JP5506243B2 - Defect inspection equipment - Google Patents

Description

  The present invention relates to a surface inspection apparatus and method for discriminating and inspecting defects such as scratches and foreign matters generated in a flattening process using a polishing or grinding technique used when manufacturing a semiconductor or a magnetic head.

  As conventional techniques for discriminating foreign substances adhering to a semiconductor wafer on which a circuit pattern is formed from circuit patterns and inspecting them, Japanese Patent Laid-Open Nos. 3-102248 (prior art 1) and Japanese Patent Laid-Open No. 3-102249 (prior art 2). )It has been known. That is, in the prior arts 1 and 2, the foreign matter on the semiconductor substrate is emphasized by oblique illumination and detected by the first photoelectric conversion element, and the edge of the circuit pattern which is the background on the semiconductor substrate is reflected by epi-illumination. The foreign matter detection signal is detected by the second photoelectric conversion element with emphasis and the foreign matter detection signal obtained from the first photoelectric conversion element is divided by the detection signal obtained from the second photoelectric conversion element. It is described that the above foreign matter is detected with emphasis on the above.

  Japanese Patent Laid-Open No. 9-304289 (Prior Art 3) is known as a prior art for separating and inspecting foreign matter adhering to the surface of a silicon wafer and crystal defects existing on the surface. That is, this prior art 3 has a low-angle light receiving system having an elevation angle of 30 ° or less with respect to the surface of the silicon wafer, and a high-angle light receiving system having an elevation angle larger than this, Scattered light obtained by irradiating the surface of the silicon wafer substantially perpendicularly is received by the low-angle light receiving system and the high-angle light receiving system, and a crystal defect is received only by the high-angle light receiving system. In addition, it is described that the light received by the low-angle light-receiving system and the high-angle light-receiving system is discriminated as an attached foreign substance and inspected.

  Japanese Laid-Open Patent Application No. 11-142127 discloses a conventional technique for distinguishing and inspecting foreign matters and scratches existing on the surface of a semiconductor wafer without misidentifying a minute dot-like recess that does not become an obstacle when creating a circuit pattern. (Prior Art 4) is known. That is, in the prior art 4, each of two different wavelengths of illumination light is condensed and irradiated to the same point on the surface of the semiconductor wafer at different low and high incident angles. The scattered light from the light spot is received separately for two wavelengths and photoelectrically converted, and the difference in intensity between the signals, that is, the intensity of the scattered light by the illumination light at a low incident angle is weakened from the point-shaped recess is utilized. Thus, it is described that a foreign substance or a flaw existing on the surface of a semiconductor wafer is inspected separately from a spot-like recess.

Japanese Patent Laid-Open No. 3-102248 Japanese Patent Laid-Open No. 3-102249 JP-A-9-304289 Japanese Patent Laid-Open No. 11-142127

  By the way, as a planarization technique used for an object to be processed (for example, an insulating film) when manufacturing a semiconductor or a magnetic head, a typical one is CMP (Chemical Mechanical Polishing). This CMP is a flattening technique in which free abrasive grains such as silica are dispersed on a polishing pad to polish the surface of a workpiece. Further, as a planarization processing technique, there is a case where a grinding technique is used in which fixed abrasive grains such as diamond are embedded in a polishing pad and grinding is performed in the same manner. In these polishing or grinding techniques, scratches having various shapes, which are polishing or grinding flaws, are generated on the surface of an object to be processed (for example, an insulating film on a semiconductor substrate (wafer)) after polishing or grinding. is there. As described above, in the manufacture of semiconductors and magnetic heads, if scratches having various shapes are generated on the surface of the workpiece, the wiring formed thereon is insufficiently etched, causing a defect such as a short circuit. Therefore, the occurrence of scratches with various shapes is monitored by observing the polished or ground wafer polished surface or ground surface, and if it occurs frequently, the polishing or grinding conditions corresponding to the scratch shape are reviewed. There must be. At the same time, when foreign matter is generated, it may cause a failure such as insulation failure or short circuit of a wiring formed thereon. When foreign objects frequently occur, measures different from scratches, such as cleaning the apparatus, are required. In other words, in a polishing or grinding process for an object to be processed (for example, an insulating film on a semiconductor substrate), foreign substances and scratches having various shapes can be separately monitored and appropriate measures can be taken for each. Necessary.

  However, in any of the above prior arts 1 to 4, scratches having various shapes generated on the surface of a workpiece (for example, an insulating film on a semiconductor substrate) that are polished or ground are given. Discriminating and inspecting the attached foreign matter is not considered.

  In addition, the scratches having various shapes have a width W of about 0.2 μm to 0.4 μm and a depth D of about several nanometers to a very deep one of about 100 nm. Then, an operator visually reviews using an electron microscope to discriminate scratches and foreign matters having various shapes, and a great amount of review time is required. For this reason, there is a delay in countermeasures against scratches or foreign matters, and a large number of wafers are continuously polished in a bad condition, resulting in a great loss in profit.

  An object of the present invention is to solve the above-mentioned problems when performing polishing or grinding processing such as CMP on an object to be processed (for example, an insulating film on a semiconductor substrate) in semiconductor manufacturing or magnetic head manufacturing. It is an object of the present invention to provide a surface inspection apparatus and method for discriminating and inspecting scratches having various shapes generated on the surface and adhering foreign substances. Another object of the present invention is to provide scratches having various shapes generated on the surface of an object to be processed (for example, an insulating film on a semiconductor substrate) when polishing or grinding such as CMP is performed. A semiconductor that can be manufactured in a highly reliable and efficient manner so that the above-mentioned defects do not occur by enabling 100% inspection or sufficient frequency sampling inspection to discriminate and inspect the adhered foreign matter. It is to provide a method for manufacturing a substrate. Still another object of the present invention is to provide a surface inspection apparatus and method for inspecting defects near the edge of a wafer in a workpiece.

In order to achieve the above object, the present invention provides a defect inspection apparatus for a first illumination system that illuminates an inspection region of the sample surface from a direction normal to or near the sample surface and the sample surface. And an illumination optical system including a second illumination system that illuminates the inspection region of the sample surface from an oblique direction, and the sample surface illuminated by the first illumination system of the illumination optical system or the second illumination system A detection optical system having detectors arranged so as to surround the illumination area, a magnitude of a detection signal of scattered light by the first illumination system, and a magnitude of a detection signal of scattered light by the second illumination system Based on the result classified by the first ratio, the second ratio of the magnitude of the detection signal of the scattered light respectively detected by the detector arranged at a position orthogonal to the inspection area of the sample surface as the center Use to classify defects And a signal processing system.

  As described above, according to the above configuration, it is possible to discriminate very shallow small scratches and foreign matters generated when the surface of the insulating film or the like is subjected to CMP processing, and further, linear large scratches and foreign matters can be distinguished. In addition, it is possible to discriminate small scratches from ties, marks, dimple marks, surface roughness, and the like, so that the cause of occurrence can be easily determined.

  According to the present invention, when a workpiece such as an insulating film is subjected to polishing or grinding such as CMP in semiconductor manufacturing or magnetic head manufacturing, it adheres to scratches having various shapes generated on the surface. There is an effect that a foreign substance can be discriminated and inspected. Further, according to the present invention, since the shape of the scratch can be classified in detail, there is an effect that the cause of the failure can be quickly identified. In addition, according to the present invention, all or a high frequency sampling inspection can be performed in the flattening polishing step, so that a defect of the polishing apparatus can be quickly found, and as a result, appropriate measures can be taken. Therefore, the yield in the polishing process can be drastically improved.

It is a schematic block diagram which shows the 1st Example of the surface inspection apparatus which concerns on this invention. It is a figure which shows the shape parameter of the scratch and foreign material which generate | occur | produce on an insulating film by CMP etc. which concern on this invention. It is a figure for demonstrating incident light projection length when the light beam d is irradiated to the scratch and foreign material which concern on this invention. It is a figure which shows the discrimination principle of the scratch and foreign material which concern on this invention. It is a figure which shows one Example of the discrimination | determination result of a scratch and a foreign material based on this invention. It is a figure which shows one Example of the discrimination processing flow of the scratch and foreign material which concern on this invention. It is a figure which shows the Example of the vertical irradiation which concerns on this invention, and pseudo | vertical vertical illumination. It is explanatory drawing of the conventional epi-illumination technique. It is a schematic block diagram which shows the modification of the two-way light reception in the 1st Example shown in FIG. It is a schematic block diagram which shows the 2nd Example of the multidirectional light reception of the surface inspection apparatus which concerns on this invention. It is the top view and front view which show arrangement | positioning of the multidirectional detection optical system shown in FIG. It is a figure for demonstrating a 2nd Example using the multi-directional detection optical system shown in FIG. It is a figure which shows distribution of the diffracted light when it illuminates with respect to the linear large scratch concerning this invention. It is a figure for demonstrating the discrimination principle by the diffracted light reception from the linear large scratch which concerns on this invention. It is a figure which shows one Example of the discrimination processing flow with a large scratch and a non-linear defect based on this invention. It is a figure which shows one Example of the whole flow of the discrimination algorithm which concerns on this invention. It is a figure which shows one Example of the discrimination processing flow of the large scratch and foreign material which concern on this invention. It is a figure for demonstrating the diffracted light distribution for every scratch shape which concerns on this invention. It is a figure which shows one Example of the scratch shape classification | category processing flow which concerns on this invention. It is a figure which shows one Example of the scratch shape classification | category result which concerns on this invention. It is a figure which shows one Example of the discrimination result layout displayed on the display apparatus which concerns on this invention. It is a figure which shows one Example of the tire mark luminance distribution histogram which concerns on this invention. It is a figure which shows one Example of the data item for spreadsheets which concerns on this invention. It is a schematic block diagram which shows the 3rd Example of the surface inspection apparatus which concerns on this invention. It is a figure which shows one Example of the diffracted light distribution in the Fourier-transform surface shown in FIG. It is a figure which shows one Example of the diffracted light distribution evaluation flow in the Fourier-transform surface shown in FIG. It is a perspective view which shows an Example at the time of applying the Example of the surface inspection apparatus which concerns on this invention to the wafer with a wiring pattern. It is a schematic block diagram which shows the 4th Example of the surface inspection apparatus which concerns on this invention. It is a figure which shows one Example of the phase delay filter shown in FIG. It is a figure which shows one Example of the phase advance filter shown in FIG. It is explanatory drawing of the phase difference which arises with the scratch and foreign material which concern on this invention. It is explanatory drawing of the phase difference and luminance production | generation principle which are produced by the scratch part which concerns on this invention. It is explanatory drawing of the phase difference and luminance production | generation principle which arise with the foreign material part which concerns on this invention. FIG. 29 is an explanatory diagram for discriminating between scratches and foreign matters by the phase difference method shown in FIG. 28. It is a perspective view which expands and shows a wafer edge part. It is a perspective view which shows the scattered light distribution from a wafer edge part at the time of epi-illumination. It is a top view which shows the scattered light distribution from a wafer edge part and a defect when epi-illumination is carried out. It is explanatory drawing of one Example for discriminating a wafer edge part and a defect. It is explanatory drawing of the other Example for discriminating a wafer edge part and a defect. It is a figure which shows one Example of the spatial filter used for FIG.

Embodiments of a surface inspection apparatus and method for the purpose of stable operation of a planarization process used in a semiconductor manufacturing process, a magnetic head manufacturing process, and the like according to the present invention will be described with reference to the drawings. First, a first embodiment of a surface inspection apparatus and method according to the present invention will be described. That is, in the first embodiment, as shown in FIG. 2, when a SiO ₂ film (object to be processed) 22 is formed on a Si wafer 21 and subjected to CMP (Chemical Mechanical Polishing), Is to discriminate the scratch 23a having a shallow depth from the foreign matter 24. By the way, the Si substrate 21 does not necessarily exist under the SiO ₂ film 22, and a wiring layer may exist. In the CMP process, polishing is performed to flatten the surface of the SiO ₂ film 22. Therefore, scratches 23a that are polishing scratches are generated on the surface of the SiO ₂ film 22 as shown in FIG. Here, the thickness of the SiO ₂ film 22 is t, the width of the scratch 23 is W, and the depth is D. As for a rough dimension, W is about 0.2 μm to 0.4 μm. The depth D is about several nanometers to about 100 nm even for very deep objects. As described above, the scratch 23a generated by CMP is characterized by a very shallow depth with respect to the width. FIG. 2B shows dimensional parameters of the foreign matter 24. Here, the foreign material 24 is modeled as a granular material having a diameter Φ. The actual foreign matter 24 is not such a beautiful spherical shape, but the scratch 23a has a depth D of about several nm to several + nm with respect to the width W (about 0.2 μm to 0.4 μm). 24 indicates that there is no extremely large difference in width and height as the scratch 23a. The present invention pays attention to the characteristic dimension ratio of the scratch 23a.

  By the way, the scratch 23 has various shapes. In particular, in CMP, polishing is performed by mixing a chemical mechanism and a mechanical mechanism. The scratch 23a generated due to a problem of mechanical polishing by a minute cutting mechanism has a minute linear shape like a so-called scratch. In addition, although rare, foreign matters other than abrasive grains are mixed during polishing, resulting in a linear large scratch 23b that is very deep with respect to the width. However, the scratch 23ab caused by a defect of a chemical factor that is an etching polishing mechanism has a shape like a dimple mortar. As described above, the shape of the scratch 23 varies depending on the cause of the failure in polishing. In other words, it becomes easier to narrow down the cause of the problem by classifying the shape of the scratch 23 in detail. In particular, when large scratches 23b due to foreign matter are frequently generated or when further huge scratches 23c are generated, it is necessary to immediately stop the polishing process and take countermeasures.

  Next, a first example of a surface inspection apparatus such as a scratch for realizing the first embodiment will be described with reference to FIGS. That is, the first embodiment of the surface inspection apparatus includes, as shown in FIG. 1, a wafer 10 that is an inspection object placed on a stage 15 whose position coordinates are measured and travel controlled in the XY directions, For example, a light source 2 comprising a light source (not limited to a laser light source) such as an Ar laser having a wavelength of 488 nm (blue wavelength), a nitrogen laser, a He—Cd laser, or an excimer laser, an optical path switching mechanism 3, and a reflection mirror An illumination optical system 1a composed of 4a, 4b, and 4c; a detection optical system 5 composed of a condenser lens 6 and a photoelectric converter 7 composed of a photomultiplier, a CCD camera, a CCD sensor, a TDI sensor, and the like; An A / D converter 16 that converts an analog luminance signal output from the photoelectric converter 7 into a digital luminance signal, and temporarily stores the digital luminance signal obtained from the A / D converter 16 An arithmetic processing unit 8 including a memory unit 17 and a comparison operation unit 18; a stage controller 14 for controlling the travel of the stage 15 based on position coordinates measured from the stage 15; and controlling the stage controller 14; The optical path switching mechanism 3 is controlled, the arithmetic processing unit 8 is further controlled, and an overall control unit 9 that receives an inspection result obtained from the arithmetic processing unit 8 is configured. The light source 2 preferably has a wavelength as short as possible as in the case of an excimer laser light source in order to discriminate and detect minute foreign matters 24 and scratches 23 generated on the CMP insulating film 22. The light emitted from the light source 2 does not directly irradiate the surface of the condenser lens 6, and passes through the wafer surface (the surface of the insulating film subjected to CMP) via the reflection mirror 4 a and the reflection mirror 4 c. Irradiate from the normal direction or from the vicinity. This is called epi-illumination 12. Alternatively, by retracting the reflection mirror 4a by the optical path switching mechanism 3, the wafer surface (the surface of the insulating film subjected to CMP) is irradiated from an oblique direction through the reflection mirror 4b. This is referred to as oblique illumination 11. In the first embodiment, epi-illumination and oblique illumination are realized by using one light source 2, a plurality of reflection mirrors 4a to 4c, and an optical path switching mechanism 3, but each has two independent independent illuminations. Individual light sources may be used. Further, the number of reflection mirrors and the presence or absence of an optical path switching mechanism are not limited. As described above, the illumination optical system 1a does not directly irradiate the surface of the condenser lens 6, and the normal direction with respect to the CMP surface on which the CMP is performed on the insulating film 22 on the wafer 10 or the It is only necessary to realize two systems of illuminations 11 and 12 from a near direction and an oblique direction (an angle of about 30 ° or less) near the wafer horizontal plane.

  Next, the detection procedure will be described. Detection is performed twice for one wafer by switching the illumination direction. Specifically, first, the incident illumination light 12 is irradiated to the CMP surface of the insulating film 22 on the wafer 10 without directly irradiating the surface of the condenser lens 6. Then, the specularly reflected light component generated from the insulating film 22 was removed without generating stray light reflected from the fine surface roughness of the surface of the condenser lens 6 or the extremely fine foreign matter adhering to the surface. In this state, only the shallow shallow scratches 23a generated by CMP on the insulating film 22 and the scattered light (low-order diffracted light component) emitted from the foreign matter 24 are constituted by the condenser lens 6 from a CCD, TDI sensor, or the like. The light is condensed on the light receiving surface of the photoelectric converter 7. Then, the output of the photoelectric converter 7 is A / D converted by the A / D conversion unit 16 to obtain the luminance value S (i) for each defect i, and then temporarily written in the storage unit 17. Next, the overall control unit 9 controls the stage 15 to irradiate the same coordinate position on the wafer surface with the oblique illumination 11 while switching the irradiation direction using the optical path switching mechanism 2. Then, in the state where the specularly reflected light component generated from the insulating film 22 is removed, scattered light (low-order diffracted light component) emitted from the extremely shallow fine scratch 23a and foreign matter 24 generated on the insulating film 22 by CMP. Only the light is condensed on the photoelectric converter 7 by the condenser lens 6. Then, the output of the photoelectric converter 7 is A / D converted by the A / D converter 16 to obtain the luminance value T (i) for each defect i, and then temporarily written in the storage unit 17. Next, the ratio R () of the detected luminance value S (i) for each defect i by the epi-illumination 12 already stored in the storage unit 17 and the detected luminance value T (i) for each defect i by the oblique illumination 11 i) is calculated by the comparison calculation unit 18. When the calculated luminance ratio R (i) is greater than a preset threshold value (judgment reference value: the discrimination line 20 shown in FIG. 5), the comparison calculation unit 18 generates a foreign object 24 and a shallow shallow scratch 23a. Discriminate and output to the overall control unit 9. As described above, the scratches 23a generated by CMP are extremely shallow and fine. Therefore, when the incident illumination light 12 is irradiated on the surface of the condenser lens 6, the weak stray light generated from the surface of the condenser lens 6 is also photoelectrically generated. When the light is received by the converter 7, it becomes difficult to distinguish it from the scattered light from the scratch 23a. Therefore, the surface of the condenser lens 6 is not irradiated with the epi-illumination light 12.

  In the first embodiment, detection by the epi-illumination 12 is performed first, and detection by the oblique illumination 11 is performed later. However, detection by the oblique illumination 11 is performed first and detection by the epi-illumination 12 is performed later. It doesn't matter. In the first embodiment, the detected luminance value T (i) by the oblique illumination 11 that is the second detection is once written in the storage unit 17 after A / D conversion. Even if the comparison operation unit 18 refers to the detected luminance value S (i) of the first incident illumination 12 already stored at the same time as the detection and stores the value T (i), the comparison operation unit 18 performs the luminance comparison calculation. It is possible to realize the invention.

  Next, the principle of discrimination for realizing the above-described embodiment according to the present invention will be described in detail with reference to FIGS. In the present invention, discrimination is performed by irradiating one defect with a light beam d from two different angles (for example, epi-illumination 12 and oblique illumination 11). First, the incident illumination light 12 is irradiated with the light flux d from the normal direction of the wafer surface or the vicinity thereof without directly irradiating the surface of the condenser lens 6. Next, the oblique illumination light 11 is irradiated with a light beam d from an angle close to the horizontal direction with respect to the wafer surface. It does not matter which one of the epi-illumination 12 and the oblique illumination 11 is performed first. Discrimination is performed by comparing the intensities of scattered light emitted from the defects 23a and 24 obtained in the two-way illumination of the light beam d. The intensity of scattered light from the defects 23a and 24 is emitted according to the amount of light source received by the defects 23a and 24. As shown in FIG. 3, it can be considered that the amount of light source received by the defects 23a and 24 is substantially proportional to the projected area of the defect dimension in the light source incident direction. In the case of the scratch 23a, this projected area is proportional to the width W during epi-illumination and to D 'during oblique illumination. Since the scratch depth D is very shallow compared to the width W, the oblique illumination projection length D ′ is much shorter than the epi-illumination projection length W ′. Therefore, the amount of light source received by the scratch 23a is weaker in the oblique illumination 11 than in the epi-illumination 12, and as a result, the amount of scattered light emitted from the scratch 23a is weaker in the oblique illumination 11. . On the other hand, in the case of the foreign matter 24, the projection lengths Φ of the oblique illumination 11 and the epi-illumination 12 are substantially equal, so the amount of scattered light emitted from the foreign matter 24 is compared with the epi-illumination and the oblique illumination. Will not change greatly. Therefore, as shown in FIG. 4, the detected brightness values of the scattered light by the epi-illumination 12 and the oblique illumination 11 are compared, and if the oblique illumination 11 is smaller than the epi-illumination 12, the scratch 23a is equivalent. Alternatively, it is possible to distinguish an object with larger oblique illumination as the foreign object 24.

By the way, since the insulating film (for example, SiO ₂ film) 22 in which the scratch 23a is generated by CMP is transparent to light, specular reflection light from the lower layer including light interference is generated. For this purpose, it is necessary to devise a method for preventing regular reflection light (including light interference light) from the surface of the insulating film 22 and its lower layer from going outside the field of view of the condensing lens 6. Of course, in the case of the oblique illumination 11, there is a contrivance to prevent regular reflection light (including light interference light) from the surface of the insulating film 22 and its lower layer from going outside the field of view of the condenser lens 6. Necessary. If a light source that emits broadband light or white light is used as the light source 2, the problem of optical interference between the regular reflection light from the surface of the insulating film 22 and the regular reflection light from the lower layer does not occur. However, in order to obtain intense scattered light from the fine scratches 23a on the insulating film 22 (particularly the depth D is shallow) and the foreign matter 24, it is preferable to use UV light or DUV light as illumination light.

  FIG. 5 is a graph showing an example of the discrimination result. This is a graph in which the horizontal axis represents the detected luminance value during epi-illumination and the vertical axis represents the detected luminance value during oblique illumination. In this case, the area below the discrimination line 20 in the figure is the area of the scratch 23a, and the area above is the area of the foreign matter 24.

  Next, an example of a flow for performing arithmetic processing using the above-described discrimination method will be described with reference to FIG. First, in step S <b> 61, the photoelectric converter 7 detects the luminance signal S (i) for each defect i caused by the epi-illumination 12 and stores it in the storage unit 17 after A / D conversion. Next, in step S <b> 62, the photoelectric converter 7 detects the luminance signal T (i) for each defect i caused by the oblique illumination 11 and stores it in the storage unit 17 after A / D conversion. In step S 63, the comparison calculation unit 18 determines the luminance signal S (i) for each defect i detected by epi-illumination stored in the storage unit 17 and the luminance signal T (i) for each defect i detected by oblique illumination. ) And the ratio R (i) with the following equation (1).

R (i) = T (i) / S (i) (Equation 1)
Here, i is an identification number assigned to each defect in order to evaluate a plurality of defects. In addition, since one defect may be detected as a plurality of defects depending on the size of the light beam d and the pixel size of the photoelectric converter 7, an expansion process (connection process) is performed on a signal indicating a defect detected in the vicinity. ) Needs to be converted into a signal indicating one defect. For this reason, the identification number i assigned to each defect is given to a signal indicating one defect subjected to the connection process.

  Furthermore, in step S64, the comparison calculation unit 18 determines that the foreign substance 24 is present if the calculated luminance ratio R (i) is greater than a preset threshold value (judgment reference value: the discrimination line 20 shown in FIG. 5). 23a and output to the overall control unit 9. In this embodiment, the detected luminance T (i) at oblique illumination is divided by the detected luminance value S (i) at epi-illumination, but conversely, the detected luminance value S (i at epi-illumination). ) May be divided by the detected luminance value T (i) during oblique illumination. In this case, if the ratio R (i) is larger than a preset threshold value (judgment reference value: the discrimination line 20 shown in FIG. 5), the scratch 23a is formed, and if the ratio R (i) is smaller, the foreign object 24 is formed.

  Next, an embodiment of a method for installing the reflection mirror 4c will be described with reference to FIG. This is a technique for detecting defects with high sensitivity by preventing stray light in the dark field detection system. The inspection of the scratch 23a requires illumination from a direction close to the normal to the surface of the wafer 10, as can be seen from the principle described above. However, in the epi-illumination technique as shown in FIG. 8 (the reflection mirror 4c ′ is placed on the lens 6), the incident light passes through the condenser lens 6 and illuminates the wafer 10, so that so-called stray light is generated. As a result, noise is generated in the detected image. Specifically, the scattered light generated from fine polishing marks on the surface of the condensing lens 6 and dust adhering to the condensing lens 6 becomes stray light. For this reason, when the minute scattered light from the defects 23a and 24 is received by the photoelectric converter 7 and observed, this stray light becomes fatal. That is, the scattered light from the extremely small scratch 23a is buried in noise caused by stray light and cannot be detected.

  Therefore, in the present invention, as shown in FIG. 7, the incident light having a strong intensity is not irradiated on the surface of the condenser lens 6, and the wafer 10 (the surface of the interlayer insulating film 22 (CMP surface) and its lower layer) The reflection mirror 4c is provided so that zero-order diffracted light, which is a specularly reflected light component from the surface of the wiring layer, the surface of the scratch 23a, and the surface of the foreign substance 24, does not enter the pupil of the condenser lens 6, that is, the NA. Devised. In FIG. 7A, the reflecting mirror 4c1 is arranged between the wafer 10 and the lens 6 substantially on the normal line of the wafer 10, and the incident illumination light 12a is viewed from the lateral direction so that the surface of the condenser lens 6 is not irradiated. The light is incident on the reflection mirror 4c1 to be reflected, and the specularly reflected light component from the wafer 10 is reflected by the reflection mirror 4c1 and is not incident on the pupil of the lens 6, but scattered light (1 A method in which scattered light (low-order diffracted light component) in a region indicated by oblique lines (in a planar shape in a plane) among lower-order diffracted light components) enters the pupil of the lens 6 will be described. The reflection mirror 4c1 has an almost elliptical outer shape. This is called scattered light detection by vertical illumination.

  7B, the reflecting mirror 4c2 is arranged between the wafer 10 and the condenser lens 6 and outside the condenser lens 6 from the NA, and the epi-illumination light 12b is provided on the surface of the condenser lens 6. So that the light is incident on the reflecting mirror 4c1 from the lateral direction so as not to be irradiated, and the specularly reflected light component from the wafer 10 is out of the pupil of the condenser lens 6, and the scattered light from the scratch 23a and the foreign matter 24 is included. The method of making the scattered light of the area | region shown with a diagonal line enter into the pupil of the lens 6 is shown. When the reflection mirror 4c2 is expanded in the circumferential direction, the illumination light illuminated by the reflection mirror 4c2 becomes annular illumination. However, as shown in FIG. 7B, when the reflecting mirror 4c2 is made a part, it becomes a part of the illumination in the annular illumination. This is called scattered light detection by pseudo vertical illumination. In FIG. 7C, a reflecting mirror or half mirror 4c3 is disposed on the condensing lens 6, a condensing lens 6 having an opening 50 formed in the center is disposed, and the vertical mirror reflected by the half mirror 4c3 is disposed. The illumination light 12a is irradiated on the insulating film CMP surface on the wafer 10 through the opening 50 without irradiating the surface of the condenser lens 6, and the specularly reflected light component from the wafer 10 is provided on the Fourier transform surface. A method is shown in which the photoelectric converter 7 receives the scattered light obtained through the condensing lens 6 among the scattered light from the scratch 23 a and the foreign matter 24 while being shielded by the spatial filter 51.

  Further, in FIG. 7D, as in FIG. 7C, the incident illumination light 12a is transmitted through the half mirror 52, passed through the opening 50 of the condenser lens 6, and vertically illuminated on the CMP surface of the wafer 10. The regular reflection light from the wafer 10 is shielded by the spatial filter 53 provided on the Fourier transform surface, and the scattered light obtained through the condenser lens 6 among the scattered light from the scratch 23a and the foreign matter 24 is reflected by the half mirror 52. A method of receiving light by the photoelectric converter 7 will be described. As described above, in FIGS. 7C and 7D, stray light is generated from the surface of the condenser lens 6 by forming the opening 50 in the center of the condenser lens 6 as in FIG. 7A. Without this, it becomes possible to detect vertical illumination and scattered light from the vertical direction. Therefore, no matter how the direction of the scratch 23a is formed in the horizontal plane, scattered light generated from the edge of the very shallow scratch 23a can be received relatively uniformly by the photoelectric converter 7, and the uniform A detected luminance value can be obtained. Furthermore, as shown in FIG. 13B, vertical illumination is preferable to pseudo-vertical illumination in order to obtain diffracted light having a strong directivity in a direction perpendicular to the large scratch 23b, which is a linear pattern.

  By the way, in the case of the scattered light detection by the vertical illumination shown in FIG. 7A, the incident light passes under the lens 6 and is clearly not irradiated on the surface of the condenser lens 6, and stray light does not occur. Further, the specularly reflected light from the wafer 10 is reflected by the reflecting mirror 4c1, and therefore does not enter the pupil of the condenser lens 6. Furthermore, the same applies to the vertical illumination shown in FIGS. 7C and 7D. Also, in the case of scattered light detection by pseudo vertical illumination in FIG. 7B, obviously incident light does not pass through the condenser lens 6. Further, since the reflection mirror 4 c 2 is disposed outside the NA of the condenser lens 6, the regular reflection light component from the wafer 10 does not enter the pupil of the condenser lens 6. That is, in any method, incident light that has high light intensity and is likely to generate stray light is not irradiated on the surface of the condenser lens 6, and epi-illumination is realized so that specularly reflected light from the wafer does not enter the condenser lens 6. doing. For this reason, stray light hardly occurs, and a detection image with a high S / N ratio can be obtained from the scratch 23 a and the foreign matter 24 generated on the CMP surface where the interlayer insulating film 22 is subjected to CMP. In addition, since the interlayer insulating film 22 is transparent to light, the light that is specularly reflected from the lower layer returns when incident on the epi-illumination, but is not incident on the NA of the lens 6 as described below. Therefore, the scratch 23a and the foreign matter 24 can be detected by a signal obtained from the photoelectric converter 7 without affecting the detection of scattered light from the scratch 23a and the foreign matter 24.

  Furthermore, the epi-illuminations 12a and 12b shown in FIG. 7 are not only for the reason for solving the stray light, but also particularly easy to receive a component having a strong scattered light intensity distribution from the scratch 23a. High detection sensitivity can be obtained. This is because the low-order diffracted light component is relatively strong in the scattered light intensity from the scratch 23a. That is, if illumination is performed from the vicinity of the normal line of the wafer surface, low-order diffracted light components are reflected from the wafer 10 and are easily collected by the condenser lens 6. As a result, it is possible to detect the scratch 23a with higher sensitivity than when using only the oblique illumination 11. Thus, by using only the vertical illumination 12a or the pseudo vertical illumination 12b, it is possible to realize a highly sensitive inspection of the scratch 23a.

  By the way, even if the reflecting mirror 4c1 is arranged in the NA of the condenser lens 6, if the reflecting mirror 4c1 is formed in an almost elliptical shape so as not to affect the imaging characteristics of the lens 6 or the like, FIG. The scattered light in the region (a zone region in plan view) indicated by the oblique lines in (a) can be condensed by the condenser lens 6 to form an image. However, if the presence of the reflection mirror 4c1 within the NA of the condenser lens 6 adversely affects the imaging characteristics, a mechanism for retracting the reflection mirror 4c1 out of the NA during vertical illumination is required. In the case of semiconductor inspection, it is necessary to eliminate as much as possible dust generated from the defect inspection apparatus. From this point of view, it is not preferable to provide the movable mechanism above the wafer. However, even in such a case, the pseudo vertical illumination 12b may be used. In the case of the quasi-vertical illumination 12b, the reflecting mirror 4c2 is outside the NA, so the imaging characteristics are never adversely affected, and there is no need to provide a separate retraction mechanism.

  Further, when the surface inspection device such as a scratch according to the present invention is used as a foreign matter inspection device using only oblique illumination, vertical illumination is unnecessary, and therefore the reflecting mirror 4c1 shown in FIG. It is also possible to effectively collect the scattered light generated from the foreign matter using all the NA of the condenser lens 6 and cause the photoelectric converter 7 to receive the scattered light. However, in order not to retract the reflecting mirror 4c1 and eliminate the generation of dust, the pseudo vertical illumination 12b whose scratch detection accuracy is somewhat lowered may be used as the vertical illumination of the surface inspection apparatus. Further, when the method shown in FIGS. 7C and 7D is used as the vertical illumination, it can be applied by stopping the vertical illumination even when used as a foreign matter inspection apparatus using only oblique illumination. It becomes. Furthermore, when used as a foreign matter inspection apparatus using only oblique illumination, if a foreign matter on a memory cell on which a periodic wiring pattern is formed is detected, the diffraction pattern based on the diffracted light from the periodic wiring pattern is shielded. Therefore, the spatial filters 51 and 53 may be replaced with linear spatial filters.

  Further, in the first embodiment, one detection optical system 5 is used, but a plurality of detection optical systems 5a and 5b may be used as shown in FIG. In particular, if the detection optical systems 5a and 5b are arranged for each irradiation direction in the direction in which the intensity of scattered light generated from the defects 23a and 24 can be detected most strongly, detection with high sensitivity becomes possible. For example, as shown in FIG. 9, as the detection optical system 5a during the epi-illumination, a lens 6a and a photoelectric converter 7a are provided in the normal direction of the wafer where the scattered light intensity is strong. The arrangement of the lens 6a can be applied to either the vertical illumination 12a described above or the pseudo epi-illumination 12b. Further, as the detection optical system 5b during oblique illumination, a condenser lens 6b and a photoelectric converter 7b are provided in the regular reflection direction of obliquely incident light having strong scattered light intensity. However, with respect to the detection optical system 5b including the lens 6b and the photoelectric converter 7b, it is necessary to prevent the regular reflection light component from the wafer 10 from being condensed. For this reason, it is preferable to install the detection optical system 5b in the specular reflection light emission direction 55, and to install the detection optical system 5b at a location where the specular reflection light component deviates from the NA of the condenser lens 6b. In the case of the present embodiment, since two detection optical systems 5a and 5b are provided, two A / D conversion units 16a and 16b and luminance storage units 17a and 17b are provided. However, of course, the same processing is possible even if the addresses are changed and stored separately in one storage unit 17, for example, one RAM.

  Next, a second embodiment of the surface inspection apparatus such as a scratch according to the present invention will be described with reference to FIGS. In the second embodiment, the configuration different from the first embodiment is in the detection optical system 5. That is, the detection optical system 5 in the second embodiment is provided with a high angle detection optical system 5a, a medium angle detection optical system 5c, and a low angle detection optical system 5b. In the first embodiment, a configuration in which the substrate mounting base 51 on which the wafer 10 as the inspection object is fixed and installed by, for example, vacuum suction is provided on the stage 15 is omitted. In the second embodiment, the stage 15 is composed of a straight stage 15a and a rotary stage 15b. That is, the stage 15 only needs to be able to transport the wafer 10 so that it can be illuminated at an arbitrary coordinate on the wafer 10. Further, in the arithmetic processing unit 8, the A / D conversion unit 16 and the storage unit 17 are also configured corresponding to the detection optical systems 5a to 5c. The illumination optical system 1a is configured in substantially the same manner as in the first embodiment. That is, for the vertical illumination 12, the illumination light reflected by the reflection mirror 4a is reflected by the reflection mirror 4d and transmitted through the half mirror 52. As shown in FIG. 7D, the aperture formed in the condenser lens 6a. The CMP surface on the wafer 10 is illuminated with the light beam d after passing through 50. The specularly reflected light generated from the wafer 10 is shielded by the spatial filter 53, and the low-order diffracted light generated from the edge of the scratch 23a and the foreign matter 24 is condensed by the condenser lens 6a of the high-angle detection optical system 5a. Light is received by the photoelectric converter 7a, and higher-order diffracted light is detected by the medium angle detection optical system 5c. Of course, the vertical illumination 12 may be configured as shown in FIGS. 7 (a), (b), and (c). The oblique illumination 11 is reflected by the reflection mirror 4b and illuminated on the CMP surface on the wafer 10 by the light beam d. Then, the diffracted light generated from the foreign matter 24 is detected by the detection optical systems 5a to 5c without detecting the specularly reflected light generated from the wafer 10.

  Next, the detection optical system of the second embodiment will be specifically described with reference to FIG. That is, one high-angle detection optical system 5a, four medium-angle detection optical systems 5c, and four low-angle detection optical systems 5b, each composed of condenser lenses 6a-6i and photoelectric converters 7a-7i. Arrange and configure. In this embodiment, photomultipliers are used as the photoelectric converters 7a to 7i. As shown in FIG. 11, nine photomals are used and arranged in a dome shape. These photomals 7a to 7i are provided with condensing lenses 6a to 6i. In this embodiment, the condensing lenses 6a to 6i and the photomultipliers 7a to 7i are used for the detection optical system. However, for example, they may be imaged on a CCD camera or a TDI sensor. Further, the number of photoelectric converters 7a to 7i is not limited to nine. Outputs of the photoelectric converters 7a to 7i are written in the storage units 17a to 17i via the A / D conversion units 16a to 16i. At the same time, the coordinate data of the wafer 10 obtained from the stage controller 14 is also written in the storage units 17a to 17i. Then, the coordinate data and the luminance data are transmitted to the comparison calculation unit 18. When inspecting the entire surface of the wafer, it is necessary to write the coordinate data and the luminance data in pairs in the storage units 17a to 17i as described above. However, coordinate data is not always necessary when inspecting with fixed coordinates. Further, the coordinate data and the luminance data need not be stored in pairs in the same storage units 17a to 17i, and may be separate storage units. Further, the coordinates need not be the same. For example, a detected number may be assigned to a detected defect and stored. It is only necessary that correspondence between the detected luminance data of oblique illumination and the detected luminance data of epi-illumination for the same defect can be obtained. In the present embodiment, the coordinate data is used to recognize the data existing at the same coordinate or the nearby coordinates among the two types of data of the epi-illumination 12 and the oblique illumination 11 as the luminance data of the same defect. Thus, the brightness values S (i) and T (i) of the epi-illumination 12 and the oblique illumination 11 are compared.

  As shown in FIGS. 11A and 11B, one detection is made in the incident direction of the epi-illumination 12 (having an angle Av from the normal direction to the wafer surface. The angle is preferably 0). An optical system 5a was provided. This is referred to as a high angle detection optical system 5a. Four high-angle detection optical systems 5a and four low-angle detection optical systems 5b are provided sequentially from the high-angle detection optical system 5a toward the wafer surface. In the present embodiment, a total of nine detection optical systems are used, but the number is not limited as means for realizing the present invention.

  In this embodiment, as shown in FIGS. 12A and 12B, all nine detection optical systems 5a to 5c are used as light receiving means when the epi-illumination 12 is used. The sum of the light reception luminances of the nine photoelectric converters 7a to 7i was used as the light reception luminance when the epi-illumination 12 was used. However, it is not necessary to use all nine as the received light brightness when reflected by the epi-illumination 12, and only the high angle detection optical system 5a, only the medium angle detection optical system 5c, or the medium angle detection with the high angle detection optical system 5a. You may use the sum of the received light quantity of the optical system 5c. In particular, the detection of the low-order diffracted light generated from the scratch 23a and the foreign matter 24 by the epi-illumination 12 may be performed only by the medium angle detection optical system 5c instead of the high angle detection optical system 5a. At this time, since regular reflection light (0th order diffracted light) is not incident on the NA of the medium angle detection optical system 5c, the sum of the light amounts of the medium angle detection optical system 5c may be simply taken. In this way, various combinations are conceivable. In this embodiment, in order to achieve the highest NA (Numerical Aperture) so that the defect detection sensitivity is highest, the received light amounts of all the detection optical systems 5a to 5c are increased. Sum was used. However, since most of the diffracted light from the scratch 23a and the foreign matter 24 is detected by the high angle detection optical system 5a and the medium angle detection optical system 5c, it is preferable to use the sum of the received light amounts of both.

  The light receiving means when the oblique illumination 11 forms an angle Vo from the wafer surface, as shown in FIGS. 12C and 12D, on the side close to the regular reflection direction from the wafer when the oblique illumination is incident. Two low-angle light receivers 6b and 7b; 6c and 7c were used. The sum of the two light reception luminances was used as the light reception luminance when the oblique illumination 11 was used. This is not limited to two low angles. In the present embodiment, in order to achieve high sensitivity, the detection optical system (the condensing lens 6b and the photoelectric converter) in the direction in which the scattered light distribution intensity is strong without receiving regular reflection light as described in FIG. 9 is used. 7b) Only 5b was selected. From the viewpoint of discrimination, it is important to detect the intensity of each scattered light using two-way illumination, and the direction of the light receiving means 5b is not so much a problem. The direction of the detection optical system (the condensing lens 6 and the photoelectric converter 7) 5 may be determined according to how far the fine foreign matter 24 must be detected from the scratch 23a. Further, when the high-angle detection optical system 5a is used as the detection optical system 5, it is necessary to prevent stray light from being generated from the surface by the incident illumination light.

  The embodiment of the discrimination method between the foreign matter 24 and the scratch 23a by the two-way illumination described above pays attention to the feature of the shape that the depth of the scratch 23a is shallow.

  However, as described above, in the case of CMP, large foreign matter may enter the CMP apparatus from the outside, which may rarely cause very deep scratches. As described above, when a large foreign matter is mixed during polishing and a scratch is generated by a large foreign matter instead of a fine abrasive grain, a scratch 23b having a depth D deep with respect to the width W is generated. In the case of the scratch 23b having such a deep depth as described above, an object that should originally be the scratch 23 is erroneously recognized as the foreign object 24 only by the discrimination method. As a result, a very large scratch that should not be overlooked cannot be recognized.

  Accordingly, a second embodiment in which the deep scratch 23b is also recognized by being distinguished from the foreign matter 24 will be described next. That is, in the first embodiment, since the ratio R (i) = T (i) / S (i) is large for both the foreign matter 24 and the deep scratch 23b, the discrimination processing flow shown in FIG. In step S64, the deep large scratch 23b is also detected in the same manner as the foreign object 24. Therefore, in the second embodiment, the large scratch 23b and the foreign matter 24 are further accurately discriminated. S (i) represents luminance data in the data for each defect i after the first connection process of inspection by the epi-illumination 12. T (i) indicates luminance data indicating the same coordinate value i among the data for each defect after the second connecting process with the oblique illumination 11.

  That is, it is noted that the scratch 23b having a depth D with respect to the width W is also long at the same time. This is because polishing in CMP is performed while rotating, and it is impossible to dig deep locally like a dent. In the second embodiment according to the present invention, attention is paid to this necessity, and among those recognized as the foreign matter 24 in step S64 in FIG. 6, those that are linearly long are further classified as large scratches 23b. This is based on the principle shown in FIG. When the large scratch 23b, which is a linear pattern, is irradiated with the light beam 12 from the normal direction as shown in FIG. 13B, for example, the diffracted light has a very strong directivity in the direction perpendicular to the linear pattern 23b. Show the distribution. FIG. 13A shows a case where the large scratch 23b, which is a linear pattern, is irradiated with the light beam 11 with an inclination from the normal direction. This is a diagonal pattern used in a foreign substance inspection apparatus. The principle of a spatial filter technique for removing a linear diffracted light pattern generated from a wiring pattern arranged regularly and repeatedly by side illumination will be described. In the present invention, as shown in FIG. 13 (b), by recognizing the strong directivity of the diffracted light from the large scratch 23b by the vertical illumination 12, the large scratch in which the defect is not a foreign matter 24 but a linear pattern. 23b is identified.

  First, the principle of discrimination between the large scratch 23b and the foreign matter 24 will be described with reference to FIGS. First, for example, the epi-illumination 12 is applied to the wafer 10, and the arithmetic processing unit 8, in step S <b> 65, includes a plurality (eight) of low angle and medium angle detection optical systems 5 b and 5 c (photoelectric converters 7 b to 7 i). ) Is selected from among the detection optical systems A having the highest luminance. Next, the arithmetic processing unit 8 refers to the luminance value Sb (i) of the detection optical system B orthogonal to the A detection optical system in step S66, and in step S67, the luminance value Sa ( i) and the luminance value Sb (i) of the detection optical system B are compared to calculate the luminance ratio (Sa (i) / Sb (i)). Next, in step S68, the arithmetic processing unit 8 compares the calculated luminance ratio (Sa (i) / Sb (i)) with a preset set value (threshold value) as a linear defect. Large scratches 23b and small ones are classified as foreign matter 24 or small scratches 23a which are non-linear defects.

  As described above, according to the second embodiment, since the large scratch 23b can be distinguished from the foreign matter 24 or the small scratch 23a in the arithmetic processing unit 8, it is combined with the first embodiment. Thus, the small scratch 23a, the foreign matter 24, and the large scratch 23b can be distinguished. Next, a description will be given of a third embodiment in which the deep scratch 23b is recognized as a foreign object 24. That is, in the third embodiment, as shown in FIG. 16, a method of extracting a long and large scratch 23c that crosses the wafer, for example, is also combined. The long scratch 23c that can be understood by visual observation can be extracted by simply evaluating the length after the connecting process in the arithmetic processing unit 8. However, the function of the present invention is not impaired even if there is no processing for extracting the giant scratch 23c.

  First, in this algorithm in the arithmetic processing unit 8, first, each of the inspection data of two times (first time: epi-illumination 12, second time: oblique illumination 11) in which the incident direction is changed is connected in each of steps S70a and S70b. Processing is performed, and giant scratch coordinate data 1 and 2 are acquired in steps S71a and S71b. This connection process is an expansion process that recognizes data having close coordinates as one defect. For example, in a 3 × 3 pixel, if a signal indicating a defect is detected in the periphery, defects that are slightly separated from each other are connected by repeating a dilation process that gives a signal indicating a defect to the center pixel a plurality of times. become. This is because one defect may be extracted as a plurality of defects due to factors such as spot size and pixel size. Next, in step S72, the arithmetic processing unit 8 refers to the coordinate data 1 and 2 indicating the defects after the connection process (for example, taking a logical sum of signals indicating the defects) to thereby convert the long defect to the giant scratch 23c. Extract as (create huge scratch data). Further, in step S72, the arithmetic processing unit 8 refers to the number of coordinate data 1 and 2 (for example, the number of pixels) indicating the defects connected by the connection process, and extracts a defect having a large area as the giant scratch 23c. Yes (create huge scratch data).

  Next, in each of steps S73a and S73b, the arithmetic processing unit 8 removes data recognized as a giant scratch from each of the first test data and the second test data, and in each of steps S61 and S62. Data after the first and second connection processing is acquired and stored in the storage unit 17. Next, the comparison operation unit 18 uses the data after the first and second connection processes obtained and stored in the storage unit 17, and in step S64, the scratch 23a by the two-way illumination shown in FIG. A discrimination process from the foreign matter 24 or the large scratch 23b is performed.

  Next, in steps S65 to S68, the arithmetic processing unit 8 performs discrimination processing for the large scratch 23b from the data discriminated to the foreign matter 24 or the large scratch 23b as shown in FIG. This discrimination processing will be described in detail with reference to FIG. That is, this discrimination processing uses a total of eight photomultipliers 7b to 7i of the low angle detection optical system 5b and the medium angle detection optical system 5c. In this case, since the solid angle with respect to the wafer 10 is different between the low angle and the medium angle, a sensitivity difference between the low angle and the medium angle is generated. In addition, the photomultipliers 7b to 7i are likely to make a difference in sensitivity characteristics of individual products. For this reason, the sensitivity balance between the eight photomals must be adjusted. Therefore, in step S74, sensitivity adjustment is performed in advance by changing the voltage applied to each of the photomultipliers 7b to 7i. In addition, when performing very delicate sensitivity balance adjustment, it is also effective to provide a gain for each of the photomultipliers 7b to 7i and to correct the detected luminance in terms of software or hardware. . However, intensity correction by this gain is not absolutely necessary. Next, using the data after correcting the sensitivity balance between the photomals, the arithmetic processing unit 8 calculates the sum of the luminances obtained between the photomals facing each other in step S75. As a result, the data of the eight photomals 7b to 7i is four. Then, in step S65, the arithmetic processing unit 8 selects the largest data (ΣSa (i)) from among the four luminance sum data, as in step S65 of FIG. Then, in steps S66 to S67, the arithmetic processing unit 8 refers to the luminance sum data Sb (i) at a position orthogonal to the orthogonal luminance ratio (ΣSa (i)) in the same manner as in steps S66 and S67 of FIG. / ΣSb (i)) is calculated. Further, in step S68, the arithmetic processing unit 8 classifies the orthogonal luminance ratio (ΣSa (i) / ΣSb (i)) as the large scratch 23b if it is larger than the set threshold value 50, and classifies it as the foreign object 24 if it is smaller. With the method described above, it is possible to classify the large scratches 23b that may be erroneously recognized as foreign matter separately from the foreign matter 24.

  Next, a fourth embodiment for classifying the defect 23a discriminated from scratch by two-way illumination in more detail will be described in detail with reference to FIGS. CMP is not only polishing by a mechanical action, but also polishing by an etching chemical action at the same time, as called chemical mechanical polishing. Usually, in the oxide film polishing process, the mechanical action is the dominant polishing condition, and the main scratch 23a is a fine crescent-like linear scratch as described as a tire mark in FIG. There are many cases where this is a continuous scratch 23aa. Conversely, when this chemical polishing action is significant, scratches 23ab on the mortar having a circular planar cross-sectional shape as described as dimple marks in FIG. 18 may occur. Further, when the polishing pad is worn out and hardened, the fine scratch group 23ac generated in a random direction as described in FIG. 18 may be generated. In this way, the scratch shape generated according to the failure factor of the polishing conditions is different. In other words, by classifying and recognizing the shape of the scratch in detail, it becomes easy to narrow down the process conditions to be improved, and the trouble countermeasure period can be greatly shortened. Therefore, in the above-described foreign object / scratch discrimination method using two-way illumination, the shape classification is performed in more detail with respect to the data discriminated from the scratch 23a. As shown in FIG. 18, this is a method of classifying scratch shapes by examining the scattered light intensity distribution in the horizontal and vertical directions in detail. As shown in FIG. 18, since the tire mark 23aa is a linear mark as described in the large scratch discrimination, diffracted light having strong directivity in the horizontal direction is generated. However, with respect to the dimple mark 23ab and the surface roughness 23ac, the directivity is hardly recognized in the horizontal scattered light intensity distribution. Therefore, the dimple marks 23ab and the surface roughness 23ac are classified using the scattered light intensity distribution in the vertical direction. This is because the dimple mark 23ab has directivity in the vertical direction but is not recognized by surface roughness. Therefore, as shown in FIG. 19, the arithmetic processing unit 8 sequentially classifies detailed shapes by evaluating the distribution of diffracted light in the horizontal direction or the vertical direction. In this embodiment, both horizontal and vertical scattered light intensity distributions are used, but either one may be used depending on the shape to be classified. That is, in steps S61 and S62, the first inspection is performed by upward detection (high angle detection optical system 5a, medium angle detection optical system 5c, and low angle detection optical system 5b) by epi-illumination 12, and the second time. This inspection is performed by the front low angle detection optical system 5b (6b, 7b; 6c, 7c) or the high angle detection optical system 5a and / or the medium angle detection optical system 5c by the oblique illumination 11.

  Next, in this embodiment, from the data classified as the scratch 23a in steps S63 and S64, the arithmetic processing unit 8 first uses the same technique shown in FIG. The scattered light intensity distribution (Sha (i) / Shb (i)) in the direction is evaluated, and those having strong directivity are classified as tire marks 23aa. In step S77, the arithmetic processing unit 8 evaluates the directivity in the vertical direction and classifies the remaining data into dimple marks 23ab and surface roughness 23ac. In this example, the directivity in the horizontal direction was performed in the same manner as the flow shown in FIG. There are various methods for evaluating the directivity in the vertical direction. For example, the detection luminance Sh (i) detected by the photoelectric converter 7a of the high angle detection optical system 5a, the low angle detection optical system 5b, and the medium angle detection optics. The ratio (Sh (i) / ΣSl (i)) with the sum ΣSl (i) of the detected luminance detected by the photoelectric converters 7a to 7i of the system 5c may be taken. FIG. 20 shows an example of the discrimination result. FIG. 20A shows the result of evaluating the horizontal luminance ratio (Sha (i) / Shb (i)) by the epi-illumination 12 and classifying it into a tire mark 23aa and a dimple mark 23ab or a rough surface 23ac. FIG. 20B shows the result of evaluating the vertical luminance ratio (Sh (i) / ΣSl (i)) by the epi-illumination 12 and classifying the surface roughness 23ac and the dimple mark 23ab. In FIG. 20B, the directivity of the diffracted light in the vertical direction increases as the diameter of the dimple mark 23ab increases. This is a finding that is consistent with a generally known Airy disc principle. It is also possible to estimate the diameter of the dimple mark from this luminance ratio. According to the fourth embodiment described above, it becomes possible to discriminate and inspect the foreign matter 24 existing on the insulating film processed flat by CMP and the scratch defect 23 having various shapes, and the result is calculated. The data is output from the unit 8 and stored in the storage device 31 connected to the overall control unit 9.

  Next, FIG. 1 and FIGS. 21 to 23 show a fifth embodiment in which it is evaluated in advance whether or not the detection defect is correctly classified in the surface inspection such as the scratch according to the present invention. Will be described. By the way, as shown in FIG. 1, a surface inspection apparatus such as a scratch includes a storage device 31, an input means 32 composed of a keyboard, a mouse, a storage medium, etc., a display device 33 composed of a display, and an SEM device, for example. A general control unit 9 for connecting a network 34 connected to the network is provided. Naturally, the storage device 31 stores the inspection result subjected to the discrimination processing in the arithmetic processing unit 8.

  Furthermore, in surface inspections such as scratches according to the present invention, it is necessary to confirm sensitivity as well, but it is necessary to evaluate in advance whether or not the classification of the detected defect as the inspection result is correctly performed by review. Don't be. For this reason, the data must be sampled based not only on the detected luminance information but also on the discrimination processing result. Specifically, among the many detected defects, the review target said to review only the defects whose classification results near the discrimination line positions (threshold values) 20 and 50 are likely to be suspicious by the SEM device (not shown). It is important to select and evaluate efficiently. That is, as will be described below, the defect coordinates (threshold values) 20 and 50 displayed on the screen 40 of the display device 33 are identified as suspicious defects, and the position coordinates of the defects are acquired. Can do. Then, by mounting this wafer 10 on an SEM apparatus and observing the defect with the SEM based on the acquired position coordinates, it is possible to review and evaluate whether it is a foreign matter 24 or various scratches 23. And if the review evaluation result by this SEM apparatus is input into the whole control part 9 via the network 34, for example, and it stores in the memory | storage device 31, the validity of the said discrimination line (threshold value) 20 and 50 can be reviewed. It becomes possible.

  Accordingly, as shown in FIG. 21, the screen 40 displayed on the display device 33 is displayed as a scratch discrimination luminance distribution graph 41, a large scratch discrimination luminance distribution graph 42, a corresponding coordinate search 43 on the defect map, and a discrimination result display. The window 44 is configured. The scratch discrimination luminance distribution graph 41 shows the relationship between the received light intensity S (i) by the epi-illumination and the received light intensity T (i) by the oblique illumination, and the fine scratch 23a and the foreign object 24 / large scratches by the threshold (discrimination line) 20. 23b is shown in a discriminated state. The large scratch discrimination luminance distribution graph 42 shows the relationship between the maximum luminance (ΣSa (i)) and the orthogonal luminance ratio (ΣSa (i) / ΣSb (i)). 23b is shown in a discriminated state. The coordinate search 43 on the defect map indicates the generation status (defect map) of the scratch 23 and the foreign matter 24 generated on the wafer 10. The discrimination result display window 44 includes the number of foreign objects 24 corresponding to the size (small, medium, large) and the number of scratches (tire marks 23aa, dimple marks 23ab, surface roughness 23ac, large scratches 23b, huge scratches 23c) 23. Is shown. The discrimination result display window 44 can also be displayed as a histogram.

  Although four types of display contents are displayed in separate windows, a plurality of graphs may be displayed in one window. Further, all of these four types may not be displayed at the same time. Although not shown, a luminance ratio screen that is a result of analyzing the contents of the scratch using the luminance ratio in the horizontal direction and the vertical direction can also be selected and displayed from the pull-down menu. Further, in the defect map corresponding coordinate search 43, when a defect on the defect map is designated by the cursor 32 or the like, the defect coordinates, the discrimination result, and the received light luminance of each of the photoelectric converters 7a to 7i are expressed by one or a plurality of items. It is possible to display. Further, on one or both of the scratch discrimination luminance distribution graph 41 and the large scratch discrimination luminance distribution graph 42, data corresponding to the defect designated by the cursor 32 or the like is blinked, or the color is changed, or By changing the size of the display mark or the like, the corresponding data can be displayed so that the operator can easily discriminate it. Similarly, when a data point is selected by the input means 32 such as a cursor on the scratch discrimination luminance distribution graph 41 or the large scratch discrimination luminance distribution graph 42, the corresponding data is displayed on the defect map. It can be distinguished on the monitor 33 by distinguishing it from other data by blinking, changing the color, or changing the size of the display mark. In addition, the defect coordinates of the data designated on the graphs 41 and 42, the discrimination results, and the information on the received light intensity of each of the photoelectric converters 7a to 7i are displayed for one or a plurality of items. In this way, by selecting arbitrary defect data on the graphs 41 and 42 or the defect map 43 and displaying the inspection information on the monitor 33, the examination of the validity of the discrimination processing is completed in a short time. It becomes possible.

  In this way, the overall control unit 9 uses the inspection results stored in the storage device 31 (coordinate data of the foreign matter 24 and the scratch 23, the discrimination results of the foreign matter 24 and the various scratches 23, and the incident light and oblique illumination to each photoelectric converter. 7a to 7i) can be displayed on the screen of the display device 33, and it is possible to review in advance whether or not the detection defects are correctly classified. . In particular, in the graph displays 41 and 42, it is possible to review the validity of the discrimination lines (threshold values) 20 and 50 for discrimination.

  Further, as shown by the discrimination result 44 in FIG. 21, the size of the defect in each classification of the foreign matter 24 and the scratch 23 is estimated according to the magnitude of the received light luminance by the photoelectric converters 7a to 7i, and classified into several categories. It is also effective to use the test result effectively. In both the foreign matter 24 and the scratch 23, a defect that is sufficiently small (small in size) with respect to the design value, such as a semiconductor wiring interval, hardly causes fatal damage to the product function. Conversely, if a large number of very large (large size) defects occur, production must be stopped immediately. That is, countermeasures differ depending on the size of the detected defect. Therefore, as shown by the discrimination result 44, a very small defect that is not a fatal defect is small in size, a defect that is likely to be fatal is medium in size, and a large defect that is supposed to be fatal is always large in size. Classified into one category. The number of categories need not be three, and one or a plurality of categories may be provided depending on the application. Further, although three categories are uniformly set for all the classifications of the foreign matter 24 and the scratch 23, a different number of categories may be provided for each classification of the foreign matter or each scratch. In addition, subtotals were obtained for each category. Further, the total was obtained for each foreign matter 24 and scratch 23. Further, a total defect including the total of the foreign matter 24 and the scratch 23 may be displayed.

  As described above, the overall control unit 9 is configured to be able to output subtotals or totals for each category of the foreign matter 24 and various scratches 23 as the discrimination result 44, so that it is actually used at the manufacturing site. In some cases, it is possible to manage the subtotal or total number of each category, and as a result, it is possible to efficiently monitor the occurrence of scratches 23 and foreign matter 24 and take appropriate countermeasures. Will be able to.

  Furthermore, the overall control unit 9 has a frequency distribution display function in order to see the defect size distribution in detail. As shown in FIG. 22, the horizontal axis represents the received light brightness during epi-illumination, and the vertical axis represents the frequency. In this figure, the frequency distribution display is performed only for the defect data classified as tire marks 23aa. Various combinations are conceivable for this, such as displaying all the defects recognized as scratches for each classification of the foreign matter 24 and the scratch 23, or all defects including the foreign matter 24 and the scratch 23 combined. Although not shown, a combination can be easily selected by a pull-down menu or the like. Moreover, you may calculate the brightness | luminance for each photoelectric conversion means 7a-7i on a horizontal axis, or the light reception luminance sum of the photoelectric conversion means 7a-7i about arbitrary combinations. Further, the same processing may be performed on data during oblique illumination. By using this frequency distribution display function, the defect distribution can be analyzed in detail, and at the same time, the scattered light distribution can be analyzed in detail.

  In addition, dedicated analysis tools are effective for analyzing these inspection data, but using commercially available spreadsheet software that can easily perform various processes on its own is effective in shortening the evaluation period. Is. Accordingly, a part or all of the selected inspection data is saved in a storage device 31 such as a hard disk or a floppy (registered trademark) disk in a format that can be read by spreadsheet software for items as shown in FIG. It was to be. In the present embodiment, the storage device 31 stores the identification number assigned to the defect, the discrimination result, the light reception brightness of each photomultiplier 7a to 7i during epi-illumination, and the light reception brightness of each photomultiplier 7a to 7i during oblique illumination. I am writing. All of these are not necessary. It may be meaningful to save defect coordinate data and the like. By reading this data with commercially available spreadsheet software, the overall control unit 9 can easily analyze the data of the detected defect, and can improve the discrimination ability in a short time. The above describes the method using the plurality of photoelectric conversion means 7a to 7i in order to evaluate the three-dimensional intensity distribution of the diffracted light.

  Next, a third embodiment of a surface inspection apparatus such as a scratch that easily obtains a two-dimensional diffracted light distribution will be described with reference to FIGS. In this embodiment, a detection optical system including lenses 108 and 106, CCD cameras 104 and 107, and a beam splitter 105 is added to the high-angle detection optical system 5a in the embodiment shown in FIG. 1 or FIG. . Therefore, also in the present embodiment, there are an epi-illumination system and an oblique illumination system, and the foreign matter 24 and the small scratch 23a are discriminated based on the luminance ratio. Further, the added CCD cameras 104 and 107 may be any means for two-dimensional photoelectric conversion such as a TDI sensor. In the present embodiment, two CCD cameras 104 and 107 are used. However, one CCD camera may be moved to two positions on the imaging plane and the Fourier transform plane. One CCD camera 104 is installed such that the camera imaging plane coincides with the imaging plane of the lens 108. The other CCD camera 107 is installed such that the camera image plane coincides with the Fourier transform plane of the lens 106. First, in the calculation unit 18a of the calculation processing unit 8, for example, a desired threshold value is obtained using the imaged image data obtained from the CCD camera 104 and A / D converted by the A / D conversion unit 16a and stored in the storage unit 17a. The position of the defects 23 and 24 is searched by converting the signal into a binarized signal and extracting a signal indicating the defect, and the searched results are output and stored in the storage device 31 as the position coordinates of the defect. The overall control unit 9 drives and controls the stage 15 by a control command from the stage controller 14 based on the searched position coordinates of the defect, and positions the defects 23 and 24 at the center of the visual field of the CCD camera 107.

  Next, the calculation unit 18b of the calculation processing unit 8 performs the following evaluation using image data obtained from the CCD camera 107, A / D converted by the A / D conversion unit 16a, and stored in the storage unit 17a. . For example, when there is a linear defect such as a large scratch 23b in the horizontal direction as shown in FIG. 25A, the diffracted light is distributed in the vertical direction in FIG. 25B on the Fourier transform plane. By evaluating this with the algorithm shown in FIG. 26, it becomes possible to evaluate the diffracted light distribution in the horizontal direction. First, in the calculation unit 18b, as shown in FIG. 25A, eight luminance evaluation regions indicated by circles in the figure are set around the point at which the 0th-order diffracted light is received. In this embodiment, the luminance evaluation area is a circle, but it is not absolutely necessary to be a circle. It may be a square or a polygon. In this embodiment, eight luminance evaluation regions are set, but there is no problem if the number is more or less than eight in accordance with the accuracy of evaluating the scattered light intensity distribution. It does not specify the number and shape. Then, the calculation unit 18b obtains the sum S (i) of the received light luminance of each pixel within each set luminance evaluation region. Using these eight luminance sums, the evaluation of the diffracted light directivity in the horizontal direction can be performed by the processing flows S65, S66, and S67 shown in FIG. 26, as in FIG. 15 or FIG. In the present embodiment, the light from the light source 2 is irradiated from a direction close to the normal line of the wafer 10, but it can also be used when irradiating from an oblique direction. In this embodiment, diffracted light in the normal direction of the wafer 10 is received, but diffracted light in an oblique direction may be received. In this embodiment, the pseudo vertical illumination in which the reflection mirror 4c is arranged at a position deviating from the NA of the lens 108 so as not to receive the regular reflection light from the wafer 10 is used. However, vertical illumination may be used.

  The method described above is a case where the wafer 10 without a wiring pattern is mainly handled. When this method is applied to the wafer 10 having the wiring pattern, a spatial filter 208 is provided as shown in FIG. However, depending on the wiring pattern, a spatial filter is not necessarily required. In this embodiment, as described in JP-A-6-258239, for example, it is necessary to remove a diffraction pattern based on diffracted light from a periodic wiring pattern using a linear spatial filter. Is described. The configuration of this embodiment will be described with reference to FIG. In this embodiment, a wafer 10 that is an object to be inspected, a light source 2, an optical path switching mechanism 3, an illuminating unit composed of reflecting mirrors 4a ", 4b, 4a ', 4a, 4c', a lens 6, a spatial filter 208, The detection unit is composed of a photoelectric conversion means 7. In this embodiment, the light beam of one light source 2 is reflected by the incident light 12 and the oblique illumination by the optical path switching mechanism 3 and the reflection mirrors 4a ", 4b, 4a ', 4a. However, the number of light sources and the number of reflection mirrors are not defined. Two light sources may be used, or more or fewer reflecting mirrors may be used. The discrimination principle and the discrimination processing method can be performed in the same manner as the above method. Further, the epi-illumination 12 may be either vertical illumination or pseudo vertical illumination shown in FIG. In the method described above, the scratch 23 and the foreign matter 24 are discriminated by paying attention to the uniqueness of the shape of the scratch 23 and analyzing the distribution and intensity of the scattered light generated therefrom in the arithmetic processing unit 8. Furthermore, the detailed shape classification of the scratch 23 was performed.

  Next, a fourth example of the surface inspection apparatus such as a scratch for realizing the first embodiment of the present invention will be described with reference to FIGS. That is, the fourth embodiment is a technique for performing unevenness discrimination processing by paying attention to the convex foreign matter 24 and the concave scratch 23. In order to avoid misunderstanding, the discrimination method between the foreign matter 24 and the scratch 23 described so far does not recognize the unevenness, but the width W and the depth D or the scratches 23 and the foreign matter 24 are limited. This is a technique that focuses on the difference in height aspect ratio.

The fourth embodiment includes an illumination optical system 1b including a light source 300 and a half mirror 302 for illuminating from the vertical direction of the wafer 10, a phase delay filter 305, a phase advance filter 306, which are installed on a Fourier transform plane. A detection optical system composed of a beam splitter 304 and a photoelectric conversion means 307 and 310 for branching light that has passed through each of them, and a difference signal processing section that takes a difference in detection luminance obtained from the photoelectric conversion means 307 and 310 308 and an arithmetic processing unit 8 configured by a discrimination processing unit 309 that recognizes irregularities from the result. In the fourth embodiment, Photomal A310 and Photomal B307 are used as the photoelectric conversion means. First, the phase lag filter 305 and the phase advance filter 306 will be described with reference to FIGS. 29 and 30. The phase delay filter 305 causes a delay in the phase of light in the vicinity where the 0th-order diffracted light is transmitted with respect to the peripheral portion. Specifically, the thickness is increased to t + d only in the vicinity where the 0th-order diffracted light is transmitted through an optical flat plate having a thickness t. Assuming that the refractive index of the plate is n, the optical path length L in the peripheral portion and the optical path length L ′ in the vicinity of the 0th-order diffracted light are expressed by the following equations (Equation 2) and (Equation 3) in consideration of the refractive index n _{0 of the} atmosphere. It can be obtained by an expression.

L = n × t + n ₀ × d (Equation 2)
L ′ = n × (t + d) (Equation 3)
That is, an optical path length difference ΔL1 expressed by the following equation (4) is generated in the light transmitted through the peripheral portion and the vicinity of the 0th-order diffracted light.

ΔL1 = L′−L = (n−n ₀ ) × d (Equation 4)
If the refractive index of the atmosphere is 1, the refractive index is approximately 1.5 when a glass material is used for the plate. Specifically, the optical path length difference ΔL represented by the following equation (5) is Arise.

ΔL1 = (1.5−1) × d = 0.5 × d (Equation 5)
Further, as shown in FIG. 30, the phase advance filter 306 makes the vicinity of the 0th-order diffracted light transmission as thin as td. Then, based on the same idea as described above, an optical path length difference ΔL2 is generated as shown in the following equation (6).

ΔL2 = (n ₀ −n) × d (Equation 6)
In the case of the above specific example, this optical path length difference ΔL2 is expressed by the following equation (Expression 7).

ΔL2 = −0.5 × d (Equation 7)
In the case of a light source having a wavelength λ, this means that the transmitted light in the vicinity of the 0th-order diffracted light with respect to the peripheral transmitted light has a phase delay θ1 (rad.), As shown in the following equations (8) and (9): And phase advance θ2 (rad.).

θ1 = ΔL1 / λ × 2π = (n−n ₀ ) × d / λ × 2π (Equation 8)
θ2 = ΔL2 / λ × 2π = (n ₀ −n) × d / λ × 2π (Equation 9)
In the above specific example, when a light source having a wavelength λ = 488 nm is used, the phase shift is specifically a numerical value represented by the following formulas (10) and (11).

θ1 = 0.5 × d / 488 × 2π (Equation 10)
θ2 = −0.5 × d / 488 × 2π (Equation 11)
Therefore, in order to cause a phase shift of θ1 = π / 2 and θ2 = −π / 2 between the phase delay θ1 and the phase advance θ2, d is then expressed by the following equations (12) and (13). What is necessary is just to make it the value shown by a formula.

θ1 = π / 2 = 0.5 × d / 488 × 2π
Therefore, d = 244 nm (Equation 12)
θ2 = −π / 2 = −0.5 × d / 488 × 2π
Therefore, d = 244 nm (Equation 13)
If the phase filters 305 and 306 designed in this way are used, the principle of discrimination between the foreign matter 24 and the scratch 23 described below can be realized.

  That is, the principle of discrimination between the foreign matter 24 and the scratch 23 will be described with reference to FIG. The laser irradiated on the wafer 10 is a plane wave having a uniform phase. Further, the light regularly reflected from the surface of the wafer having no defect is a plane wave having a uniform phase. This regular reflection light is referred to as reference reflection light. The light reflected from the foreign matter 24 that is a convex portion has a shorter optical path length than the reference reflected light. Therefore, the phase of the reflected light from the convex portion is advanced with respect to the reference reflected light. Conversely, the reflected light from the scratch 23, which is a recess, has an optical path length longer than that of the reference light by the amount of the dent, and the phase is delayed. That is, when the surface including the irregularities is irradiated with the laser beam 12, the phase of the laser reflected from the irregularities is delayed by the concave portion and the convex portion is advanced compared to the phase of the laser reflected from the flat portion. . Therefore, in the present invention, two types of phase filters 305 and 306 are inserted in the Fourier transform plane, and the phase advance and delay due to the unevenness are detected by the difference signal processing unit 308, and the detected phase advance and Based on the delay, the discrimination processing unit 309 discriminates the unevenness, so that it is possible to discriminate between the foreign matter 24 and the scratch 23, and the storage device 31 connected to the overall control unit 9 together with the position coordinates. Can be memorized.

  Furthermore, it demonstrates in detail using FIGS. 32-34. First, the difference signal strength for detecting the scratch 23 that is a recess in the difference signal processing unit 308 and the discrimination processing unit 309 of the arithmetic processing unit 8 will be described in detail with reference to FIG. This phase vector diagram shows a phase delay in the clockwise direction and a phase advance in the counterclockwise direction with reference to the phase of the reference reflected light that does not undergo phase change. The photomultiplier A 310 receives the light transmitted through the phase advance filter 306. The photomultiplier B307 receives the light transmitted through the phase delay filter 305. The specularly reflected light component from the wafer surface is collected in the form of dots on the Fourier transform plane and passes through a phase change region formed at the center of the phase filter. For this reason, the phase of the reference light transmitted through the phase advance filter 306 is a vector 321 advanced 90 degrees counterclockwise on the phase vector diagram. Further, the scattered light generated from the scratch 23 becomes substantially parallel light on the Fourier transform plane and passes through the peripheral regions of the phase filters 306 and 305. For this reason, the scattered light from the scratch 23 whose phase is relatively delayed with respect to the reference transmitted light becomes a vector whose phase is shifted clockwise as shown on the left side of the phase vector diagram of FIG. On the imaging surface, the reference reflected light 321 with the advanced phase and the scattered light 231 with the delayed phase interfere to form an image. In this phase vector diagram, the sum of the reference light vector 321 whose phase is advanced by 90 ° by the phase filter 306 and the scattered light vector 231 causing the phase delay due to the concave portion, and the photomal A detection light vector 310a in the figure are It represents an imaging vector caused by the interference. That is, the brightness of the image formation is the length of the photomal A detection light vector 310a.

  Similarly, the luminance of the image formed through the phase delay filter 305 is indicated by the sum of the reference light vector 322 whose phase is delayed by 90 ° by the phase filter 305 and the scattered light vector 231 in which the phase delay is caused by the concave portion. The photomal B detection light vector 307a on the right side of the phase vector. When the magnitudes of these two vectors 310a and 307a are compared, the photomal B detection light vector 307a is larger. That is, the magnitude of the combined vector is larger when the phase of the reference light is shifted in the same direction as the phase shift of the scattered light detected by the photomultiplier B307. Accordingly, in the difference signal processing unit 308 of the arithmetic processing unit 8, the difference signal obtained by subtracting the photomal B detection luminance 307 a from the photomal A detection luminance 310 a becomes negative. Can be discriminated.

  FIG. 34 shows an example of a dark field image when a two-dimensional photoelectric conversion unit, for example, a CCD camera is used as the photoelectric conversion units 310 and 307. Each luminance distribution indicates a luminance profile in the a-a ′ or b-b ′ section on the image data. The left half of FIG. The dark field image transmitted through the phase advance filter 306 is darker than the dark field image transmitted through the phase delay filter 305. In this way, when the two-dimensional photoelectric conversion means is used, the difference signal processing unit 308 of the arithmetic processing unit 8 may take the difference between the maximum values in the detected luminance profile.

  Next, the difference signal intensity for detecting the foreign matter 24 that is a convex portion in the difference signal processing unit 308 and the discrimination processing unit 309 of the arithmetic processing unit 8 will be described in detail with reference to FIG. In the foreign matter 24 that is a convex portion, the phase of the scattered light 241 advances with respect to the reference light. For this reason, as for the intensity at the imaging plane, the light reception intensity 310 b of the photomultiplier A 310 through the phase advance filter 306 is stronger than the light reception intensity 307 b of the photomultiplier B 307 through the phase delay filter 305. Therefore, the difference signal obtained in the difference signal processing unit 308 becomes positive and can be discriminated as the foreign matter 24 that is a convex portion in the discrimination processing unit 309. Then, as shown in the right half of FIG. 34, in the case of the foreign matter 24, the dark field image transmitted through the phase advance filter 306 becomes brighter. In this way, when the two-dimensional photoelectric conversion means is used, the difference signal processing unit 308 of the arithmetic processing unit 8 may take the difference between the maximum values in the detected luminance profile.

  As described above, the difference signal of the luminance signal obtained from the difference signal processing unit 308 is positive / negative depending on the uneven shape of the defect composed of the scratch 23 and the foreign matter 24. Therefore, the discrimination processing unit 309 can discriminate between the scratch 23 and the foreign matter 24 by determining whether this is positive or negative. Further, it is possible to easily convert the difference signal intensity into depth or height information in the discrimination processing unit 309. In the first to fourth embodiments described above, only the case where the scratch 23 is included as a defect type has been described. In the first half of the first to fourth embodiments, shape classification using the directivity of diffracted light was performed only for those classified in advance as scratches 23 or only for those classified as foreign matter 24. However, of course, the arithmetic processing unit 8 can easily realize the defect shape classification by using the function of classifying the defect of only the foreign matter 24 and the defect of only the scratch 23 using this directivity. is there. It is also possible to use a combination of the unevenness discrimination method based on the phase difference described in the second half of the first to fourth embodiments and the diffracted light distribution evaluation method described in the first half. According to the invention described above, an ADC (Automatic Defect Classification) or an on-the-fly ADC can be realized.

  Next, an embodiment of defect inspection near the wafer edge will be described with reference to FIGS. First, the case where this embodiment is applied to the condensing detection optical system shown in FIGS. 10 and 11 will be described. That is, FIG. 35 shows a case where a defect 402 such as a scratch 23 or a foreign object 24 is attached in the vicinity of the wafer edge portion 403. In this case, when the illumination light 11, 12 such as a laser is applied to the defect 402 such as the scratch 23 or the foreign material 24, the edge 403 is also included in the light beam d. Further, the scattered light from the wafer edge 403 is distributed on the vertical plane in the edge normal direction as shown in FIG. Therefore, as shown in FIG. 37, when viewed from above the wafer, the scattered light 404 from the edge 403 of the wafer 10 is strongly distributed in the direction of the edge normal. Further, the scattered light 405 from the defect 402 does not show a remarkable directivity. Therefore, as shown in FIG. 38, if one or two of the detection optical system positioned in the edge tangential direction, the B detection optical system in FIG. 38, or the B ′ detection optical system is used, compared to the scattered light 404 from the edge 403. It becomes possible to detect the scattered light 405 from the defect 402 with high intensity. Further, as the detection optical system, one or more of a C detection optical system, a C ′ detection optical system, a D detection optical system, and a D ′ detection optical system may be used. Further, the comparison calculation unit 18 in the calculation processing unit 8 determines the strength of the directivity of the scattered light by calculating the detection luminance ratio between the A detection optical system and the A ′ detection optical system and the detection optical system. Then, it may be determined whether only the edge 403 or the defect 402 is included.

  Next, the case where this embodiment is applied to the imaging detection optical system shown in FIG. 24 will be described with reference to FIG. This is a case where the spatial filter 407 shown in FIG. 40 is installed on the Fourier transform plane and the spatial filter light-shielding portion 408 is installed in the wafer edge normal direction. As shown in FIG. 36, the scattered light from the wafer edge portion 403 is distributed with a strong directivity from the wafer edge portion 403 in the normal direction. Therefore, by inserting the spatial filter 407, the scattered light from the edge portion 403 is shielded, and the scattered light from the defect 402 can be received by the photoelectric conversion means 107. In the present embodiment, a case where an Rθ stage (horizontal rotation stage) is provided as the stage 15 is shown. In this case, the direction of the wafer edge irradiated with the detection optical system and illumination light such as laser light is always relatively constant including the orientation flat. Therefore, the positions of the detection optical systems 5b and 5c used in the condensing optical system may be constant without changing the light shielding direction by the spatial filter 407 in the imaging optical system.

  Further, when the stage 15 is constituted only by the XY stage, it is necessary to devise such as changing the detection optical systems 5b and 5c used in accordance with the edge direction of the wafer, or making the spatial filter 407 a viewpoint. As described above, according to the above-described embodiment, foreign matter in the vicinity of the wafer edge can be detected with high sensitivity. Therefore, a high yield can be achieved by instantly finding a defect in a process that easily attaches foreign matter to the peripheral portion of the edge. It becomes possible.

DESCRIPTION OF SYMBOLS 1a, 1b ... Illumination optical system, 2 ... Light source, 3 ... Optical path switching mechanism, 4a-4d, 4c '... Reflection mirror, 5 ... Detection optical system, 5a ... High angle detection optical system, 5b ... Low angle detection optical system, 5c: Medium angle detection optical system, 6, 6a to 6i ... Condensing lens, 7, 7a to 7i ... Photoelectric converter (photomal), 8 ... Arithmetic processing unit, 9 ... Overall control unit, 10 ... Inspected object ( Wafer), 11 ... oblique illumination light, 12 ... epi-illumination light, 12a ... vertical illumination light, 12b ... pseudo-vertical illumination light, 14 ... stage controller, 15 ... stage, 16, 16a-16i ... A / D converter, 17,17A～17i ... storage unit, 18 ... comparing unit (comparison determination unit), 21 ... substrate, 22 ... insulating film (SiO ₂ film), 23 ... scratch, 23a ... small scratches, 23aa ... tire marks, 23ab ... Dimple marks, 23ac ... rough surface, 23 ... large scratch, 23c ... huge scratch, 24 ... foreign matter, 20, 50 ... discrimination line (threshold), 31 ... storage device, 32 ... input means, 33 ... display device, 34 ... network, 40 ... screen, 41 ... scratch discrimination Luminance distribution graph, 42 ... Large scratch discrimination luminance distribution graph, 43 ... Defect map corresponding coordinate search, 44 ... Discrimination result display window (histogram display), 51 ... Substrate installation table, 104 ... CCD camera, 105 ... Beam splitter, DESCRIPTION OF SYMBOLS 106 ... Lens, 107 ... CCD camera, 108 (6) ... Lens, 208 ... Spatial filter, 300 ... Light source, 302 ... Half mirror, 303 ... Lens, 304 ... Beam splitter, 305 ... Phase delay filter, 306 ... Phase advance filter 307 ... Photomal B, 308 ... Difference signal processing unit, 309 ... Discrimination processing unit, 31 ... photomultiplier A.

Claims (8)

A first illumination system that illuminates the inspection area of the sample surface from a normal direction or a direction near the sample surface , and a second illumination that illuminates the inspection area of the sample surface from an oblique direction with respect to the sample surface An illumination optical system equipped with a system;
A detection optical system having detectors respectively disposed so as to surround an illumination area of the sample surface illuminated by the first illumination system or the second illumination system of the illumination optical system;
Based on the result of classification by the first ratio between the magnitude of the detection signal of the scattered light by the first illumination system and the magnitude of the detection signal of the scattered light by the second illumination system , A defect inspection apparatus comprising: a signal processing system that classifies defects using a second ratio of detection signals of scattered light respectively detected by detectors arranged at orthogonal positions as a center; . The defect inspection apparatus according to claim 1,
In the signal processing system, the detection signal of the scattered light detected by the detector having the highest scattered light luminance from the illumination region, and the detector having the highest scattered light luminance with respect to the inspection region of the sample surface, A defect inspection apparatus that classifies defects using a detection signal of scattered light detected by a detector arranged at an orthogonal position. The defect inspection apparatus according to claim 1,
In the signal processing system, a detector disposed on a first plane perpendicular to the sample surface passing through the inspection region, and orthogonal to the first plane and passing through the inspection region on the sample surface A defect inspection apparatus that classifies defects using detection signals of scattered light respectively detected by a detector disposed on a second surface perpendicular to the second surface. The defect inspection apparatus according to any one of claims 1 to 3,
The defect is characterized in that the signal processing system classifies the defect by comparing a ratio of detection signals of scattered light respectively detected by the detectors arranged at the orthogonal positions with a predetermined threshold value. Inspection device. The defect inspection apparatus according to any one of claims 1 to 4,
2. The defect inspection apparatus according to claim 1, wherein the detection optical system includes two or more detectors arranged at different elevation angles with respect to the sample surface. The defect inspection apparatus according to claim 5,
The defect inspection apparatus, wherein the detectors arranged at the orthogonal positions are arranged at the same elevation angle with respect to the sample surface. The defect inspection apparatus according to any one of claims 1 to 6,
The defect inspection apparatus, wherein the signal processing system corrects the magnitude of a detection signal of scattered light detected by the detector. A defect inspection apparatus according to any one of claims 1 to 7 ,
The signal processing system classifies a defect into a scratch and a foreign object by comparing a ratio of magnitudes of detection signals of scattered light detected by the detectors arranged at the orthogonal positions with a predetermined threshold value. Defect inspection device characterized by.

Images