JP4772815B2 - Correction pattern image generation apparatus, pattern inspection apparatus, and correction pattern image generation method - Google Patents

Description

  The present invention relates to a correction pattern image generation device, a pattern inspection device, and a correction pattern image generation method. For example, the present invention relates to a correction pattern image generation device, a pattern inspection device, and a correction pattern image generation method applied to defect inspection of a semiconductor lithography mask. .

  In recent years, the circuit line width required for a semiconductor element has been increasingly narrowed as a large scale integrated circuit (LSI) is highly integrated and has a large capacity. These semiconductor elements use an original pattern pattern (also referred to as a mask or a reticle, hereinafter referred to as a mask) on which a circuit pattern is formed, and the pattern is exposed and transferred onto a wafer by a reduction projection exposure apparatus called a stepper. It is manufactured by forming a circuit.

  In addition, improvement in yield is indispensable for manufacturing an LSI that requires a large amount of manufacturing cost. One of the major factors that reduce the yield is a pattern defect of a mask used when an ultrafine pattern is exposed and transferred onto a semiconductor wafer. In recent years, with the miniaturization of LSI pattern dimensions formed on semiconductor wafers, the dimensions that must be detected as pattern defects have become extremely small. Therefore, it is necessary to improve the accuracy of a pattern inspection apparatus that inspects defects in a transfer mask used in LSI manufacturing.

  Methods for inspecting pattern defects are roughly classified into die-to-die comparison (Die to Die comparison: DD comparison) inspection and die-to-database comparison (Die to Database comparison: DB comparison) inspection. The DD comparison inspection is a method for detecting a defect by comparing measurement data (inspection reference pattern image data and inspection pattern image data) of two dies on a reticle. The DB comparison inspection is a method for detecting defects by comparing die measurement data (inspected pattern image data) and die design data (inspection reference pattern image data) generated from LSI design CAD data. It is.

  In the DB comparison inspection, the sample is placed on the stage of the pattern inspection apparatus, and the stage is moved so that the light beam scans on the sample and the inspection is performed. The sample is irradiated with a light beam by a light source and an illumination optical system. The light transmitted or reflected by the sample is imaged on the sensor via the optical system. The image picked up by the sensor is sent to the comparison circuit as measurement data. The comparison circuit compares the measured data and the design data after aligning the images according to an appropriate algorithm, and determines that there is a pattern defect if they do not match.

  Here, in recent years, with the miniaturization of the pattern on the lithography mask, it has become necessary to detect minute defects that are buried in the image misalignment of images to be compared, image expansion / contraction, undulation, sensing noise, and the like. . Therefore, it is necessary to align the inspection reference pattern image and the inspection pattern image with high accuracy. Furthermore, image correction is important for detecting these defects. For this reason, conventionally, in the first stage of comparing and inspecting two images of the inspection reference pattern image and the pattern image to be inspected, after the alignment of both images, the correction of the expansion / contraction of the image (for example, see Patent Document 1) and the image Image correction such as swell correction, resizing correction, and noise averaging processing was sequentially performed.

  However, repeating such corrections causes a cumulative error, which is a major factor for image deterioration. Furthermore, if the inspection reference pattern image and the pattern image to be inspected are too close to each other after correction, it becomes difficult to detect the defect. That is, excessive correction is counterproductive.

Further, Patent Document 2 discloses image correction using an input / output prediction model that integrates final fine adjustment alignment from the sub-pixel level and image correction, and has less image degradation and is effective image correction. Has been. In this image correction, for example, alignment and image correction in units of sub-pixels are realized at the same time using a linear prediction model using the inspection reference pattern image as input data and the pattern image to be inspected as output data. In this case, a model parameter (coefficient) is identified by creating a relational expression of a matrix from image data and solving simultaneous equations. Then, based on the linear prediction model to which the identified model parameter is applied, the inspection reference pattern image is corrected to generate a corrected pattern image.
JP 2000-241136 A Japanese Patent No. 3965189

  The present invention provides an apparatus and a method for generating a correction pattern image in which an image of an inspection reference pattern is appropriately corrected in order to overcome the above-described problems and appropriately inspect an inspection pattern image of a sample to be inspected. The purpose is to do.

A correction pattern image generation device according to one aspect of the present invention is a correction pattern image generation device that corrects an inspection reference pattern image used for pattern inspection of a sample, and aligns an inspection pattern image with the inspection reference pattern image. A relative shift amount calculation unit that calculates a relative shift amount in units of subpixels, and an image shift unit that shifts the inspection reference pattern image in one of the vertical and horizontal shift directions based on the relative shift amount. And a first equation generation unit that generates a first simultaneous equation describing an input / output relationship using a first linear prediction model for the inspection pattern image and the shifted inspection reference pattern image, A first parameter estimating unit for obtaining a parameter of the first simultaneous equation, and the parameter of the first simultaneous equation being substituted A first correction pattern image generation unit that generates a first correction pattern image obtained by correcting the inspection reference pattern image in a direction perpendicular to the shift direction using a linear prediction model of 1; A second equation generation unit that generates a second simultaneous equation describing an input / output relationship using a second linear prediction model for the image and the first correction pattern image; and the second simultaneous equation The first correction pattern image is the same as the shift direction by using a second parameter estimation unit for obtaining the parameters of the second linear prediction model in which the parameters of the second simultaneous equations are substituted. A second correction pattern image generation unit that generates a second correction pattern image corrected in the direction , wherein the first linear prediction model has an area of 1 × n (n is an integer of 3 or more) pixels. The input is a 1 × n-order model, and the second linear prediction model is an n × first-order model having an input of an n × 1 pixel area, or the first linear prediction model is n × 1-order model that receives the region of the n × 1 pixel, and said second linear prediction model Ru Ah at 1 × n following model that receives the region of 1 × n pixels It is characterized by that.

  A pattern inspection apparatus according to an aspect of the present invention includes the correction pattern image generation apparatus.

A correction pattern image generation method according to an aspect of the present invention is a correction pattern image generation method for correcting an inspection reference pattern image used for pattern inspection of a sample, and aligns an inspection pattern image and the inspection reference pattern image. Performing a relative shift amount calculating step for calculating a relative shift amount in units of sub-pixels, an image shift step for shifting the inspection reference pattern image in either the vertical or horizontal direction based on the relative shift amount, and A first equation generating step for generating a first simultaneous equation describing an input / output relationship using a first linear prediction model for the pattern image to be inspected and the shifted inspection reference pattern image; A first parameter estimating step for obtaining parameters of the simultaneous equations, and a parameter of the first simultaneous equations A first correction pattern image generation step of generating a first correction pattern image obtained by correcting the inspection reference pattern image in a direction perpendicular to the shift direction using the first linear prediction model, A second equation generating step for generating a second simultaneous equation describing an input / output relationship using a second linear prediction model for the inspection pattern image and the first correction pattern image; A second parameter estimation step for obtaining parameters of two simultaneous equations, and the second linear prediction model in which the parameters of the second simultaneous equations are substituted, and the first correction pattern image is shifted by the shift includes a second correction pattern image generating step of generating a second correction pattern image corrected in the same direction as the direction, the said first linear prediction model A 1 × n-order model that receives a region of × n (n is an integer of 3 or more) pixels, and the second linear prediction model is an n × first-order that has a region of n × 1 pixels as an input. Or the first linear prediction model is an n × 1 order model with an input of an n × 1 pixel region, and the second linear prediction model is 1 × n pixels. characterized Oh Rukoto at 1 × n following model which receives the area.

  Here, the first linear prediction model is a 1 × n-order model that receives an area of 1 × n (n is an integer of 3 or more) pixels, and the second linear prediction model is n ×. The first linear prediction model is an n × first order model with an input of an n × 1 pixel area, and It is desirable that the second linear prediction model is a 1 × n-order model with a 1 × n pixel region as an input.

  ADVANTAGE OF THE INVENTION According to this invention, in order to test | inspect suitably the to-be-inspected pattern image of a test object sample, it becomes possible to provide the apparatus and its method which produce | generate the correction | amendment image which corrected the image of the test | inspection reference pattern appropriately.

  Even in the correction pattern image generation method described above with reference to Patent Document 2, for example, with respect to a hatched pattern or the like, there is a situation in which appropriate correction cannot be performed due to an indefinite pattern shift direction. sell. In addition, it is difficult to determine whether correction is appropriate for a hatched pattern or the like. In such a case, the use of an inspection reference pattern image that is not appropriately corrected causes a decrease in the inspection accuracy of the inspection target sample.

  Therefore, in the following embodiments, when performing image correction in a pattern inspection apparatus such as a lithography mask inspection apparatus, the vertical direction and the horizontal direction are corrected separately. According to this image correction, an appropriate inspection reference pattern can be corrected. Hereinafter, an effective image correction apparatus, pattern inspection apparatus, and image correction method adapted to various pattern images will be described.

[First embodiment]
FIG. 2 is a block diagram illustrating a configuration of the pattern inspection apparatus according to the first embodiment. In FIG. 2, a pattern inspection apparatus 100 for inspecting a defect of a sample using a substrate such as an exposure mask or wafer on which a pattern is formed includes an optical image acquisition unit 150 and a control system circuit 160. The optical image acquisition unit 150 includes an XYθ table 102, a light source 103, an enlargement optical system 104, a photodiode array 105, a sensor circuit 106, a laser length measurement system 122, an autoloader 130, and an illumination optical system 170. In the control system circuit 160, the control computer 110 serving as a computer transmits a position circuit 107, a comparison circuit 108, a development circuit 111, a reference circuit 112, an autoloader control circuit 113, and a table control circuit 114 via a bus 120 serving as a data transmission path. , Magnetic disk device 109, magnetic tape device 115, flexible disk device (FD) 116, CRT 117, pattern monitor 118, and printer 119. The XYθ table 102 is driven by an X-axis motor, a Y-axis motor, and a θ-axis motor.

  In FIG. 2, description of components other than those necessary for describing the present embodiment is omitted. It goes without saying that the pattern inspection apparatus 100 usually includes other necessary configurations.

  FIG. 1 is a block diagram showing an internal configuration of the comparison circuit of FIG. In FIG. 2, the comparison circuit 108 includes a memory 142, a memory 144, an alignment unit 140, a correction pattern image generation unit 200, and a comparison unit 146.

  The memory 142 stores reference data (inspection standard pattern image data) input from the reference circuit 112. The memory 144 stores measurement data (inspected pattern image data) input from the sensor circuit 106. Here, the correction pattern image generation unit 200 is an example of a correction pattern image generation apparatus.

  The correction pattern image generation unit 200 includes a relative shift amount calculation unit 202, an image shift unit 204, a first equation generation unit 206, a first parameter estimation unit 208, a first correction pattern image generation unit 210, and a second correction pattern image generation unit 210. Equation generating unit 212, second parameter estimating unit 214, and second correction pattern image generating unit 216.

  In FIG. 1, description of components other than those necessary for describing the present embodiment is omitted. The alignment unit 140, the correction pattern image generation unit 200, or the comparison unit 146 may include other configurations. The functions of the alignment unit 140, the correction pattern image generation unit 200, and the comparison unit 146 may be configured by software that can be executed by a computer, for example. However, the present invention is not limited to this. For example, it may be implemented by hardware using an electric circuit. Or you may make it implement by the combination of the hardware and software by an electrical circuit. Alternatively, a combination of such hardware and firmware may be used. Similarly, the relative shift amount calculation unit 202, the image shift unit 204, the first equation generation unit 206, the first parameter estimation unit 208, the first correction pattern image generation unit 210, the second equation generation unit 212, the first The functions of the second parameter estimation unit 214 and the second correction pattern image generation unit 216 may be configured by software executable by a computer, for example. Or you may carry out by the hardware by an electric circuit. Or you may make it implement by the combination of the hardware and software by an electrical circuit. Alternatively, a combination of such hardware and firmware may be used.

  Hereinafter, the operation of the pattern inspection apparatus 100 will be described with reference to FIG. First, as an optical image acquisition step, the optical image acquisition unit 150 acquires an optical image of the photomask 101 serving as a sample on which a pattern is formed based on the design data. Specifically, the optical image is acquired as follows.

  A photomask 101 to be inspected is placed on an XYθ table 102 provided so as to be movable in a horizontal direction and a rotation direction by motors of XYθ axes. Then, the pattern formed on the photomask 101 is irradiated with light by an appropriate light source 103 disposed above the XYθ table 102. The light beam emitted from the light source 103 irradiates the photomask 101 via the illumination optical system 170. A magnifying optical system 104, a photodiode array 105, and a sensor circuit 106 are arranged below the photomask 101, and the light transmitted through the photomask 101 is optically imaged on the photodiode array 105 via the magnifying optical system 104. Is imaged and incident. The magnifying optical system 104 may be automatically focused by an unillustrated autofocus mechanism.

  FIG. 3 is a diagram for explaining an optical image acquisition procedure. As shown in FIG. 3, the inspection area is virtually divided into a plurality of strip-shaped inspection stripes having a scan width W in the Y direction, and each of the divided inspection stripes is continuously scanned. Thus, the operation of the XYθ table 102 (FIG. 2) is controlled, and an optical image is acquired while moving in the X direction.

  In the photodiode array 105 (FIG. 2), images having a scan width W as shown in FIG. 3 are continuously input. Then, after acquiring the image of the first inspection stripe, the image of the scan width W is continuously input in the same manner while moving the image of the second inspection stripe in the opposite direction. When an image in the third inspection stripe is acquired, the image moves while moving in the direction opposite to the direction in which the image in the second inspection stripe is acquired, that is, in the direction in which the image in the first inspection stripe is acquired. To get. In this way, it is possible to shorten a useless processing time by continuously acquiring images.

  The pattern image formed on the photodiode array 105 is photoelectrically converted by the photodiode array 105 and further A / D (analog-digital) converted by the sensor circuit 106. For example, a sensor such as a TDI (Time Delay Integrator) sensor is installed in the photodiode array 105. By continuously moving the XYθ table 102 serving as a stage in the X-axis direction, the TDI sensor images the pattern of the photomask 101 serving as a sample. These light source 103, magnifying optical system 104, photodiode array 105, and sensor circuit 106 constitute a high-magnification inspection optical system.

  The XYθ table 102 is driven by the table control circuit 114 under the control of the control computer 110. It can be moved by a drive system such as a three-axis (XY-θ) motor that drives in the X, Y, and θ directions. As these X-axis motor, Y-axis motor, and θ-axis motor, for example, a step motor can be used. The movement position of the XYθ table 102 is measured by the laser length measurement system 122 and supplied to the position circuit 107. The photomask 101 on the XYθ table 102 is automatically conveyed from the autoloader 130 driven by the autoloader control circuit 113, and is automatically discharged after the inspection is completed.

  Measurement data (pattern image data to be inspected: optical image) output from the sensor circuit 106 is sent to the comparison circuit 108 together with data indicating the position of the photomask 101 on the XYθ table 102 output from the position circuit 107. The measurement data is, for example, 8-bit unsigned data, and represents the brightness gradation of each pixel. The measurement data is compared for each image data of 512 pixels × 512 pixels, for example.

  On the other hand, design data used at the time of pattern formation of the photomask 101 is stored in a magnetic disk device 109 which is an example of a storage device. The design data is read from the magnetic disk device 109 to the development circuit 111 through the control computer 110. Then, the development circuit 111 converts the read design data of the photomask 101 to be inspected into binary or multivalued image data, and the image data is sent to the reference circuit 112.

  Here, the design data is a rectangle or triangle as a basic figure. For example, the coordinates (x, y) at two vertex positions of the figure, or a figure code serving as an identifier for distinguishing the figure type such as a rectangle or a triangle. The graphic data defining the shape, size, position, etc. of each pattern graphic is stored in the information. When design data as such graphic data is input to the expansion circuit 111, it is expanded to data for each graphic. Then, a graphic code indicating a graphic shape of the graphic data, a graphic dimension, and the like are interpreted. Then, it is developed into binary or multi-value graphic pattern data as a pattern arranged in a grid having a grid of a predetermined quantization dimension as a unit.

In other words, the design data is read, and the occupation ratio occupied by the graphic indicated by the graphic data in the design data is calculated for each cell formed by virtually dividing the inspection area as a cell having a predetermined dimension as a unit, and the n-bit Occupancy data is generated and output to the internal pattern memory. For example, it is preferable to set one square as one pixel. If a resolution of 1/2 ⁸ (= 1/256) is given to one pixel, 1/256 small areas are allocated by the figure area arranged in the pixel, and the occupation ratio in the pixel is set. Calculate. Then, it is generated as 8-bit occupancy data and stored in an internal pattern memory.

  Then, the reference circuit 112 creates reference data (inspection standard pattern image data) for comparison with measurement data from the graphic image data sent from the development circuit 111. The reference data to be compared is created as image data of 512 pixels × 512 pixels, for example, like the measurement data.

  Here, reference data is created based on design data in order to perform “die to database inspection”, but the present invention is not limited to this. A “die to die inspection” can also be performed. In that case, reference data may be created based on another measurement data (optical image) to be compared. Then, the reference data is sent to the comparison circuit 108.

  The comparison circuit 108 captures reference data and measurement data. The reference data (inspection standard pattern image data) is stored in the memory 142 shown in FIG. Measurement data (pattern image data to be inspected) is stored in the memory 144. Then, the alignment unit 140 shifts pixel by pixel and corrects the pixel position deviation. That is, the positional deviation between both images is driven to less than one pixel. The evaluation function represented by the sum of squares of the gradation differences of the corresponding pixels of both images may be adjusted in advance to a position where the evaluation function is minimized.

  After that, after correcting the reference data (inspection standard pattern image data) by a method described in detail later to generate a corrected pattern image, the corrected pattern image data and measurement data are compared according to a predetermined algorithm to determine the presence or absence of defects. To do.

  FIG. 4 is a flowchart showing the main steps of the correction pattern image generation method according to the present embodiment. 4, the correction pattern image generation method includes a relative shift amount calculation step S10, an image shift step S12, a first equation generation step S14, a first parameter estimation step S16, a first correction pattern image generation step S18, and a first correction pattern image generation step S18. A series of steps of an equation generation step S20, a second parameter estimation step S22, and a second correction pattern image generation step S24 are performed. After that, as an inspection process, a comparison step S26 for comparing the second correction pattern image with the pattern image to be inspected is performed.

  In the relative shift amount calculation step S10, the relative shift amount calculation unit 202 (FIG. 1) aligns the inspection pattern image and the inspection reference pattern image, and compares the inspection pattern image and the inspection reference pattern image in units of sub-pixels. The shift amount is calculated.

ここで、ｐ、ｑは次式で与えられる。

本実施の形態においても、画像のシフト演算はバイキュービック補間法で行うものとするが、画像のシフト演算はバイキュービック補間法に限定されず、直線補間法やスプライン補間法などを用いることができる。 In general, when the horizontal shift amount in sub-pixel units is l and the vertical shift amount is m, an image u ″ obtained by shifting the image u by l and m is expressed by the following formula called bicubic interpolation method ( 1-1) and equation (1-2). As a method of obtaining the shift amount, for example, there is a method of obtaining the shift amount that minimizes the sum of squares of the gradation difference after shifting by shifting the shift amount within a predetermined range.

Here, p and q are given by the following equations.

Also in this embodiment, the image shift calculation is performed by the bicubic interpolation method. However, the image shift calculation is not limited to the bicubic interpolation method, and a linear interpolation method, a spline interpolation method, or the like can be used. .

  In this embodiment, hereinafter, the image to be inspected is y, the inspection reference image is u, and the inspection reference pattern image is shifted based on the relative shift amount obtained in the relative shift amount calculation step S10, that is, the image is shifted. The inspection reference image is represented by v, the first correction pattern image is represented by w, and the second correction pattern image is represented by x.

  In the image shift step S12, the image shift unit 204 (FIG. 1) shifts the inspection reference pattern image u only in the vertical direction based on the relative shift amount obtained in S10. Specifically, the inspection reference pattern image v is generated by shifting the inspection reference pattern image only in the vertical direction by setting l = 0 in the expressions (1-1) and (1-2).

  Next, in a series of steps of the first equation generation step S14, the first parameter estimation step S16, and the first correction pattern image generation step S18, the inspection reference pattern image u is shifted in the direction shifted in the image shift step S12. Correct the vertical direction. In this embodiment, since the inspection reference pattern image u is shifted in the vertical direction in the image shift step S12, correction in the direction perpendicular to this, that is, in the horizontal direction is performed. In the image shift step S12, as a result of performing the alignment in the vertical direction in units of sub-pixels, it is possible to perform correction with higher accuracy in the horizontal direction.

  First, in the first equation generation step S14, the first equation generation unit 206 uses the data of the inspection pattern image y and the data of the shifted inspection reference pattern image v corresponding to the first linear prediction. Generate simultaneous equations based on the model. Hereinafter, a method for setting the first linear prediction model with the shifted inspection reference pattern image v as input data and the inspection pattern image y as output data will be described.

Here, a 1 × 5 order linear prediction model using a 1 × 5 pixel region is used as the first linear prediction model. The suffixes used in this model are shown in Table (1) below.

As shown in Table (1), the suffix is taken, and the input data (gradation value) and output data (gradation value) are v (i, j) and y (i, j) or v _{i, j} and y _{i, respectively. , J.} FIG. 5 is a diagram for explaining the first linear prediction model. Let the suffixes (coordinates) of the pixel 20 of interest in the pattern image 10 to be inspected be i and j. Then, in the shifted inspection reference pattern image 12, the suffixes of a total of five pixels, that is, the pixel 26 having the same coordinates and the two pixels 28 on the left and right of the pixel 26 are set as shown in Table (1). Then, a relational expression such as Expression (3) shown below is set for a set of pixel data in a 1 × 5 region.

Here, a _{0 to} a ₄ to be multiplied by the inspection reference pattern image data v (i, j) shifted in each coordinate in the relational expression shown in Expression (3) are model parameters (coefficients) to be identified. The expression (3) means that the data y _k = y (i, j) of one pixel 20 in the pattern image 10 to be inspected corresponds to one pixel 26 of the corresponding shifted inspection reference pattern image 12 and its left and right. This means that it can be expressed by a linear combination of 1 × 5 pixel data composed of the pixels 28. Thus, by performing the convolution operation with the linear prediction model, it is possible to correct the swell and noise of the image. When Expression (3) is expressed in a vector format (matrix format), it can be expressed as Expression (4-1) to Expression (4-3).

ここで、Ａ＝［ｘ_１，ｘ_２，…ｘ_ｎ］^Ｔ、ｙ＝［ｙ_１，ｙ_２，…ｙ_ｎ］^Ｔ、ｋ＝１，２，…ｎである。 Here, in the equations (4-1) to (4-3), the coordinates (i, j) of the shifted inspection reference pattern image v and the inspection pattern image y can be scanned to obtain five sets of data. Model parameters can be identified. Here, from a statistical point of view, n (> 5) sets of data are prepared as in the following equation (5-1), and 5 based on the least square method as in the following equation (5-2). It is also preferable to solve α-dimensional simultaneous equations and identify α. As a method of solving such an equation, there is a maximum likelihood estimation method in addition to the least square method, and any method may be used.

Here, A = [x ₁ , x ₂ ,... X _n ] ^T , y = [y ₁ , y ₂ ,... Y _n ] ^T , k = 1, 2,.

For example, if the shifted inspection reference pattern image 12 and the inspected pattern image 10 are each 512 × 512 pixels, about 250,000 sets of data can be obtained, and a sufficient number can be ensured statistically. Can do. The first equation generation unit 206 (FIG. 1) calculates simultaneous equations of the equations (5-1) and (5-2) from the pattern image to be inspected and the pattern image to be inspected, that is, a 5 × 5 matrix (A ^T A) and a five-element vector A ^T y are generated.

In the first parameter estimation step S16, the first parameter estimation unit 208 calculates the parameter α (coefficient) by solving the simultaneous equations of the equations (5-1) and (5-2). That is, the parameters a _{0 to} a ₄ (coefficients) are calculated.

As the first correction pattern image generation step S18, the first correction pattern image generation unit 210 uses the first linear prediction model in which the parameters of the first simultaneous equations are substituted, and shifts in the image shift step S12. A first correction pattern image w is generated by correcting the previous inspection reference pattern image u in a direction perpendicular to the shift direction of the image shift step S12, here the horizontal direction. That is, the first correction pattern image data w _{i, j} at the coordinates (i, j) is calculated using the parameter a _j obtained in the first parameter estimation step S16 as shown in the following equation (6). To do. Then, the calculation result is output to the second equation generation unit 212. Here, u _{i, j} is inspection reference pattern image data.

  As described above, the first correction pattern image generation unit 210 applies the correction in the horizontal direction to the inspection reference pattern image u using the parameter α obtained by the first parameter estimation unit 208. One correction pattern image w is generated.

  Next, the first correction pattern image w is shifted in the image shift step S12 in a series of steps of the second equation generation step S20, the second parameter estimation step S22, and the second correction pattern image generation step S24. Correct in the same direction as the direction. In this embodiment, since the image shift step S12 is shifted in the vertical direction, the same correction in the vertical direction is performed. That is, the vertical correction is performed on the first correction pattern image w corrected in the horizontal direction.

  First, in the second equation generation step S20, the second equation generation unit 212 uses the data of the inspection pattern image y and the data of the first correction pattern image w corresponding to the second linear prediction model. Generate simultaneous equations based on. Hereinafter, a method for setting the second linear prediction model by regarding the first correction pattern image w as input data and regarding the inspection pattern image y as output data will be described.

Here, a 5 × 1 linear prediction model using a 5 × 1 pixel region is used as the second linear prediction model. The suffixes used in this model are shown in Table (2) below.

As shown in Table (2), a suffix is taken, and input data (gradation value) and output data (gradation value) are w (i, j) and y (i, j), respectively. FIG. 6 is a diagram for explaining the second linear prediction model. The suffixes (coordinates) of the pixel 20 of interest in the pattern image to be inspected 10 are i and j. Then, in the first correction pattern image 14, the suffixes of a total of five pixels, that is, the pixel 22 having the same coordinates and the two pixels 24 above and below the pixel 22, are set as shown in Table (2). Then, a relational expression such as Expression (7) below is set for a set of pixel data in a 5 × 1 region.

として、次の式（９）の第２の連立方程式を生成する。

ここで、Ａ＝［ｘ_１，ｘ_２，…ｘ_ｎ］^Ｔ、ｙ＝［ｙ_１，ｙ_２，…ｙ_ｎ］^Ｔ、ｋ＝１，２，…ｎである。 Here, as in the first equation generation step S14

As a result, a second simultaneous equation of the following equation (9) is generated.

Here, A = [x ₁ , x ₂ ,... X _n ] ^T , y = [y ₁ , y ₂ ,... Y _n ] ^T , k = 1, 2,.

  As the second parameter estimation step S22, the second parameter estimation unit 214 solves the simultaneous equations of Equation (9) and multiplies the first correction pattern image w of each coordinate in the simultaneous equations by the parameter β (coefficient). An operation for calculating is performed.

In the second correction pattern image generation step S24, the second correction pattern image generation unit 216 uses the obtained second correction model parameter b _j as shown in the following equation (10), and coordinates (i , J) of the second correction pattern image data x _{i, j} is calculated. Then, the calculation result is output to the comparison unit 146. Here, w _{i, j} is the first correction pattern image data.

  As described above, the second correction pattern image generation unit generates the second correction pattern image x from the first correction pattern image w using the parameter β obtained by the second parameter estimation unit. .

As the comparison step S26, the comparison unit 146 compares the second correction pattern image data x _{i, j} with the inspected pattern image data y (i, j) using a predetermined algorithm. Then, the presence or absence of a defect is inspected, and the result is output.

  FIG. 7 is a diagram for explaining the operation of the present embodiment. Here, a case where correction for matching pattern A with pattern B is performed will be described as an example. Pattern A and pattern B are individually included in two different images, and are displayed in an overlapping manner in the figure. When the correction parameter is two-dimensional, that is, when the correction is shifted simultaneously in the horizontal and vertical directions, the amount of shift in the direction parallel to the pattern edge is arbitrary as shown by the broken line arrow in the figure. As a result, the contribution of the shift parameter obtained by the estimation in the direction parallel to the edge of the pattern becomes indefinite, and a good shift parameter may not be obtained. For example, if there is a defect image in the shifted image, the defect image may be blurred in a direction parallel to the edge of the pattern.

  On the other hand, when the correction parameters are limited to one dimension as in the first pattern image generation and the second pattern image generation of the present embodiment, that is, the correction shift is either horizontal or vertical. In the case of being fixed, there is a place where the edge positions of the two patterns coincide as indicated by solid arrows in the figure. Thus, since the degree of freedom of the coefficient to be estimated is constrained by the edge position of the pattern, a good shift parameter can be obtained. However, if the image is occupied only by a horizontal or vertical pattern, the horizontal or vertical parameter may be indefinite. Even in this case, since the shift parameter is one-dimensional, it can be easily determined that the parameter is abnormal by determining the center of gravity of the parameter. Moreover, Formula (1-1), (1-2) and Formula (6), (10) are the same shape. Therefore, the abnormal parameter can be replaced with the aforementioned bicubic interpolation parameter p or q, and a good corrected image can be obtained by the operation.

  As described above, the second correction pattern image x generated in the above-described step is subjected to the inspection reference image u while avoiding the deterioration of the defect signal due to the indefinite shift direction of the correction. This corresponds to an image corrected in accordance with the inspection pattern image y. In particular, due to the effect of sequentially estimating the horizontal direction and the vertical direction, good estimation can be performed when a certain direction component, such as a hatched pattern, becomes indefinite. Therefore, by comparing the second correction pattern image x obtained here with the pattern to be inspected y, a highly accurate inspection can be performed.

  In the present embodiment, an example has been described in which the inspection reference pattern image is shifted in the vertical shift direction based on the relative shift amount, and then the horizontal correction and the vertical correction are sequentially performed. . However, the same effect can be obtained by shifting the inspection reference pattern image in the horizontal shift direction based on the relative shift amount, and then sequentially performing the vertical correction and the horizontal correction.

  In addition, the first linear prediction model is a 1 × 5 order model using a 1 × 5 pixel region, and the second linear prediction model is a 5 × 1 order model using a 5 × 1 pixel region. The case of a model has been described as an example. From the viewpoint of the accuracy of correction and the calculation processing time, about 1 × 5th order and 5 × 1st order are appropriate. However, the 1 × 5 order model and the 5 × 1 order model are not necessarily used, and a 1 × n order model and an n × 1 order model in which n is an arbitrary integer of 3 or more should be adopted. Is possible. The first linear prediction model is an n × 1 order model using an area of n × 1 pixels, and the second linear prediction model is an 1 × n order model using an area of 1 × n pixels. It does not matter as a model.

  FIG. 8 is a diagram for explaining another optical image acquisition procedure. In the configuration of FIG. 2, the photodiode array 105 that simultaneously enters the number of pixels of the scan width W (for example, 2048 pixels) is used. However, the present invention is not limited to this, and as shown in FIG. Each time a constant pitch movement is detected by a laser interferometer while scanning at a constant speed in the direction, a laser scanning optical device (not shown) scans the laser beam in the Y direction, detects transmitted light, A procedure for obtaining a two-dimensional image for each size area may be used.

  In the above description, what is described as “˜circuit” or “˜step” can be configured by a computer-operable program. Or you may make it implement by not only the program used as software but the combination of hardware and software. Alternatively, a combination with firmware may be used. When configured by a program, the program is recorded on a recording medium such as a magnetic disk device, a magnetic tape device, an FD, or a ROM (Read Only Memory). For example, the table control circuit 114, the expansion circuit 111, the reference circuit 112, the comparison circuit 108, and the like that constitute the arithmetic control unit in FIG. 2 may be configured by an electrical circuit or processed by the control computer 110. It may be realized as software that can. Moreover, you may implement | achieve with the combination of an electrical circuit and software.

  The embodiments have been described above with reference to specific examples. However, the present invention is not limited to these specific examples. For example, in each embodiment, transmitted light is used, but reflected light or transmitted light and reflected light may be used simultaneously. Although the reference image to be the inspection standard pattern image is generated from the design data, data of the same pattern captured by a sensor such as a photodiode array may be used. In other words, a die to die inspection or a die to database inspection may be used.

  In addition, although descriptions are omitted for parts and the like that are not directly required for the description of the present invention, such as a device configuration and a control method, a required device configuration and a control method can be appropriately selected and used. In addition, any correction pattern image generation apparatus, pattern inspection apparatus, or correction pattern image generation method that includes the elements of the present invention and that can be appropriately modified by those skilled in the art is included in the scope of the present invention.

It is a block diagram which shows the internal structure of the comparison circuit of embodiment. It is a block diagram of the pattern inspection apparatus of an embodiment. It is a figure for demonstrating the acquisition procedure of the optical image of embodiment. It is a flowchart figure shown about the principal part process of the correction | amendment pattern image generation method in embodiment. It is a figure for demonstrating the 1st linear prediction model of embodiment. It is a figure for demonstrating the 2nd linear prediction model of embodiment. It is a figure explaining the effect | action of embodiment. It is a figure for demonstrating the acquisition procedure of another optical image.

Explanation of symbols

100 pattern inspection apparatus 108 comparison circuit 140 alignment unit 146 comparison unit 200 correction pattern image generation unit 202 relative shift amount calculation unit 204 image shift unit 206 first equation generation unit 208 first parameter estimation unit 210 first correction pattern Image generation unit 212 Second equation generation unit 214 Second parameter estimation unit 216 Second correction pattern image generation unit

Claims (3)

A correction pattern image generation device for correcting an inspection reference pattern image used for pattern inspection of a sample,
A relative shift amount calculation unit that performs alignment between the inspection pattern image and the inspection reference pattern image and calculates a relative shift amount in sub-pixel units;
An image shift unit that shifts the inspection reference pattern image in either the vertical or horizontal shift direction based on the relative shift amount;
A first equation generation unit that generates a first simultaneous equation describing an input / output relationship using a first linear prediction model for the inspection pattern image and the shifted inspection reference pattern image;
A first parameter estimation unit for obtaining a parameter of the first simultaneous equations;
Using the first linear prediction model into which the parameters of the first simultaneous equations are substituted, a first correction pattern image is generated by correcting the inspection reference pattern image in a direction perpendicular to the shift direction. A first correction pattern image generation unit;
A second equation generation unit for generating a second simultaneous equation describing an input / output relationship using a second linear prediction model for the inspection pattern image and the first correction pattern image;
A second parameter estimation unit for obtaining a parameter of the second simultaneous equation;
Using the second linear prediction model in which the parameters of the second simultaneous equations are substituted, a second correction pattern image is generated by correcting the first correction pattern image in the same direction as the shift direction. A second correction pattern image generation unit that
Equipped with a,
The first linear prediction model is a 1 × n-order model that receives an area of 1 × n (n is an integer of 3 or more) pixels, and the second linear prediction model has n × 1 pixels. An n × 1 order model with a region as input, or
The first linear prediction model is an n × 1 order model that receives an n × 1 pixel region, and the second linear prediction model uses a 1 × n pixel region as an input 1 × n. characterized Oh Rukoto in the following model correction pattern image generating apparatus. A pattern inspection apparatus comprising the correction pattern image generation apparatus according to claim 1 . A correction pattern image generation method for correcting an inspection reference pattern image used for pattern inspection of a sample,
Relative shift amount calculating step for performing alignment between the pattern image to be inspected and the inspection reference pattern image, and calculating a relative shift amount in sub-pixel units;
An image shift step of shifting the inspection reference pattern image in either the vertical or horizontal direction based on the relative shift amount;
A first equation generating step for generating a first simultaneous equation describing an input / output relationship using a first linear prediction model for the inspection pattern image and the shifted inspection reference pattern image;
A first parameter estimating step for obtaining a parameter of the first simultaneous equations;
Using the first linear prediction model in which the parameters of the first simultaneous equations are substituted, a first correction pattern image is generated by correcting the inspection reference pattern image in a direction perpendicular to the shift direction. A first correction pattern image generation step,
A second equation generation step for generating a second simultaneous equation describing an input / output relationship using a second linear prediction model for the inspection pattern image and the first correction pattern image;
A second parameter estimating step for obtaining a parameter of the second simultaneous equation;
Using the second linear prediction model in which the parameters of the second simultaneous equations are substituted, a second correction pattern image is generated by correcting the first correction pattern image in the same direction as the shift direction. A second correction pattern image generation step,
Equipped with a,
The first linear prediction model is a 1 × n-order model that receives an area of 1 × n (n is an integer of 3 or more) pixels, and the second linear prediction model has n × 1 pixels. An n × 1 order model with a region as input, or
The first linear prediction model is an n × 1 order model that receives an n × 1 pixel region, and the second linear prediction model uses a 1 × n pixel region as an input 1 × n. correction pattern image generating method comprising Oh Rukoto in the following model.

Images