JP2012119512A - Substrate quality evaluation method and apparatus therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a substrate quality evaluation method which enables quick quality evaluation in an ELA process matched with quality evaluation values by a visual inspector or with final picture quality, and an apparatus therefor.SOLUTION: A substrate 5 to be evaluated is moved continuously in one direction while light is impinged on the substrate 5 from an oblique direction, whereby an image based on a primary diffracted light, which is produced by a polycrystalline silicon thin film formed on the substrate 5, is photographed to obtain a primary diffracted light image. Also, an image of scattered light near the optical axis of regular reflected light or transmitted light from the substrate 5 is photographed to obtain a scattered light image near the optical axis. The primary diffracted light image and the scattered light image near the optical axis thus obtained are processed to extract plural features, and at lease one or more of the extracted plural features are used to calculate a quality evaluation value of the polycrystalline silicon thin film formed on the substrate 5 according to preset evaluation standards, to evaluate the quality of the polycrystalline silicon thin film formed on the substrate 5.

Description

  The present invention relates to an inspection for evaluating the quality of a substrate from an image (detected image) of an object to be inspected obtained using light or a laser, and in particular, the quality of crystalline silicon formed on a TFT substrate, an organic EL substrate, or the like. The present invention relates to a substrate quality evaluation method and apparatus suitable for evaluation.

  At the production site of flat panels represented by TFT liquid crystal display devices and organic EL display devices, a portion of the glass substrate is irradiated at a low temperature by irradiating an amorphous silicon film formed on the glass substrate with a pulsed excimer laser. There is a step of forming crystalline silicon in the region. This process is referred to as an ELA (Excimer Laser Anneal) process. Here, in order to determine the final panel quality by changing the size of the crystalline silicon particles due to fluctuations in the irradiation laser energy, the quality of the crystalline silicon (particle size and variation) is evaluated, and an excimer laser annealing device is used. It is necessary to optimize the laser power of the laser and eliminate the poor quality substrate. For this reason, visual evaluation is performed by an inspector, but because it is a sensory test, the evaluation result varies depending on the inspector, and the quality is not stable. There is a need. One of the methods is using an electron microscope (SEM), but it is necessary to cut the substrate, takes time for evaluation, and is not practical on the production line. There is also an apparatus that irradiates the substrate with microwaves and lasers and evaluates the crystallinity by the dielectric constant of the crystal, but it is difficult to inspect all the substrates. In addition, as another technique for evaluating the quality of crystalline silicon at high speed in a production line, Japanese Patent Application Laid-Open No. 2006-19408 (Patent Document 1) irradiates the surface of crystalline silicon with visible light and reflects the reflected light. A method for acquiring an image and inspecting the superiority or inferiority of the crystal quality on the crystalline silicon surface from the luminance uniformity of the image, the contrast of the stripe pattern, and the like is described. Japanese Laid-Open Patent Publication No. 2001-308209 (Patent Document 2) discloses that an excimer laser annealing condition is optimized by irradiating light on a substrate subjected to excimer laser annealing treatment and monitoring diffracted light from the substrate. Are listed.

JP 2006-19408 A JP 2001-308209 A

  As described above, the quality evaluation of crystalline silicon is often a sensory evaluation by a visual inspector. For this reason, there are subtle differences in the determination criteria depending on the inspector, and it is necessary to set a determination criterion that summarizes them and update it appropriately. In addition, at the time of starting the process, it is assumed that a new quality deterioration factor may occur, and the determination criteria may change. Further, in the manufacturing process of the flat panel, which is an object to be inspected, a lighting inspection is performed as a final image quality evaluation after many steps after the ELA step. For this reason, there is a possibility that it is necessary to appropriately change the quality judgment criteria of the crystalline silicon while reflecting the result of the final image quality in consideration of the correlation between the quality of the crystalline silicon and the final image quality.

  In the method described in Patent Document 1, it is not described that the inspection standard is flexibly dealt with in accordance with a change in process, a change in inspection conditions, and the like.

  In the method described in Patent Document 2, only the use of information on the diffracted light from the substrate is described, so that the inspection standard can be flexibly changed according to a change in process or a change in inspection conditions. There is no mention of correspondence.

  The object of the present invention is to make it possible to flexibly cope with inspection standards in accordance with process changes, changes in inspection conditions, etc., in the ELA process, and is consistent with quality evaluation values and final image quality of visual inspectors. Another object of the present invention is to provide a quality evaluation method and apparatus for a substrate.

  In order to solve the above-mentioned problems, the present invention is an apparatus for evaluating the quality of a polycrystalline silicon thin film formed on a substrate, and the substrate to be evaluated can be placed and continuously moved in at least one direction. A first-order diffracted light image is obtained by irradiating light on the table means and the substrate placed on the table means from an oblique direction to capture an image of the first-order diffracted light generated by the polycrystalline silicon thin film formed on the substrate. In addition, an image acquisition unit that captures an image of scattered light near the optical axis of specularly reflected light or transmitted light from the substrate and acquires a scattered light image near the optical axis, and first-order diffracted light acquired by the image acquisition unit A feature extracting unit that extracts a plurality of features by processing an image of the image and a scattered light image near the optical axis, and at least one of the plurality of features extracted by the feature extracting unit; Pre-set evaluation criteria Thus it was constructed and a quality evaluation value calculating means for calculating the quality evaluation value of the polycrystalline silicon thin film formed on a substrate.

  Further, in order to solve the above-mentioned problems, in the present invention, a method for evaluating the quality of a polycrystalline silicon thin film formed on a substrate is obliquely applied to the substrate while continuously moving the substrate to be evaluated in one direction. The first-order diffracted light image is obtained by capturing an image of the first-order diffracted light generated by the polycrystalline silicon thin film formed on the substrate by irradiating light from the direction, and the optical axis of specularly reflected light or transmitted light from the substrate A scattered light image near the optical axis is acquired by capturing an image of the scattered light in the vicinity, and a plurality of features are extracted by processing the acquired image of the first-order diffracted light image and the scattered light image near the optical axis. The quality evaluation value of the polycrystalline silicon thin film formed on the substrate is calculated using at least one or more of the extracted features according to the evaluation criteria set in advance.

  According to the present invention, it is possible to output a quality evaluation value consistent with the determination by visual observation / evaluation. In addition, the final quality can be predicted in advance by setting the criteria for calculating the quality evaluation value of the board in advance by setting multiple board images and the quality of the board after product completion. Became.

It is a block diagram which shows the schematic structure as one Embodiment of a quality evaluation apparatus. It is a block diagram which shows the concept of a structure of a quality evaluation apparatus. It is sectional drawing of a board | substrate which shows the state which irradiated the laser to the polycrystalline-silicon film | membrane on a board | substrate, and the 1st-order diffracted light was generated. It is the graph which expressed notionally the relationship between the laser power for annealing and the crystal grain diameter of a polycrystalline silicon thin film. It is a flowchart which shows the flow of the process which calculates the quality evaluation value of a board | substrate. It is a flowchart which shows the concept of the process which calculates the quality evaluation value of a board | substrate. It is a conceptual diagram of the image of the board | substrate with which the crystalline silicon used as evaluation object was formed. It is a flowchart which shows an example of the process which sets a determination standard from a user's viewpoint. It is a flowchart which shows an example of the process which sets a criterion from the quality evaluation value at the time of final lighting inspection. It is a block diagram which shows the schematic structure as another form of the illumination optical system of a quality evaluation apparatus. It is a flowchart which shows an example of the process which sets the determination standard of the quality of a board | substrate from the image obtained from the some optical system. It is a flowchart which shows an example of the process which calculates the quality evaluation value of a board | substrate from the image obtained from the some optical system.

  Embodiments of a substrate quality evaluation method and apparatus according to the present invention will be described with reference to the drawings. First, an embodiment of a quality evaluation apparatus for a substrate on which crystalline silicon is formed as an object to be inspected will be described.

FIG. 2 is a conceptual diagram showing an embodiment of a substrate quality evaluation apparatus 100 according to the present invention. The optical system 1 includes an illumination unit 4 and a plurality of detection units 7a and 7b. The illuminating unit 4 irradiates the evaluation object 5 (the substrate on which the crystalline silicon is formed) with the illumination conditions (for example, the irradiation angle, the illumination direction, the illumination wavelength, the polarization state, etc.) adjusted. Scattered light is generated from the inspection object 5 by the illumination light emitted from the illumination unit 4. Here, the light scattered in the direction of the detector 7a is referred to as scattered light 6a, and the light scattered in the direction of the detector 7b is referred to as scattered light 6b. The scattered light 6a and the scattered light 6b are detected as scattered light intensity signals by the detection unit 7a and the detection unit 7b, respectively. Each of the detected scattered light intensity signals is amplified and A / D converted by the A / D conversion unit 2 and input to the image processing unit 3.
The image processing unit 3 includes a preprocessing unit 8-1 and an evaluation value calculation unit 8-2 as appropriate. For the scattered light intensity signal input to the image processing unit 3, the pre-processing unit 8-1 calculates a determination criterion described later. The evaluation value calculation unit 8-2 performs processing described later according to the determination criterion generated by the preprocessing unit 8-1, calculates the quality evaluation value of the substrate, and outputs it to the overall control unit 9.

  Each of the scattered light 6a and the scattered light 6b indicates a scattered light distribution generated in the direction of the detection unit 7a and the detection unit 7b, respectively. If the optical conditions of the illumination light by the illumination unit 4 are different, the scattered light 6a and the scattered light 6b generated thereby are different from each other. In this embodiment, the optical properties and characteristics of scattered light generated by certain illumination light are referred to as scattered light distribution of the scattered light. More specifically, the scattered light distribution refers to a distribution of optical parameter values such as intensity, amplitude, phase, polarization, wavelength, and coherency with respect to the emission position, emission direction, and emission angle of the scattered light.

  Next, FIG. 1 shows a block diagram as an embodiment of a specific quality evaluation apparatus for realizing the configuration shown in FIG. That is, the quality evaluation apparatus 100 according to the present embodiment includes an illumination unit 4 that irradiates illumination light obliquely onto an object to be inspected (for example, a substrate 5 on which crystalline silicon is formed), and 1 from the substrate 5. A detection optical system (first-order diffracted light detection system) 7a that forms an image of scattered light in the direction of the next-order diffracted light and a detection optical system that forms an image of scattered light in the vicinity of the specularly reflected light from the sample 5 (scattered light near the specularly reflected light) (Detection system) 7b, an optical system 1 having sensor portions 11a and 11b that receive optical images formed by the respective detection optical systems and convert them into image signals, and images obtained by the sensor portions 11a and 11b. An A / D converter 2 that amplifies the signal and performs A / D conversion, an image processor 3, and an overall controller 9 are configured. Here, the details of the processing will be described using scattered light as an example. The same processing is performed on the diffracted light.

  The substrate 5 is mounted on a stage (XYZ-θ stage) 13 that can move and rotate in the XY plane and move in the Z direction perpendicular to the XY plane. The XYZ-θ stage 13 is It is driven by the mechanical controller 14. At this time, the substrate 5 is mounted on the XYZ-θ stage 13, and the scattered light from the foreign matter on the object to be inspected is detected while moving the XYZ-θ stage 13 in the horizontal direction. Thus, the detection result is obtained as a two-dimensional image.

The illumination light source 41 of the illumination unit 4 may use a laser or a lamp. Moreover, the wavelength of the light of each illumination light source may be a short wavelength, or may be light with a broad wavelength (white light). When short-wavelength light is used, light (Ultra Violet Light: UV light) having a wavelength in the ultraviolet region (160 to 400 nm) can be used in order to increase the resolution of the image to be detected (detect fine defects). . When the laser is used as the light source, if it is a single wavelength laser, the illumination unit 4 can be provided with means 43 for reducing coherence. The illumination unit 4 further includes a collimator lens 42 that collimates the light emitted from the illumination light source 41, and a beam that is long in one direction (in the case of FIG. 1, long in a direction perpendicular to the paper surface). A cylindrical lens 44 for converting into a shape is provided. Illumination conditions (for example, irradiation angle, illumination azimuth, illumination wavelength, polarization state, etc.) are selected or automatically selected by the user, and the illumination driver 12 performs setting and control according to the selection conditions.
Of the scattered light emitted from the substrate 5, the first-order diffracted light reflected by the regular projection pattern of the polycrystalline silicon film formed on the substrate 5 by ELA and the scattered light in the vicinity thereof are sent to the detection optical system 7a. And converted into an image signal by the sensor unit 11a. In addition, the light scattered from the substrate 5 in the direction of the specularly reflected light is shielded from the specularly reflected light by the spatial filter 71b, and the scattered light near the specularly reflected light is converted into an image signal by the sensor unit 11b via the detection optical system 7b. Converted. The detection optical system 7a includes an objective lens 72a, an optical filter 73a, and an imaging lens 74a, and is condensed and imaged on the sensor unit 11a. Similarly to 7a, the detection optical system 7b also has an objective lens 72b, an optical filter 73b, and an imaging lens 74b.
The sensor units 11a and 11b are a one-dimensional image sensor (CCD linear sensor) or a time delay integration image sensor (TDI image) configured by two-dimensionally arranging a plurality of one-dimensional image sensors. Sensor), the scattered light distribution is detected by the image sensor in synchronization with the movement of the XYZ-θ stage 12, and a two-dimensional image is obtained. Reference numerals 73a and 73b denote optical filters, such as an optical element capable of adjusting the light intensity such as an ND filter and an attenuator, a polarizing optical element such as a polarizing plate, a polarizing beam splitter, and a wave plate, or a bandpass filter and a dichroic mirror. Or a combination of them, and controls any one of the light intensity, polarization characteristics, and wavelength characteristics of the detection light, or a combination thereof.

  The image processing unit 3 calculates a quality evaluation value on the substrate 5 that is an object to be inspected, and a standard for calculating the quality evaluation value for the image signals input from the sensor units 11a and 11b. A pre-processing unit 8-1 that sets a quality evaluation value, an evaluation value calculation unit 8-2 that calculates a quality evaluation value from a detected image based on a set quality criterion, and accepts parameters and data input from the outside The data teaching unit 8-3 is set to the preprocessing unit 8-1 and the evaluation value calculation unit 8-2. In the image processing unit 3, for example, the data teaching unit 8-3 is configured by connecting the database 15.

  The overall control unit 9 includes a CPU (incorporated in the overall control unit 9) that performs various controls, receives display parameters and the like from the user, and displays display means and input means that display detected board images, quality evaluation values, and the like. A user interface unit (GUI unit) 16 and a substrate 17 processed by the image processing unit 3 and a storage device 17 for storing the feature amount, image, quality criterion, and the like are appropriately connected. The mechanical controller 14 drives the XYZ-θ stage 13 based on a control command from the overall control unit 9. The image processing unit 3 and the detection optical systems 7a and 7b are also driven by commands from the overall control unit 9.

  The substrate 5 to be evaluated is, for example, a glass substrate, and a large number of crystalline silicon is regularly formed on the surface. The overall control unit 9 continuously moves the substrate 5 by the XYZ-θ stage 13, and in synchronization with this, captures the image of the substrate from the sensor units 11a and 11b, and obtains two types of scattered light obtained. For each of the images (6a, 6b), various features are calculated, and a quality evaluation value is calculated by comparing the calculated feature value with a preset criterion.

  In FIG. 3A, a light source 350 is formed on a substrate 300 in a state where a part of an amorphous silicon thin film 310 formed on a glass substrate 300 is annealed with an excimer laser to form a polycrystalline silicon thin film 320 having a uniform crystal grain size. Illuminating light 351 is incident on the substrate 300 at an incident angle θ1 (an angle with respect to the normal direction of the substrate 300), and the first-order diffracted light 352 is generated in the θ2 direction on the surface side of the substrate 300.

  The grain size of the polycrystalline silicon thin film 320 formed by this excimer laser annealing depends on the irradiation energy of the excimer laser (the product of the laser power density and the irradiation time). That is, when the power of the laser irradiating the amorphous silicon thin film 310 is increased, as shown in FIG. 3B, the crystallization of the amorphous silicon thin film 310 starts to progress from a certain energy level, and the polycrystalline silicon thin film 320 grows. As the power of the irradiated laser is further increased, the grain size of the polycrystalline silicon thin film 320 grows larger.

  Here, when the polycrystalline silicon thin film 320 has a uniform grain size, protrusions are formed on the surface of the polycrystalline silicon thin film 320 at a substantially constant pitch P in accordance with the crystal grain size (FIG. 3A drawing). Projections are formed at a constant pitch in a direction perpendicular to the surface). The pitch P of the protrusions on the film surface varies depending on the crystal grain size of the polycrystalline silicon thin film 320.

  On the other hand, in the configuration shown in FIG. 3A, the incident angle θ1 of the illumination light emitted from the light source 350 with the wavelength λ to the substrate 300 and the first-order diffracted light generated from the substrate 300 on which the polycrystalline silicon thin film 320 is formed. Between the emission angle θ2 and the pitch P of the protrusions on the surface of the polycrystalline silicon thin film 320,

The relationship expressed by

  That is, if there is variation (spatial distribution) or variation (time-dependent change) in the power of the excimer laser when annealing the amorphous silicon thin film 310, the grain size of the polycrystalline silicon thin film 320 changes. As a result, the emission angle θ2 of the first-order diffracted light generated from the substrate 300 irradiated from the direction of θ1 changes. Accordingly, the grain size of the polycrystalline silicon thin film 320 can be estimated by detecting the first-order diffracted light 352 and obtaining the angle θ2.

Next, an example of the processing flow of the board quality evaluation value calculation unit 8-2 of the image processing unit 3 will be described. FIG. 4 shows an outline of the flow of processing for calculating the quality evaluation value from the image of the first-order diffracted light distribution obtained from the sensor unit 11a by scanning the stage 13 with respect to the substrate 5 shown in FIG. The evaluation value calculation unit 8-2 receives the determination criterion 410 set in advance and the image 420 of the substrate 5 detected by the sensor unit 11a. In the evaluation value calculation unit 8-2, first, noise that hinders evaluation is removed in order to reveal features that affect quality (S401). Examples of noise include electrical noise of the sensor, luminance shading that occurs depending on the detected elevation angle, and uneven illumination. Noise removal methods include smoothing and removal of a specific frequency band. Next, one or a plurality of feature amounts are calculated for the substrate image from which noise has been removed (S402).
Here, when the target is crystalline silicon, stripe-shaped unevenness due to fluctuations in the irradiation energy of the pulse excimer laser, luminance distribution in the substrate, and the like are examples of characteristics that affect quality. Therefore, as an example of the feature value to quantify these,
Feature (1): Contrast feature calculated from luminance projection in a striped uneven direction (2): For each pixel in the image, there is a distribution in the image of the luminance gradient intensity. FIG. 6 shows an example. 61 is an image of a substrate on which crystalline silicon is formed. The longitudinal direction of the laser formed long in one direction to irradiate the substrate having the amorphous silicon thin film formed on the surface by the excimer laser annealing apparatus is the X direction, and the scanning direction (feeding direction) is the Y direction. If the quality of the polycrystalline silicon formed by laser annealing deteriorates due to fluctuations in the irradiation laser energy or the distribution of energy in the longitudinal direction of the laser formed long in one direction, the stripe shape along the X direction as indicated by 61 Unevenness becomes noticeable. Therefore, assuming that the luminance value of each point of the detected image is f (x, y), the feature (1) calculates the average projection luminance of each pixel in the X direction as in (Equation 2), and calculates the average projection luminance. The amount of change in the Y direction (Equation 3) is taken as the contrast.

  The feature (2) is a feature in which the luminance gradient intensity is calculated for each pixel as in (Equation 4), and the histogram is obtained in the target region.

  Features (1) and (2) are examples of feature amounts, and any other feature may be used as long as it represents the features of the substrate. Then, all or some of the plurality of feature quantities are compared with a predetermined determination criterion (S403), and a board quality evaluation value is determined based on the determination criterion (S404). Next, the quality of the laser irradiation condition of the laser annealing apparatus is judged based on the quality evaluation value (S405), and the result is output (S406).

  FIG. 5 is a specific conceptual diagram for determining the quality evaluation value of the substrate in S404. 510 and 520 are images acquired under the same detection conditions for different substrates. When such an image of the substrate 5 is input to the image processing unit 3, first, feature amounts A, B,... Are calculated for each image of the substrate 5. Here, an example in which a quality evaluation value is calculated using two feature amounts A and B is shown. In the present invention, the determination criterion using the feature amounts A and B is determined in advance by the preprocessing unit 8-1.

Reference numeral 530 in FIG. 5 is an example of a feature space indicating a determination criterion. That is, a range that each quality evaluation value can take is set in a two-dimensional feature space with the features A and B as axes. 530 is
If “ThA1 <feature A ≦ ThA2” and “ThB1 <feature B ≦ ThB2” Then Level1
As described above, this is an example of a determination criterion in which each evaluation value range is divided by a rectangle so as to be parallel to the axis so that the evaluation value range can be determined by a threshold value. Then, the quality evaluation value is determined depending on where the feature amount calculated from the image of the evaluation target board is plotted on the feature space in which the range of each evaluation value (Level 1 to Level 4 of 530 in the figure) is set. .

  In the feature space 530 of FIG. 5, the feature amount calculated from the image of the substrate 510 is plotted in the Level 1 region, and the feature amount calculated from the image of the substrate 520 is plotted in Level 3. Evaluation values are determined as Level 1 and Level 3 (S404). In FIG. 5, the determination criterion in the feature space 530 is indicated by a rectangular area divided by a straight line parallel to the feature axis. However, it is also possible to divide by a straight line having an inclination with respect to the feature axis. Further, instead of a straight line, it is also possible to divide each evaluation value area with a curve in the feature space.

The area setting is
If “ThA1 <feature A ≦ ThA2” and “ThB1 <feature B ≦ ThB2” Then Level1
Such a threshold (ThA1, ThA2, ThB1, ThB2) can be set by the user. Also, a feature space such as 530 can be displayed on the user interface unit 16, and the range of each evaluation value can be directly set by the user on the GUI.

  In this embodiment, the criterion for determining the quality evaluation value is not limited to the setting in the two-dimensional feature space 530 shown in FIG. 5, but can be set in a higher-dimensional feature space. For example, in a three-dimensional feature space that uses three types of feature quantities, the quality criterion can be set as a plane or a curved surface. Furthermore, in a higher-dimensional feature space, the quality criterion is a hyperplane or a hypersurface. In this case, it is difficult for the user to set threshold values for a plurality of feature axes. Therefore, the present invention has a determination criterion setting function based on supervised learning.

  FIG. 7 is an example of a processing flow when the determination criterion 410 is set from the viewpoint of the user. First, the user visually observes the substrate 5 (S701), and sets a quality evaluation value on the substrate 5 (S702). This is repeated, and the substrate 5 covering the quality evaluation value and the evaluation value are prepared as a set. Next, an image of the evaluated substrate 5 is acquired by the optical system 1 (S703). The user sets the obtained image of the substrate 5 and the evaluation value of the substrate 5 set by the user as a set, and prepares a plurality of sets so as to cover each evaluation value, and inputs them to the image processing unit 3. When a plurality of images of the substrate 5 and a plurality of evaluation values are input to the image processing unit 3, the evaluation values set in S702 in the preprocessing unit 8-1 are known, that is, for the supervised substrate image. Then, noise is removed (S704), and a feature amount is calculated (S705). The feature amount is the same as that calculated by the evaluation value calculation unit 8-2. Then, the feature amount calculated from the image and the entire set of evaluation values of the image are learned (S706), and the determination criterion 710 is output.

  The learning executed in S706 uses various general classifiers such as a parametric method and a nonparametric method. As an example, assuming that the distribution of each evaluation value plotted in the feature space follows a normal distribution, the distribution of evaluation values calculated from the feature values of each image and the points plotted in the feature space are input as teachers There is a method of calculating a determination criterion by obtaining a probability of entering the image. This is because, if the feature quantities calculated from n images that are certain evaluation values are x1, x2,..., Xn, the discriminant function φ for making them the correct evaluation values is as follows: It is given in 6).

  Although the example in which the evaluation value of each substrate set by the user is determined from the observation result of the substrate itself has been described, the user may observe the image detected in 703 and give the evaluation value as a teacher. In any case, an evaluation value that includes the user's criteria is given to the board image as a teacher, learning is performed using a discriminator, and a determination criterion is generated, so that it matches the result of visual evaluation by an actual inspector. It is within the scope of the present invention to output a quality evaluation value. The determination criterion 710 created by the preprocessing unit 8-1 is used as the determination criterion 410 input to the quality evaluation value calculation unit 8-2 described in the processing flow of FIG.

  The setting of the determination criterion 710 performed by the preprocessing unit 8-1 is not limited to providing the user evaluation criterion set in S702 as a teacher. FIG. 8 shows an example. First, the substrate on which the crystalline silicon is formed is visually observed. (S801) On the other hand, the substrate 5 is imaged by the optical system 1 to obtain an image of the substrate 5 (S803) and stored. Furthermore, the quality evaluation value at the final lighting inspection after the visually observed substrate has undergone each manufacturing process is set as a set with the acquired image (S802), and a plurality of sets are prepared so as to cover each evaluation value. Input to process 3. In the subsequent processing, as in FIG. 7, the preprocessing unit 8-1 removes noise from the substrate image (S704), calculates the feature amount (S705), and calculates the feature amount calculated from the image and the image. The entire set of final lighting evaluation values is learned (S706), and the determination criterion 810 is output.

  In this example, the evaluation value containing the criteria for the final lighting test is given to the image of the board as a teacher, learning is performed using a discriminator, and a determination criterion is generated, so that quality evaluation that is consistent with the final quality is achieved. Output the value. In addition, the evaluation value given as a teacher for learning may be an evaluation value desired to be matched, such as an evaluation value obtained by observation with an electron microscope (SEM).

  As described above, the example of the form of the specific quality evaluation apparatus that realizes the configuration shown in FIG. 2 has been described according to FIG. 1, but the example of the quality evaluation apparatus 900 of another form of the optical system 1 is illustrated. 9 shows. That is, a configuration provided with an optical system 91 as shown in FIG. 9 will be described as another form of the quality evaluation apparatus according to the present embodiment. The optical system 91 is configured to irradiate the glass substrate 95 on which polycrystalline silicon is formed with the illumination unit 94 disposed on the back surface. In this case, in the detection optical system (first-order diffracted light detection system) 97a, the first-order diffracted light by the reflected light from the regular projection-like pattern of the polycrystalline silicon film formed on the glass substrate 95 is imaged. The image is detected by the unit 911a to obtain an image of the first-order diffracted light. On the other hand, in the detection optical system (scattered light detection system in the vicinity of the light transmitted through the glass substrate) 97b, the light transmitted through the glass substrate is shielded to form an image of the scattered light in the vicinity, and the image is detected by the sensor unit 911b. An image of scattered light near the transmitted light is obtained.

  The configuration of the illumination unit 94 is the same as the configuration of the illumination unit 4 described in FIG. 1, and includes an illumination light source 941, a collimating lens 942, an optical filter 943, and a cylindrical lens 944. The configuration of the detection optical system 97a for detecting the first-order diffracted light is the same as that of the first-order diffracted light detection system 7a described with reference to FIG. 1, and includes an objective lens 972a, an optical filter 973a, and an imaging lens 974a. Condensed and imaged on 911a. The configuration of the detection optical system 97b that detects the scattered light in the vicinity of the transmitted light that has passed through the glass substrate 95 is the same as that of the regular reflected light vicinity scattered light detection system 7b described in FIG. A light blocking filter 971b for blocking transmitted light, an objective lens 972b, an optical filter 973b, and an imaging lens 974b are provided, and condensed and imaged on the sensor unit 911b. As an advantage of irradiating the glass substrate 95 from the back surface, for example, the arrangement relationship between the illuminating unit 4 and the first-order diffracted light detection system 7a in FIG. 1 must be limited so that the structures do not interfere with each other. The first-order diffracted light detection system 97a and the illumination unit 91 can be arranged without restriction. In the configuration shown in FIG. 9, when the first-order diffracted light detection system 97a interferes with the detection optical system 97b that detects scattered light in the vicinity of the transmitted light, the detection optical system that detects scattered light in the vicinity of the transmitted light. The optical axis 97b may be installed so as to be inclined with respect to the optical axis of the transmitted light transmitted through the glass substrate 95. In this case, since the transmitted light transmitted through the glass substrate 95 that is shifted with respect to the optical axis of the detection optical system 97b is not put into the sensor unit 911b, the light shielding filter 971b is unnecessary.

Here, by having a plurality of detection systems, it is possible to obtain images of diffracted light and scattered light at various detection elevation angles, and this makes it possible to obtain a more accurate quality evaluation value of the substrate. An example will be described. First, one detection system (for example, the first-order diffracted light detection system 7a) can obtain an image of scattered light from the substrate 5 with high resolution. This makes it possible to evaluate the crystal grain size of the polycrystalline silicon film due to the excess or deficiency of the excimer laser irradiation power in the ELA process (quality evaluation value A). The other detection system (for example, the scattered light detection system 7b in the vicinity of the optical axis of specular reflection light or substrate transmission light) can obtain a scattered light signal having an intensity corresponding to the crystal grain size from the substrate 5 at a low resolution. Thereby, it is possible to evaluate the non-uniformity of the crystalline silicon due to the missing pulse of the pulse excimer laser or the stability of the laser shot (quality evaluation value B). The quality evaluation value A makes it possible to evaluate whether the laser power set by the excimer laser annealing apparatus in the ELA process is good or bad, and to set an optimum power value. Further, the quality evaluation value B makes it possible to evaluate the quality of the crystalline silicon for all the products. By making the judgment data based on the teaching data at the time of learning as the quality at the time of final lighting inspection, the final image quality of each substrate after the ELA process can be predicted in advance, and quality defects can be prevented.
As described above, instead of calculating each evaluation value from a plurality of images obtained from different detection systems, information of a plurality of images obtained from a plurality of different detection systems is integrated to obtain 1 It is also possible to calculate one quality evaluation value.

FIG. 10 shows an example of the procedure. First, detection is performed under different image acquisition conditions for one substrate 100, and images of a plurality of substrates 100 (for example, an image 100a by first-order diffracted light, an image 100b by regular reflected light or scattered light near the optical axis of transmitted light from the substrate, and so on). ) Is set (S1001), and the evaluation value 102 of the polycrystalline silicon film formed on the substrate from which this image has been acquired (any one from the user's viewpoint, based on the SEM observation result, etc.) is set. (S1002). The plurality of images 100 a and 100 b of the substrate 100 and the set evaluation value data are input to the image processing unit 3 as a set in association with each other. When a plurality of sets of board images and evaluation values are input to the image processing unit 3, the preprocessing unit 8-1 performs noise removal processing S1003a and S1003b on the respective images (100a, 100b), and features Is calculated (S1004a, S1004b). The noise removal processing S1003a and S1003b may be different processing. Also, the feature amounts calculated in the feature amount calculation S1004a and S1004b may be the same and of different types or different types. Then, all the sets of feature amounts calculated from a plurality of images and evaluation values of the substrate are learned for one substrate (S1005), and a criterion is output (S1006).
FIG. 11 shows an outline of the flow of processing for calculating the quality evaluation value of the substrate using the determination criterion calculated in FIG. In the evaluation value calculation unit 8-2, the determination criteria 1101 and the sensor unit 11a and the sensor unit 11b (or the sensor unit 911a and the sensor unit 911b) set in advance according to the processing flow of FIG. Images 1100a and 1100b are input (S1101, S1100a, S1100b). The evaluation value calculation unit 8-2 performs noise removal processing on each of the images 1100a and 1100b (S1102a and S1102b). Different methods for removing noise may be used. Next, a plurality of feature amounts are calculated for each of the board images from which noise has been removed (1103a, 1103b). The feature amounts calculated in the feature amount calculation steps S1103a and S1103b may be the same type or different. Then, all or some of the plurality of feature amounts individually calculated in the feature amount calculation steps S1103a and S1103b are compared with the criterion 1101 set in advance as shown in FIG. 10 (S1104), A quality evaluation value of the substrate is determined based on the determination criterion (S1105). Thereby, the quality evaluation value of a board | substrate can be determined from the some image detected on different optical conditions. Next, the quality of the laser irradiation condition of the laser annealing apparatus is determined based on the quality evaluation value (S1106), and the result (evaluation value) is output (S1107).
According to the present invention, it is possible to obtain an image similar to an inspector directly observing a substrate visually, and to generate a determination criterion by learning a human evaluation result as a known quality evaluation value for the image. It is possible to output a quality evaluation value that is consistent with human judgment.
Furthermore, by learning the final lighting inspection result as a known quality evaluation value and generating a determination criterion, it is possible to predict the final quality in advance for the substrate after the ELA process.
As a result, the optimum conditions of the ELA apparatus can be determined from the quantitative evaluation values, and the substrate quality can be stabilized. In addition, quality control, ELA device defect detection, and prevention of quality defects can be realized from the quantitative evaluation values of the total number of substrates.

  As described above, the embodiment of the present invention has been described by taking the quality evaluation for the crystalline silicon formed on the substrate as an example, but it can also be applied to image quality inspection of flat panels such as TFT liquid crystal panels and organic EL panels. is there. Further, the object of evaluation is not limited to the flat panel glass substrate and the image quality at the time of lighting, and can be applied to, for example, a quality inspection of a semiconductor wafer as long as the quality is quantified.

  DESCRIPTION OF SYMBOLS 1 ... Optical system 2 ... A / D conversion part 3 ... Image processing part 4a, 4b ... Illumination part 5 ... Board | substrate 7a, 7b ... Detection part 8-1 ... Before Processing unit 8-2 ... Evaluation value calculation unit 11a, 11b ... Sensor unit 9 ... Overall control unit 13 ... Stage.

Claims (10)

An apparatus for evaluating the quality of a polycrystalline silicon thin film formed on a substrate,
A table means for placing a substrate to be evaluated and continuously moving in at least one direction;
The substrate placed on the table means is irradiated with light from an oblique direction, and an image of the first-order diffracted light generated by the polycrystalline silicon thin film formed on the substrate is taken to obtain a first-order diffracted light image. Image acquisition means for capturing an image of scattered light near the optical axis of specularly reflected light or transmitted light from the substrate and acquiring a scattered light image near the optical axis;
Feature extraction means for processing the image of the first-order diffracted light image acquired by the image acquisition means and the image of the scattered light image near the optical axis to extract a plurality of features;
Quality evaluation for calculating a quality evaluation value of a polycrystalline silicon thin film formed on the substrate according to an evaluation criterion set in advance using at least one or more of the plurality of features extracted by the feature extraction means A board quality evaluation apparatus comprising: a value calculation means.   The evaluation criterion to be set in advance is further provided with determination criterion setting means for creating by learning using an image of the substrate on which the polycrystalline silicon thin film obtained by the image acquisition means is formed. 1. The board quality evaluation apparatus according to 1.   The determination criterion setting unit creates the determination criterion by learning an evaluation value set for the image of the board in association with a feature amount of the image of the board. Board quality evaluation equipment.   Laser irradiation conditions for determining the quality of the laser irradiation conditions of the laser annealing apparatus formed with the polycrystalline silicon thin film based on the quality evaluation values of the polycrystalline silicon thin film formed on the substrate calculated by the quality evaluation value calculating means 4. The board quality evaluation apparatus according to claim 1, further comprising pass / fail judgment means.   The image acquisition means captures an image of scattered light near the optical axis of specularly reflected light or transmitted light from the substrate, and shields the specularly reflected light or transmitted light from the substrate with a light shielding plate. 5. The substrate quality evaluation apparatus according to claim 1, wherein an image of scattered light not shielded by the plate is captured. A method for evaluating the quality of a polycrystalline silicon thin film formed on a substrate,
While the substrate to be evaluated is continuously moved in one direction, the substrate is irradiated with light from an oblique direction, and an image of the first-order diffracted light generated by the polycrystalline silicon thin film formed on the substrate is captured. While acquiring the next diffracted light image, capturing the image of the scattered light near the optical axis of the specularly reflected light or transmitted light from the substrate to obtain the scattered light image near the optical axis,
A plurality of features are extracted by processing the acquired image of the first-order diffracted light image and the image of the scattered light image near the optical axis,
A substrate characterized in that a quality evaluation value of a polycrystalline silicon thin film formed on the substrate is calculated according to an evaluation criterion set in advance using at least one or more of the extracted features. Quality evaluation method.   7. The substrate according to claim 6, wherein the setting of the evaluation criterion in advance is creation and setting by learning using an image of the substrate on which the obtained polycrystalline silicon thin film is formed. Quality evaluation method.   Setting the evaluation criterion in advance includes learning the evaluation value set for the acquired image of the substrate on which the polycrystalline silicon thin film is formed in association with the feature amount of the image of the substrate. 8. The substrate quality evaluation method according to claim 6, wherein a determination criterion is created.   9. The quality of laser irradiation conditions of a laser annealing apparatus in which the polycrystalline silicon thin film is formed is determined based on the calculated quality evaluation value of the polycrystalline silicon thin film formed on the substrate. A method for evaluating the quality of a substrate according to any one of the above.   Taking an image of scattered light near the optical axis of specularly reflected light or transmitted light from the substrate, the specularly reflected light or transmitted light from the substrate was shielded by a light shielding plate, and was not shielded by the light shielding plate The substrate quality evaluation method according to claim 6, wherein an image of scattered light is captured.

Images