy can be improved. Furthermore, in the bonding defect inspection of the bonded wafer, it becomes possible to detect a void finer than the current detection limit value. These can be said to be the problems of the present invention.

  The present invention has been made in view of such a problem. By using infrared rays, it is possible to detect a void finer than the current detection limit value while improving throughput, and to identify the void and the particle. An object of the present invention is to provide a crystal wafer defect inspection method and apparatus that can be used.

  The present inventor has found that fine voids and particles can be identified by filtering an image obtained by infrared rays transmitted through a silicon wafer, and based on this, a new silicon wafer as described below can be identified. The inventors have invented a crystal defect inspection method and apparatus.

  (1) A silicon wafer crystal defect inspection method for inspecting a silicon wafer for crystal defects from an image obtained by irradiating the silicon wafer with infrared rays and imaging the infrared rays transmitted through the silicon wafer. A low-magnification inspection step for obtaining a low-magnification inspection image having a fixed visual field area on the silicon wafer using a lens, and a position confirmation step for confirming and storing a crystal defect position by image processing the low-magnification inspection image And using a high-magnification objective lens, the position of the stored crystal defect is positioned within a fixed visual field region, and the objective distance is changed to obtain a high-magnification inspection image of the visual field region Determining the shape of the crystal defect based on the step and the high-magnification inspection image, from the crystal defect shape information, circular or non-circular irregular shape Crystal defect inspection method of a silicon wafer comprising a determining step the crystal defects or voids or particles, with or.

  The crystal defect inspection method for a silicon wafer according to the invention of (1) inspects a crystal defect of the silicon wafer from an image obtained by irradiating the silicon wafer with infrared rays and imaging the infrared rays transmitted through the silicon wafer. First, in the low-magnification inspection step, a low-magnification inspection image having a certain visual field area on the silicon wafer is acquired using a low-magnification objective lens.

  Infrared rays are in the spectral region in the wavelength range of 0.7 μm to 1000 μm, and infrared rays in the wavelength range of 1.0 μm to 1000 μm can pass through the silicon wafer. For example, infrared light is irradiated from a xenon lamp onto a silicon wafer on an XY stage, and the infrared light transmitted through the silicon wafer is imaged by an infrared camera. An infrared digital CCD camera is preferably used as the infrared camera. For the illumination of the xenon lamp, it is preferable that the silicon wafer is irradiated with light of uniform intensity by the light guide plate. For example, the CCD camera can image light having a wavelength up to 1100 nm, and transmitted infrared light having a wavelength range of 1000 nm to 1100 nm is imaged by the CCD camera.

  First, an image of a certain visual field region on a silicon wafer is acquired by a CCD camera using a low-magnification objective lens. The visual field region can be set as appropriate. For example, the visual field region may be about 50 mm □ (□ indicates the length of one side of the square), and the entire measurement region of the silicon wafer is captured by moving the XY stage.

  In the crystal wafer defect inspection method according to the invention of (1), in the next position confirmation step, the low magnification inspection image is subjected to image processing to confirm the position of the crystal defect, and the position is stored. Next, in the high magnification inspection step, using the high magnification objective lens, the position of the stored crystal defect is positioned in a fixed visual field region, the objective distance is changed, and a high magnification inspection image of the visual field region is obtained. get. Next, in the determination step, the shape of the crystal defect is obtained based on the high magnification inspection image. Then, from the shape information of the crystal defect, it is determined whether the crystal defect is a void or a particle with a circular shape or an indefinite shape other than a circular shape.

  In the position confirmation step, image processing is performed on the low-magnification inspection image obtained in the low-magnification inspection step to detect crystal defects. Generally, the crystal defect image obtained from the low-magnification inspection image is small, and it is difficult to determine whether the crystal defect is a void or a particle from its shape. However, the position of the crystal defect obtained in the low-magnification inspection image can be specified from the difference in brightness from the surrounding (background). In the position confirmation step, the position of the crystal defect obtained by the image processing is confirmed, and the position is stored. For example, the center of the silicon wafer to be inspected may be stored in the memory as the coordinates of the position of the crystal defect with the origin of the XY coordinates as the origin.

  In the next high magnification inspection step, the low magnification objective lens is switched to the high magnification objective lens. The low-magnification objective lens and the high-magnification objective lens may be switched by the mechanical moving means without changing the focal point. In the high magnification inspection step, the silicon wafer is moved to the position of the crystal defect stored in the memory. Note that the switching of the objective lens and the movement of the silicon wafer may be performed simultaneously. In the high magnification inspection step, the image of the crystal defect is enlarged, so that the shape can be recognized. For example, in the high-magnification inspection step, a high-magnification inspection image of crystal defects in which the visual field area is enlarged to about 20 mm □ is acquired.

  In the next determination step, the shape of the crystal defect is obtained based on the high magnification inspection image. For example, with only a general binarization process, in the high-magnification inspection image, the outline of the void opened on the objective lens side is clear, and the outline of the void opened on the infrared irradiation side is unclear. However, the outline of crystal defects can be made clear by filtering the high-magnification inspection image described later. From the contour information, it is determined whether the crystal defect is a void or a particle with a circular shape or an indefinite shape other than a circular shape. Crystal defects having a shape close to a circle are determined as voids, and crystal defects having an indefinite shape different from a circle are determined as particles. In automating this determination, it may be determined using a feature extraction process described later whether the crystal defect obtained in the high-magnification inspection image is a circular shape or an indefinite shape other than a circular shape.

  In the method for inspecting a crystal defect of a silicon wafer according to the invention of (1), since the crystal defect position is determined by a high magnification inspection image after the position of the crystal defect is acquired by a low magnification inspection image, silicon having excellent throughput is obtained. A wafer crystal defect inspection method can be provided.

  (2) The high-magnification inspection step includes a luminance correction step for correcting luminance unevenness in the background of the visual field region by image processing, and a binarization processing step for binarizing the image obtained by the luminance correction. (1) The crystal defect inspection method for a silicon wafer according to (1).

  For example, the surface of a silicon wafer is chemically polished, the surface state is rough, and an image pattern is generated in the background in a normal transmitted infrared image, and it is difficult to determine the shape of a fine crystal defect. However, after filtering the high-magnification inspection image obtained in the high-magnification inspection step and correcting the luminance unevenness, it is possible to determine the shape of fine crystal defects by binarizing the image obtained by the luminance correction It becomes. For example, the background filtering process described above can be implemented with existing image processing software.

  The crystal defect inspection method for a silicon wafer according to the invention of (2) corrects the luminance unevenness of the background by filtering the high magnification inspection image obtained in the high magnification inspection step, so that the current detection limit Void finer than the value can be detected.

  (3) In the determination step, based on the high-magnification inspection image, a peripheral length that is the length of the outline of the crystal defect and an area surrounded by the outline are obtained, and a numerical value that is characterized by circularity is a perfect circle In the case of (2), a calculation step for calculating the area perimeter length ratio that is the maximum value and a threshold determination step for determining that the crystal defect is a particle when the area perimeter length ratio is equal to or less than a certain threshold value (2) Crystal defect inspection method for silicon wafers.

For example, when the radius of the perfect circle is “r”, the circumference P of the perfect circle is calculated as P = 2πr. Further, the area S of the perfect circle is calculated by S = πr ² . Here, P ² = (2πr) ² = 4π ² r ² = 4π × πr ² . That is, P ² = 4πS, and a relationship of (4πS) / P ² = 1 is derived. In the calculation step, when the numerical value extracted with the circularity (4π × area / (peripheral length) ² ) × 100 is a perfect circle, the area perimeter length ratio that is the maximum value “100” is calculated, and in the threshold determination step The crystal defect is determined to be a particle when the area perimeter length ratio is equal to or less than a certain threshold value. The crystal defect may be determined as a void when the area perimeter length ratio is equal to or greater than a certain threshold value.

  As a result of implementation, the threshold value is equivalent to “63”, and the expansion of the visual field based on the high-magnification inspection image is about 20 mm □, and the depth of focus is preferably equal to or greater than the thickness of the silicon wafer, as will be described later. It is considered that the area perimeter length ratio based on image analysis can be calculated with higher accuracy by obtaining a visual field area in which the visual field area is enlarged to about 20 mm □ or more.

  (4) The silicon wafer crystal defect inspection method according to any one of (1) to (3), wherein the high-power objective lens has a focal depth equal to or greater than a thickness of the silicon wafer.

  In the crystal defect inspection method for a silicon wafer according to the invention of (4), since the depth of focus of the high-magnification objective lens is equal to or greater than the thickness of the silicon wafer, the void opened to the objective lens side without blurring the outline of the void, The outlines of the voids opened on the infrared irradiation side and the voids inside the silicon wafer can be clearly captured.

  (5) The silicon wafer according to any one of (1) to (4), wherein the silicon wafer can determine whether the crystal defect is a void or a particle with an electric resistivity of 6 mΩ · cm as a lower limit. Method.

  Silicon wafers are said to have different infrared absorption levels due to differences in electrical resistivity, that is, the infrared absorption levels of dopants (impure additives) contained in silicon wafers are different. In the method for inspecting a crystal defect of a silicon wafer according to the invention, it was demonstrated that the silicon wafer to be inspected can determine whether the crystal defect is a void or a particle with an electric resistivity of 6 mΩ · cm as a lower limit. A silicon wafer having a high resistivity exceeding 6 mΩ · cm easily transmits infrared rays, so that it is possible to determine whether a crystal defect is a void or a particle.

  (6) The silicon wafer crystal defect inspection method according to any one of (1) to (3), wherein the silicon wafer includes a bonded wafer, and the crystal defect includes a bonded defect of the bonded wafer.

  Here, the void includes the meaning of bubbles that become non-bonded portions on the overlapping surface of the bonded wafers, and also includes the meaning of voids that become crystal defects in the silicon wafer. Regardless of how these occur, the void is a hollow portion remaining on the wafer, and both are captured in a circular shape in an infrared image. Then, by using the silicon wafer crystal defect inspection method described in any one of (1) to (3), it is possible to detect voids in the bonded wafer with higher sensitivity than in the conventional method.

  (7) A silicon wafer crystal defect inspection apparatus for inspecting a silicon wafer for crystal defects from an image obtained by irradiating the silicon wafer with infrared rays and imaging the infrared rays transmitted through the silicon wafer. A low-magnification objective lens that acquires a low-magnification inspection image having a fixed visual field area, a position confirmation unit that performs image processing on the low-magnification inspection image to confirm and store the position of a crystal defect, and the stored crystal defect Is positioned in a fixed visual field region, a high-magnification objective lens that acquires a high-magnification inspection image of the visual field region, and the shape of the crystal defect is obtained based on the high-magnification inspection image, Determination means for determining whether the crystal defect is a void or a particle with a circular shape or an irregular shape other than a circular shape from defect shape information. Crystal defect inspection apparatus of N'weha.

  In the crystal defect inspection apparatus for a silicon wafer according to the invention of (7), the position confirmation means includes an XY stage on which the silicon wafer is placed, and includes a control device for commanding and controlling the XY stage to be movable. Further, the position confirmation means may include an image processing apparatus that includes a CCD camera that captures infrared light transmitted through the silicon wafer through a low-magnification objective lens, and that processes an image signal captured by the CCD camera. The position of the crystal defect of the silicon wafer specified by the image processing apparatus is stored as coordinates in a memory provided in the control apparatus.

  In the crystal defect inspection apparatus for a silicon wafer according to the invention of (7), the determination means converts the image signal picked up by the CCD camera into a digital signal, and executes a filtering process and a binarization process, A feature extraction unit that extracts the shape of the crystal defect based on the magnification inspection image; and a determination / output unit that is determined based on a reference for which the extraction result is set. The preprocessing unit, the feature extraction unit, and the determination / output unit can be provided in the image processing apparatus. The determination means includes an image processing program that executes a series of these processes, and may include image processing software that performs background filtering.

  In the method for inspecting a crystal defect of a silicon wafer according to the present invention, since the position of the crystal defect is acquired with a low-magnification inspection image, the shape of the crystal defect is determined with a high-magnification inspection image. A defect inspection method can be provided. Further, the silicon wafer crystal defect inspection method according to the present invention filters the high-magnification inspection image obtained in the high-magnification inspection step, so that it is possible to detect voids finer than the current detection limit value.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is a main flowchart showing the procedure of an embodiment of the method for inspecting a crystal defect of a silicon wafer according to the present invention. FIG. 2 is a flowchart showing a procedure in the high magnification inspection step according to the embodiment. FIG. 3 is a flowchart showing the procedure of the determination step according to the embodiment.

  FIG. 4 is a cross-sectional view of a silicon wafer to be inspected. FIG. 5 is a block diagram showing an example of an apparatus for carrying out the silicon wafer crystal defect inspection method according to the present invention. FIG. 6 is a diagram showing an example of an inspection image acquired by a CCD camera. 6A shows an image of a crystal defect acquired by a low magnification inspection image, FIG. 6B shows an image of a void acquired by a high magnification inspection image, and FIG. 6C shows a high magnification inspection. The image of the particle acquired by the image is shown. FIG. 7 is a graph showing the relationship between the position and the luminance in the transmitted infrared of a silicon wafer acquired with a high-magnification inspection image. In FIG. 7, the vertical axis indicates luminance, and the horizontal axis indicates the distance from the origin in the number of pixels.

  FIG. 8 is a graph showing the relationship between the position and the luminance in transmitted infrared rays of a plurality of silicon wafers having low resistivity. The vertical axis in FIG. 8 indicates luminance, and the horizontal axis indicates the distance from the origin in terms of the number of pixels. FIG. 9 is a graph showing the relationship between the position and luminance in the transmitted infrared of a silicon wafer having a plurality of normal resistivities. The vertical axis in FIG. 9 indicates luminance, and the horizontal axis indicates the distance from the origin in terms of the number of pixels. FIG. 10 is a graph showing the relationship between the position and the luminance in the transmitted infrared of silicon wafers having different resistivities. The vertical axis in FIG. 10 indicates luminance, and the horizontal axis indicates the distance from the origin in terms of the number of pixels.

  First, the configuration of an apparatus for carrying out the crystal defect inspection method for a silicon wafer according to the present invention will be described. In FIG. 4, the silicon wafer W to be inspected is formed in a disk shape having a diameter of 200 mm, for example. The silicon wafer W may have a spherical void V formed with crystal growth. In the silicon wafer W obtained by slicing the ingot, a void V exists in the silicon wafer W, a void V appears so as to open in the front surface W1 of the silicon wafer W, or a void exists in the back surface W2 of the silicon wafer W. Or V appears to open. In FIG. 4, the front surface W1 and the back surface W2 of the silicon wafer W are relative distinctions, and are simply distinguished with reference numerals for convenience of explanation.

  On the other hand, in FIG. 4, indefinitely shaped particles Pa may adhere to the front surface W1 or the back surface W2 of the silicon wafer W. In general, the silicon wafer W in which even one void V is detected is discarded as a crystal defect product. However, since the silicon wafer W to which the particle Pa is attached can be removed by cleaning, the silicon wafer W is treated as a good product in the next process. It is also possible to send to. Therefore, it is necessary to clearly identify the void V and the particle Pa and inspect the silicon wafer W. In the following description, the crystal defect to be inspected indicates an abnormal part in the previous stage where the void V and the particle Pa are identified.

  In FIG. 5, the silicon wafer W is placed on the XY stage 11. The XY stage 11 can be moved independently in the XY directions on the gantry 12 by a command from the control device 17. By moving the XY stage 11, the entire measurement region of the silicon wafer W is captured. The light source 13 irradiates the silicon wafer W on the XY stage 11 with infrared rays. For example, a xenon lamp is used as the light source 13, and the xenon lamp is preferably illuminated with light of uniform intensity on the silicon wafer W by a light guide plate (not shown).

  In FIG. 5, the infrared light transmitted through the silicon wafer W is imaged by the infrared camera 15 through the low-magnification objective lens 14a or the high-magnification objective lens 14b. The low-magnification objective lens 14a and the high-magnification objective lens 14b are switched by the mechanical moving means without changing the focal point. The objective lens 14 a and the objective lens 14 b are switched by a command from the control device 17. An infrared digital CCD camera is preferably used as the infrared camera 15. In the embodiment, the infrared CCD camera 15 can image light having a wavelength of up to 1100 nm, and the infrared CCD camera 15 captures transmitted infrared light having a wavelength range of 1000 nm to 1100 nm.

  In FIG. 5, the image signal captured by the CCD camera 15 is sent to the image processing device 16. The image signal picked up by the CCD camera 15 is converted into a digital signal by the pre-processing unit 16a. For example, this image data is transmitted to the feature extraction unit 16b after filtering processing and binarization processing are executed. The feature extraction unit 16b extracts the shape of the crystal defect based on the high-magnification inspection image as will be described later. Then, the determination / output unit 16 c makes a determination based on the set reference and outputs the determination result to the control device 17.

  In FIG. 5, the image processing device 16 may include an image processing program that executes a series of processes, and may include image processing software that performs background filtering processing described later. The image processing device 16 can also be configured to include a memory for storing an image processing program and image processing software. The control device 17 may be provided with a memory (not shown) in which the position of the crystal defect is stored as coordinates. The monitor device 18 can display the image processed result and the inspection result.

  Next, the procedure of the silicon wafer crystal defect inspection method according to the present invention will be described. In FIG. 1, first, in step S1, an image of a fixed visual field region on the silicon wafer W is acquired by the infrared CCD camera 15 using the low-magnification objective lens 14a. FIG. 6A shows a low-magnification inspection image obtained using the low-magnification objective lens 14a. The visual field area X × Y of the low-magnification inspection image can be set arbitrarily, but the visual field area is preferably about 50 mm □, and the entire measurement area of the silicon wafer W is captured by moving the XY stage 11. The Here, step S1 is referred to as a low magnification inspection step S1.

  Next, in step S2, the low-magnification inspection image is subjected to image processing to confirm the presence or absence of crystal defects in the low-magnification inspection image. In step S2, when no crystal defect is detected in the low magnification inspection image, the process returns to the previous stage of the low magnification inspection step S1. That is, by moving the XY stage 11, the next low-magnification inspection image is acquired. When no crystal defect is detected in the low-magnification inspection image in the entire measurement region of the silicon wafer W, the inspection is terminated, and the silicon wafer W is transferred to the next process as a non-defective product. In FIG. 1, illustration of the procedure after acquisition of all the low-magnification inspection images is omitted. In step S2, when a crystal defect is detected in the low magnification inspection image, the process proceeds to step S3.

  Next, in step S3, the low magnification inspection image is subjected to image processing to confirm the position of the crystal defect, and the position is stored. For example, as shown in FIG. 6A, the image of the crystal defects D1 and D2 obtained from the low-magnification inspection image is small, and it is difficult to determine whether the crystal defects D1 and D2 are voids or particles based on their shapes. It is. However, the positions of the crystal defects D1 and D2 obtained in the low-magnification inspection image can be specified from the difference in brightness from the periphery in the low-magnification inspection image. The position is stored in the memory in the control device 17 as the coordinates of the XY stage 11. Here, step S3 is referred to as position confirmation step S3. Next, the process proceeds to step S4.

  In the next step S4, using the high-magnification objective lens 14b, the position of the stored crystal defect is positioned within a fixed visual field region, the objective distance is changed, and a high-magnification inspection image of this visual field region is acquired. Is done. In step S4, the low magnification objective lens 14a is switched to the high magnification objective lens 14b (see FIG. 5). Here, the low-magnification objective lens 14a and the high-magnification objective lens 14b are switched by the mechanical movement means according to a command from the control device 17 without changing the focal point. In step S <b> 4, the silicon wafer W is moved to the position of the crystal defect stored in the memory in the control device 17. Note that switching between the low-magnification objective lens 14a and the high-magnification objective lens 14b and the movement of the silicon wafer W may be performed simultaneously.

  In step S4, since the crystal defect image is enlarged, the shape can be recognized. For example, FIGS. 6B and 6C show high-magnification inspection images obtained using the high-magnification objective lens 14b. The visual field area xxy of the high-magnification inspection image can be arbitrarily set, but the visual field area is preferably about 20 mm □, and the XY stage 11 is moved to move the visual field area XXY in the low-magnification inspection image. Is captured. Here, step S4 is referred to as a high magnification inspection step S4. Next, the process proceeds to step S5.

  In the next step S5, the shape of the crystal defect is obtained based on the high magnification inspection image. For example, with only a general binarization process, a high-magnification inspection image has a clear outline of a void V (see FIG. 4) opened on the objective lens side, and a void V opened on the infrared irradiation side (FIG. 4). The contour of (see) was unclear. However, by appropriately filtering the high-magnification inspection image, the outline of crystal defects can be clear. From the contour information, it is determined whether the crystal defect is a void or a particle with a circular shape or an indefinite shape other than a circular shape. As shown in FIG. 6B, the crystal defect D1 having a nearly circular shape is determined as a void. On the other hand, as shown in FIG. 6C, the crystal defect D2 having an indefinite shape different from the circular shape is determined as a particle. Here, step S5 is referred to as determination step S5. In automating the determination step S5, it is possible to determine whether the crystal defect obtained by the high-magnification inspection image is a circle or an indeterminate shape other than a circle using a calculation formula described later. Then, the process proceeds to step S6.

  In the next step S6, the presence or absence of voids in the high-magnification inspection image is confirmed. When no void is detected in the high magnification inspection image, the process returns to the previous stage of the high magnification inspection step S4. That is, by moving the XY stage 11, the next high magnification inspection image is acquired. When no void is detected in all the high-magnification inspection images on the silicon wafer W, the inspection is terminated, and the silicon wafer W is cleaned and transferred to the next process as a non-defective product, for example. In FIG. 1 to FIG. 3, the procedure after acquisition of all the high-magnification inspection images is not shown. In step S6, when a void is detected in the high-magnification inspection image, a series of inspections is terminated, and this silicon wafer W is discarded as a defective product, for example.

  As described above, the method for inspecting a crystal defect of a silicon wafer according to the present invention is excellent in throughput because the crystal defect shape is determined by a high magnification inspection image after the position of the crystal defect is acquired by a low magnification inspection image. A crystal defect inspection method for a silicon wafer can be provided.

  Next, details of the high magnification inspection step S4 will be described. In FIG. 2, in step S41 following the position confirmation step S3, a high-magnification inspection image is acquired using the high-magnification objective lens 14b. Next, in step S42, background luminance unevenness is corrected by the preprocessing unit 16a (see FIG. 5). Here, step S42 is referred to as luminance correction step S42. Next, in step S43, the high-magnification inspection image in which the background luminance unevenness is corrected is binarized by the pre-processing unit 16a (see FIG. 5). Here, step S43 is referred to as binarization processing step S43. In a determination step S5 following the binarization processing step S43, the shape of the crystal defect is determined from the binarized high magnification inspection image.

  For example, the surface state of the silicon wafer W is rough, and an image group is generated in the background in a normal transmitted infrared image, and it is difficult to determine the shape of a fine crystal defect. However, after the high-magnification inspection image obtained in the high-magnification inspection step S4 is filtered to correct luminance unevenness, the image obtained by the luminance correction is binarized to determine the shape of fine crystal defects. It was proved that it was possible. For example, existing image processing software can be used, and by incorporating this into the image processing device 16 (see FIG. 5), the background luminance unevenness can be corrected by this image processing software.

  FIG. 7 is an example showing a luminance profile (contour) of each cross section of a void and a particle in a high-magnification inspection image after correcting the luminance unevenness of the background. In FIG. 7, the broken line Sv indicates the luminance profile of the silicon wafer having voids, and the broken line Sp indicates the luminance profile of the silicon wafer having particles. As shown in FIG. 7, the brightness profiles of voids and particles are significantly lower in brightness than the surroundings, and the contour can be detected. However, the polygonal line Sv and the polygonal line Sp are substantially similar, and it is difficult to distinguish the shape from each brightness profile of the void and the particle, and it is preferable to distinguish the shape using a calculation formula and a threshold as described later. . In the broken line Sv, a phenomenon in which the luminance is increased at the center of the void is observed, but such a phenomenon is not observed in the broken line Sp of the particle. It can be considered that voids and particles can be distinguished from such a difference in phenomenon.

  As described above, the crystal wafer defect inspection method according to the invention corrects the luminance unevenness of the background by filtering the high magnification inspection image obtained in the high magnification inspection step. Void finer than the value can be detected.

Next, details of the determination step S5 will be described. In FIG. 3, in step S51 following the high-magnification inspection step S4, based on the high-magnification inspection image, the peripheral length that is the length of the contour of the crystal defect and the area surrounded by this contour are obtained, and (4π × area / ( Perimeter length) When the numerical value calculated with a circularity of ² ) × 100 is a perfect circle, the area perimeter length ratio at which the maximum value is “100” is calculated. Here, step S51 is referred to as calculation step S51.

  Next, in step S52, the crystal defect is determined to be a particle when the area perimeter length ratio is equal to or smaller than a certain threshold value. The crystal defect may be determined as a void when the area perimeter length ratio is equal to or greater than a certain threshold value. As a result of the implementation, this threshold value is assumed to be “63”, but can be set as appropriate. Here, step S52 is referred to as threshold determination step S52, and the result of threshold determination step S52 is sent to step S6. By calculating the area perimeter length ratio and comparing it with a predetermined threshold value, void inspection can be automated.

  Next, using a semiconductor internal test apparatus using infrared rays and image processing software, the void size and position, the influence of the resistivity of the silicon wafer, and the discriminable limit of particles that can be identified by the method of the present invention were investigated. The results will be explained.

  As a semiconductor internal test apparatus using infrared rays, an existing semiconductor internal test apparatus was used. Using this semiconductor internal test equipment, a silicon wafer on a manual stage is illuminated with a xenon lamp, and infrared rays that have passed through the silicon wafer are imaged with a camera. Voids opened in the surface or voids generated inside the silicon wafer were observed (see FIG. 4). As the camera, an existing infrared digital CCD camera was used. The CCD camera can detect transmitted infrared rays in the wavelength range of 1000 nm to 1100 nm, and has a feature of high sensitivity because of its high S / N ratio. The CCD camera having a high S / N ratio does not exist in the performance at the time when the inventions of Patent Documents 1 and 2 were made, and the improvement of the S / N ratio and conversion to a digital signal by this infrared digital CCD camera are possible. It is considered that this was a factor that sufficiently exerted the performance of subsequent image processing.

  Existing image processing software can be used as the image processing software. This image processing software is software that can extract feature quantities such as defects of a silicon wafer by the following procedure and can be used for defect inspections such as spot and unevenness of a planar object. In addition, the background filtering process described above is performed as a pre-process, and after extracting the inspection area, luminance unevenness is corrected by processes such as shading correction, contrast enhancement, and repeated pattern removal. Any existing product can be used as long as it performs a proper process. This time, we selected the appropriate ones.

  The procedure of image processing by the image processing software is as follows.

(1) Pre-processing: inspection area extraction, shading correction, contrast enhancement, repeated pattern removal, etc. (2) Defect inspection processing: filters (all 20 types such as average, maximum, minimum), feature extraction filter, binarization, 2 Value image processing, feature amount calculation (defect coordinates, area, density, perimeter, etc.)
(3) Result data processing: coordinate conversion, OK / NG determination (number, area, distance), data output (screen, file)

  Next, Table 2 shows the characteristics of the investigated silicon wafer (sample). In Table 2, the void diameter is the aperture diameter, which is a value measured with an optical microscope. Hereinafter, samples 1 to 4 may be referred to as low resistivity silicon wafers, and samples 5 to 8 may be referred to as normal resistivity silicon wafers.

  The void of sample 1 is placed on the front side (CCD camera side), sample 1 is placed on the stage of the inspection device, and transmitted infrared light is captured in the 52 mm square field of view. Since it was blocked, the brightness was lower than that of the surrounding area, and the image was observed as a dark image. However, the front-side particles also blocked the transmitted infrared light and were observed as dark images. Since there is no difference in looking at the shape by enlarging the void and particle images as they are, it is not possible to distinguish the void from the particles.

  Therefore, when the low-magnification inspection image of the 52 mm □ field of view is changed to the high-magnification inspection image of the 21 mm □ field of view and the image capturing the transmitted infrared light is observed, both the void and particle images are dark. However, in the void image, bright spots were observed inside a substantially circular shape, whereas the particles were observed as dark amorphous images. Therefore, it is estimated that the shape can be distinguished in the high-magnification inspection image.

  When the void opening of a low resistivity silicon wafer (samples 1 to 4) is on the front side and a high-magnification inspection image of a 21mm □ field of view is observed, it is possible to distinguish the shape of the void and the particle. However, it was difficult to distinguish the shape of the void and the particle in a visual field region of 21 mm □ or larger. Since this result was the same as the samples 1 and 2 on the mirror surface and the samples 3 and 4 on the chemically polished surface, it is considered that the detection of voids by transmitted infrared light does not depend on the state of the silicon wafer surface. Further, even when the void opening of the low resistivity silicon wafer (samples 1 to 4) was observed with the back side, the same result as described above was shown.

  Next, the void opening of sample 1 is set to the back side (irradiation side), sample 1 is placed on the stage of the inspection apparatus, and the image obtained by capturing the transmitted infrared light in the 52 mm □ field of view is the back surface of sample 1 Both the void and particle images were observed as dark images, but the shapes of the images were determined to be indistinguishable.

  Furthermore, when the low-magnification inspection image of the 52 mm □ field of view in sample 1 is changed to the high-magnification inspection image of the 21 mm □ field of view and the image capturing the transmitted infrared light is observed, voids and particles on the front surface Each of the shapes could clearly capture the outline, but the void and particle shapes on the back surface were unclear and could not be captured. Therefore, it is considered that a device for image processing is required to distinguish the shapes of voids and particles.

  A normal silicon wafer (samples 5 to 8) with a void opening on the front side and an image captured of transmitted infrared light is observed. Sample 5 with a void diameter of 61 μm has a 52 mm square field of view. The position of the void can be confirmed by the low magnification inspection image of the region, and the shape of the void can be confirmed by the high magnification inspection image of the visual field region of 21 mm □. However, in the sample 6 having a void diameter of 37 μm, the position of the void could not be confirmed in the visual field region of 52 mm □. The samples 7 and 8 having void diameters of 67 μm and 84 μm are considered to be difficult to distinguish from the particles because the shape is unclear for the size of the voids.

  Furthermore, when the void opening of the silicon wafer with normal resistivity (samples 5 to 8) is set to the back side and the image capturing the transmitted infrared light is observed, sample 5 with a void diameter of 61 μm is 52 mm □. The position of the void can be confirmed in the visual field region of FIG. 1. In the visual field region of 21 mm □, the outline of the void was not always clear, but the circular shape seems to be confirmed. However, in the sample 6 having a void diameter of 37 μm, the position of the void could not be confirmed in the visual field region of 52 mm □.

  FIGS. 8 to 10 show graphs obtained by extracting the luminance profiles of the respective cross sections from the void and particle images observed in the high-magnification inspection image of the 21 mm □ field-of-view region among the above observation images. In FIG. 8A, the brightness profile of each void is drawn with the void opening of the low resistivity silicon wafer (samples 1 to 4) facing the front side. In FIG. 8B, the luminance profile of each void is depicted with the void opening of the low resistivity silicon wafer (samples 1 to 4) on the back side. The polygonal lines Sm1 to Sm4 shown in FIG. 8 correspond to the voids opened in the samples 1 to 4, respectively.

  In FIG. 9A, the brightness profile of each void is depicted with the void opening of the normal resistivity silicon wafer (samples 5, 6, 8) facing the front. In FIG. 9B, the brightness profile of each void is depicted with the void opening of the silicon wafer of normal resistivity (samples 5, 6 and 8) on the back side. The polygonal lines Sm5, 6, and 8 shown in FIG. 9 correspond to the voids opened in the samples 5, 6, and 8, respectively.

  The semiconductor test equipment used for image observation automatically corrects the display range when converting the captured luminance difference to black and white, so the image depends on the difference in resistivity of the silicon wafer. There is no difference in brightness. However, it was observed that the resistivity of the silicon wafer was lowered, the intensity of the transmitted infrared light was weakened, and the contrast between the void and its periphery was reduced (see FIGS. 8 and 9).

  In FIG. 10A, the luminance profile of each particle of Samples 1 and 5 is drawn with the particle on the front side. In FIG. 10B, the luminance profile of each particle of Samples 1 and 5 is drawn with the particle on the back side. The polygonal lines Sm1 · 5 shown in FIG. 10 correspond to the particles adhering to the samples 1 · 5. As shown in FIGS. 8 to 10, the luminance profile of the particles is similar to the luminance profile of the void, and it is considered difficult to distinguish the void and the particle using the luminance profile. In some samples, a phenomenon was observed in which the brightness profile of the void increased in the center, but in all samples, this phenomenon did not occur in the brightness profile of the void.

  As a result of examining the feasibility of void detection based on the above observation results, the following detection procedure was validated.

  First, the position (coordinates) of crystal defects (voids or particles) is confirmed using a low-magnification lens (fixed focus). Next, an enlarged image of the crystal defect is captured using a high-magnification lens (focus fixed), and it is determined whether it is a void or a particle from its shape. When the void is detected, the silicon wafer is regarded as a defective product, and the silicon wafer where no void is detected is regarded as a non-defective product.

  However, as a factor that hinders the distinction between crystal defect shapes, the parallelism of the backlight is inadequate, so the outline of the voids that open on the front side gives a clear image, but the back side It is mentioned that the outline of the voids opened in the image becomes an unclear image. Furthermore, the difference in the transmittance of infrared rays due to the difference in the resistivity of the silicon wafer and the difference in the backlight intensity depending on the site may cause uneven brightness in the background of the image. As these countermeasures, firstly, securing the number of pixels of 10 pixels or more can be mentioned. If the number of pixels of 10 pixels or more can be ensured, even a somewhat unclear image can be determined by image processing, so a high-magnification inspection image of a viewing area of 21 mm □ or more can be obtained. Next, it is possible to correct luminance unevenness by image processing. And it is necessary to examine image analysis.

  For example, sample 1 is a silicon wafer having a low resistivity and hardly transmits infrared light, so that the brightness contrast when detecting voids is small (see FIG. 8). Therefore, if particles similar to the void are attached to the sample 1, it is considered that it is difficult to extract the void by image analysis. Here, in the case where the sample 1 has voids and particles, an attempt was made to extract only voids using the above-described image processing software. The results will be described below.

  The presence or absence of background luminance correction and the influence on the subsequent binarization processing were examined for an image obtained by imaging a void that opened on the front side of sample 1. When the background luminance was not corrected, the positions of voids and particles could not be specified even when the binarization process was performed due to uneven luminance of the background. On the other hand, when the background luminance was corrected, the positions of voids and particles could be specified after the binarization process. However, the void and particle images were displayed as bright spots by inverting the contrast.

  With the void opening of sample 1 facing the front side, sample 1 was placed on the stage of the inspection device, and the image of the captured infrared light captured in the 52 mm square field of view was analyzed. Whereas it was 68.72 ", two area perimeter ratios of particles," 52.97 "and" 76.93 ", were detected. The area perimeter ratio is 100 in a perfect circle, and it is difficult to extract only voids in the measurement of the area perimeter ratio by image analysis of the low-magnification inspection image.

  With the void opening of sample 1 facing the front side, sample 1 was placed on the stage of the inspection apparatus, and the image of the captured infrared light captured in the 21 mm square field of view was analyzed. The area perimeter length ratio of the particles was “54.82” while it was “63.02”. From this result, it is considered possible to extract voids that are open to the front side by measuring the area perimeter length ratio by image analysis of the high-magnification inspection image.

  Furthermore, the area perimeter length ratio of the particles adhering to the front side of Sample 1 is calculated as “76.93” in the 52 mm □ field of view, and is calculated as “54.82” in the 21 mm □ field of view. By changing the area perimeter length ratio is different. However, the area perimeter length ratio of the voids opened on the front side of the sample 1 is calculated as “68.72” in the 52 mm □ visual field region and is calculated as “63.02” in the 21 mm □ visual field region. A stable value with little fluctuation was calculated. For this reason, it is considered possible to extract a void that opens on the front side even if the magnification is changed.

  Next, the void opening of sample 1 was placed on the back side, sample 1 was placed on the stage of the inspection apparatus, and the image of the captured infrared light captured in the 52 mm □ field of view was subjected to image analysis. The ratio was “64.11”, whereas the area perimeter length ratio of the particles adhering to the back side was “74.54”. The area perimeter ratio is reversed from the actual situation of voids and particles, and it is difficult to extract only voids that are open on the back side in the measurement of the area perimeter ratio by image analysis of the low magnification inspection image.

  The void of sample 1 was placed on the back side, sample 1 was placed on the stage of the inspection device, and the image of the transmitted infrared light captured in the 21 mm square field of view was analyzed. The area perimeter length ratio of the particles was “60.87” while it was “63.81”. From this result, it is considered possible to extract voids opened on the back side by measuring the area perimeter length ratio by image analysis of the high magnification inspection image.

  From the above image analysis results, the following image processing procedure is considered appropriate for the extraction of voids.

  Using a high-magnification lens (fixed focus), an enlarged image of a crystal defect is captured, and uneven brightness in the background of the enlarged image is corrected. Next, the corrected image is binarized. Next, the peripheral length (contour length) and area of the crystal defect are obtained from the binarized image, and the area perimeter ratio of the crystal defect is calculated. And when this calculated value is more than a predetermined threshold (for example, 63), it determines with a void.

Based on the observation result of the above transmitted infrared light and the image analysis result, an automatic void inspection apparatus according to the following specifications can be realized. In addition, the lower limit value of the resistivity of samples 1 to 8 is 6 mΩ · cm, and it was proved that it is possible to determine whether the crystal defect is a void or a particle in all the samples. Since a silicon wafer having a high resistivity exceeding 6 mΩ · cm easily transmits infrared rays, it can be determined whether the crystal defect is a void or a particle.
Void size: diameter 80μm or more Void position: surface (front and back surfaces) and internal silicon wafer standard: detectable in chemical polishing state,
The resistivity is usually the resistivity (several Ω · cm) and
For low resistivity (several Ω · cm) Throughput: 1 min / sheet or less

  As described above, the method for inspecting the crystal defect of the silicon wafer using infrared rays has been described. However, if the apparatus for carrying out the method for inspecting the crystal defect of the silicon wafer according to the present invention is applied, voids generated in the silicon wafer are detected. The distribution can also be measured, and the yield in the previous process can be improved by analyzing the distribution of voids.

It is a main flowchart which shows the procedure of one Embodiment in the crystal defect inspection method of the silicon wafer by this invention. It is a flowchart which shows the procedure in the high magnification test | inspection step by the said embodiment. It is a flowchart which shows the procedure of the determination step by the said embodiment. It is sectional drawing of the silicon wafer inspected. It is a block diagram which shows an example of the apparatus for enforcing the crystal defect inspection method of the silicon wafer by this invention. It is the figure which showed an example of the test | inspection image acquired with a CCD camera. It is a graph which shows the relationship between the position in the transmitted infrared rays of the silicon wafer acquired with a high-magnification inspection image, and a brightness | luminance. It is a graph which shows the position and the brightness | luminance in the transmitted infrared rays of the silicon wafer which has several low resistivity. It is a graph which shows the position and the brightness | luminance in the transmitted infrared rays of the silicon wafer which has a some normal resistivity. It is a graph which shows the position and the brightness | luminance in the transmission infrared of the silicon wafer which has a different resistivity.

Explanation of symbols

14a Low magnification objective lens 14b High magnification objective lens S1 Low magnification inspection step S3 Position confirmation step S4 High magnification inspection step S5 Determination step Pa Particle V Void W Silicon wafer

Claims (7)

A silicon wafer crystal defect inspection method for inspecting a silicon wafer for crystal defects from an image obtained by irradiating the silicon wafer with infrared light and imaging the infrared light transmitted through the silicon wafer,
A low-magnification inspection step for obtaining a low-magnification inspection image having a constant visual field area on the silicon wafer using a low-magnification objective lens;
A position confirmation step for confirming and storing the position of the crystal defect by image processing the low-magnification inspection image;
A high-magnification inspection step of obtaining a high-magnification inspection image of the visual field region by using a high-magnification objective lens to change the objective distance by positioning the position of the stored crystal defect in a fixed visual field region; ,
Determining the shape of the crystal defect based on the high-magnification inspection image, and determining from the shape information of the crystal defect whether the crystal defect is a void or a particle in a circular or indeterminate shape other than a circle. Crystal defect inspection method for silicon wafers.   The high-magnification inspection step includes a luminance correction step of correcting background luminance unevenness of the visual field region by image processing, and a binarization processing step of binarizing the image obtained by the luminance correction. The method for inspecting crystal defects in a silicon wafer according to claim 1.   In the determination step, based on the high-magnification inspection image, a peripheral length that is the length of the outline of the crystal defect and an area surrounded by the outline are obtained, and when the numerical value that is extracted by the circularity is a perfect circle, The silicon wafer according to claim 2, comprising: a calculation step for calculating an area perimeter length ratio that is a maximum value; and a threshold determination step for determining that the crystal defect is a particle when the area perimeter length ratio is equal to or less than a predetermined threshold value. Crystal defect inspection method.   The silicon wafer crystal defect inspection method according to claim 1, wherein the high-power objective lens has a focal depth equal to or greater than a thickness of the silicon wafer.   The silicon wafer crystal defect inspection method according to claim 1, wherein the silicon wafer can determine whether the crystal defect is a void or a particle with an electric resistivity of 6 mΩ · cm as a lower limit.   4. The method for inspecting a crystal defect of a silicon wafer according to claim 1, wherein the silicon wafer includes a bonded wafer, and the crystal defect includes a bonded defect of the bonded wafer. A silicon wafer crystal defect inspection apparatus that inspects the silicon wafer for crystal defects from an image obtained by irradiating the silicon wafer with infrared light and imaging the infrared light transmitted through the silicon wafer,
A low-magnification objective lens that obtains a low-magnification inspection image having a constant field of view in the silicon wafer;
Position confirmation means for confirming and storing the position of the crystal defect by image processing the low-magnification inspection image;
A high-magnification objective lens that positions the stored crystal defects in a fixed visual field area and obtains a high-magnification inspection image of the visual field area;
Determining means for determining the shape of the crystal defect based on the high-magnification inspection image, and determining whether the crystal defect is a void or a particle in an indefinite shape other than a circle or a circle from the shape information of the crystal defect; Crystal defect inspection system for silicon wafers.

Images