JP2011196952A - Inspection device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection device and an inspection method capable of detecting a flaw at high speed while precisely aligning the pattern of an optical image and the pattern of a reference image.SOLUTION: The inspection method includes the steps of acquiring the optical image of a sample from an image sensor, determining an equation expressing the relationship of α, β, and the squared-sum of the difference between the optical image and the reference image on either the optical image or the reference image of determination reference with the movement quantities in the X direction and the Y direction as α (0≤α<1) and β (0≤β<1) respectively, and determining suitable movement quantities for aligning from a pair of (α, β) that the squared-sum of the differences obtained from this equation is minimum. The suitable movement quantities for aligning may be determined by solving the partial differential equation for α and β on this equation.

Description

  The present invention relates to an inspection apparatus and an inspection method, and more particularly to an inspection apparatus and an inspection method for performing defect inspection by comparing an optical image and a reference image.

  In recent years, the circuit line width required for a semiconductor element has become increasingly narrower as the large scale integrated circuit (LSI) is highly integrated and has a large capacity. The semiconductor element uses an original pattern pattern (a mask or a reticle, which will be collectively referred to as a mask hereinafter) on which a circuit pattern is formed, and the circuit is exposed and transferred onto a wafer by a reduction projection exposure apparatus called a stepper. Manufactured by forming. For manufacturing a mask for transferring such a fine circuit pattern onto a wafer, an electron beam drawing apparatus capable of drawing the fine pattern is used. Attempts have also been made to develop a laser beam drawing apparatus for drawing using a laser beam. The electron beam drawing apparatus is also used when drawing a circuit pattern directly on a wafer.

  Yield improvement is indispensable for the manufacture of LSIs that require a large amount of manufacturing costs. On the other hand, recent typical logic devices are required to form a pattern having a line width of several tens of nanometers. Here, as a major factor for reducing the yield, there are mask pattern defects and variations in process conditions during exposure transfer. Until now, with the miniaturization of the LSI pattern dimension formed on a semiconductor wafer, it has been performed to absorb the variation margin of process conditions by increasing the dimensional accuracy of the mask. For this reason, in mask inspection, the dimensions that must be detected as pattern defects are miniaturized, and it is necessary to detect position errors of extremely small patterns. For this reason, high inspection accuracy is required for inspection apparatuses that detect defects in transfer masks used in LSI manufacturing.

  Defect detection methods include a die-to-die inspection method and a die-to-database inspection method. The die-to-die inspection method compares the same patterns of chips with different masks when a plurality of chips having the same pattern configuration are arranged in a part or the whole of the same mask. Inspection method. According to this method, since the mask patterns are directly compared, highly accurate inspection can be performed with a relatively simple apparatus configuration. However, it is impossible to detect a defect that exists in common in both patterns to be compared. On the other hand, the die-to-database inspection method is an inspection method in which reference data generated from design pattern data used for mask manufacture is compared with an actual pattern on the mask. Since a mechanism for generating a reference image is required, the apparatus becomes large, but a strict comparison with design pattern data can be performed. This method can be used only when there is only one chip transfer area in one mask.

  In die-to-database inspection, light emitted from a light source is irradiated onto a mask to be inspected via an optical system. The mask is placed on a table, and light irradiated as the table moves scans the mask. The light transmitted or reflected through the mask forms an image on the image sensor via the lens, and the optical image captured by the image sensor is sent to the comparison unit as measurement data. In the comparison unit, the measurement data and the reference data are compared according to an appropriate algorithm. If these data do not match, it is determined that there is a defect (see, for example, Patent Document 1).

  By the way, it is known that a positional deviation of the pattern occurs between the optical image and the reference image due to an error of a laser interferometer for detecting the position of the stage on which the mask is placed, an error of the straightness of the stage, or the like. Yes. Therefore, pattern alignment (frame alignment) is performed before pattern comparison in these images. Conventionally, while shifting the pattern in the X-axis direction and the Y-axis direction in units of about 1 pixel or (1/16) pixel, the square sum of the differences between the two images is taken, and the position where the value is the minimum is obtained. The position was adjusted as the optimum position (see, for example, Patent Document 2).

JP 2008-112178 A JP 2009-168553 A

  However, in recent masks that have been miniaturized, if the pattern is shifted by a unit of about 1 pixel or (1/16) pixel, alignment cannot be performed with sufficient accuracy, and it is determined whether or not it is a defect. May be difficult.

  On the other hand, in order to obtain sufficient alignment accuracy, it is necessary to increase the amount of calculation for alignment. In other words, the calculation time is prolonged or the scale of hardware is increased, resulting in an increase in cost.

  The present invention has been made in view of these points. That is, an object of the present invention is to provide an inspection apparatus and an inspection method capable of detecting a defect at high speed while performing alignment between an optical image pattern and a reference image pattern with high accuracy.

  Other objects and advantages of the present invention will become apparent from the following description.

According to a first aspect of the present invention, an image of a sample on which a pattern is formed is formed on an image sensor to obtain an optical image of the sample;
A portion for aligning the optical image with a reference image as a reference for determination;
Comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold,
The portions to be aligned are α and β with respect to either one of the optical image and the reference image, where the movement amounts in the X and Y directions are α (0 ≦ α <1) and β (0 ≦ β <1), respectively. And a part for obtaining an equation representing the relationship between the square sum of the difference between the optical image and the reference image,
The present invention relates to an inspection apparatus including a portion for obtaining an optimal movement amount for alignment from a set of (α, β) that minimizes the sum of squares of differences obtained from this equation.

According to a second aspect of the present invention, an image of a sample on which a pattern is formed is formed on an image sensor to obtain an optical image of the sample;
A portion for aligning the optical image with a reference image as a reference for determination;
Comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold,
The portions to be aligned are α and β with respect to either one of the optical image and the reference image, where the movement amounts in the X and Y directions are α (0 ≦ α <1) and β (0 ≦ β <1), respectively. And a part for obtaining an equation representing the relationship between the square sum of the difference between the optical image and the reference image,
The present invention relates to an inspection apparatus characterized by having a part for obtaining a movement amount optimum for alignment by solving partial differentials of α and β for this equation.

A third aspect of the present invention is an inspection method for illuminating a sample on which a pattern is formed, forming an image of the sample on an image sensor via an optical system, and determining the presence or absence of defects.
Obtaining an optical image of the sample from the image sensor;
For either one of the optical image and the reference image serving as a reference for determination, the movement amounts in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively, and α and β Obtaining an equation representing the relationship between the square sum of the difference between the optical image and the reference image;
A step of obtaining an optimal amount of movement for alignment between the optical image and the reference image from a set of (α, β) that minimizes the sum of squares of the differences obtained from this equation;
Aligning the optical image and the reference image according to the optimum amount of movement;
A step of comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold value.

According to a fourth aspect of the present invention, there is provided an inspection method for illuminating a sample on which a pattern is formed, forming an image of the sample on an image sensor via an optical system, and determining the presence or absence of a defect.
Obtaining an optical image of the sample from the image sensor;
For either one of the optical image and the reference image serving as a reference for determination, the movement amounts in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively, and α and β Obtaining an equation representing the relationship between the square sum of the difference between the optical image and the reference image;
Obtaining the optimal amount of movement for alignment of the optical image and the reference image by solving the partial differential of α and β for the equation;
Aligning the optical image and the reference image according to the optimum amount of movement;
A step of comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold value.

  In the third aspect or the fourth aspect of the present invention, the reference image is a reference image created from pattern design data, or an optical image of the same pattern as a pattern in a region different from the optical image on the sample. Preferably there is.

  ADVANTAGE OF THE INVENTION According to this invention, the inspection apparatus and the inspection method which can detect a defect at high-speed can be provided, aligning with the pattern of an optical image and the pattern of a reference | standard image with high precision.

It is a block diagram of the inspection apparatus in this Embodiment. It is a conceptual diagram which shows the flow of the data in this Embodiment. It is a figure explaining a filter process. It is a figure explaining the acquisition procedure of mask measurement data. It is an example of an optical image that is centered on a two-dimensional Gaussian function having a spread of 20 pixels and sampled in units of one pixel. FIG. 6 is an example of a reference image corresponding to the optical image in FIG. 5, in which 0.7 pixels in the X direction and 0.3 pixels in the Y direction are translated and sampled at the same position as in FIG. 5. It is the figure which expanded the center part of FIG. It is the figure which expanded the center part of FIG. It is a figure which shows the relationship between nine sets of ((alpha), (beta)) calculated | required by this Embodiment, and the sum of squared differences. It is a figure which shows the relationship between the group of ((alpha), (beta)) calculated | required in this Embodiment, and the sum of squared differences. It is the figure which connected the point with the same value of a sum of squares with the same line in FIG. It is a figure explaining the structure of the comparison circuit of this Embodiment.

  FIG. 1 is a configuration diagram of an inspection apparatus according to the present embodiment. In this embodiment mode, a mask used in a photolithography method or the like is an inspection target.

  As illustrated in FIG. 1, the inspection apparatus 100 includes an optical image acquisition unit A and a control unit B.

  The optical image acquisition unit A includes a light source 103, an XYθ table 102 that can move in the horizontal direction (X direction, Y direction) and the rotation direction (θ direction), an illumination optical system 170 that constitutes a transmissive illumination system, and magnification optics. A system 104, a photodiode array 105, a sensor circuit 106, a laser length measurement system 122, and an autoloader 130 are included.

  In the control unit B, a control computer 110 that controls the entire inspection apparatus 100 performs a position circuit 107, a comparison circuit 108, a reference circuit 112, a development circuit 111, and an autoloader control unit 113 via a bus 120 serving as a data transmission path. Are connected to a table control circuit 114, a magnetic disk device 109 as an example of a storage device, a magnetic tape device 115, a flexible disk device 116, a CRT 117, a pattern monitor 118, and a printer 119. The XYθ table 102 is driven by an X-axis motor, a Y-axis motor, and a θ-axis motor controlled by the table control circuit 114. As these motors, for example, step motors are used.

  Design pattern data serving as database-based reference data is stored in the magnetic disk device 109, read out as the inspection progresses, and sent to the development circuit 111. In the development circuit 111, the design pattern data is converted into image data (bit pattern data). Thereafter, this image data is sent to the reference circuit 112 and used to generate a reference image to be a standard image.

  In FIG. 1, constituent components necessary for the present embodiment are illustrated, but other known components necessary for inspecting the mask may be included. In this embodiment, a die-to-database inspection method is taken as an example, but a die-to-die inspection method may be used. In this case, one optical image of the same pattern in a different area in the mask is handled as a reference image.

  FIG. 2 is a conceptual diagram showing the flow of data in the present embodiment.

  As shown in FIG. 2, CAD data 401 created by a designer (user) is converted into design intermediate data 402 in a hierarchical format such as OASIS. The design intermediate data 402 stores design pattern data created for each layer and formed on each mask. Here, in general, the inspection apparatus 100 is not configured to directly read OASIS data. That is, unique format data is used for each manufacturer of the inspection apparatus 100. Therefore, the OASIS data is input to the inspection apparatus 100 after being converted into format data 403 unique to each inspection apparatus for each layer. The format data 403 can be data unique to the inspection apparatus 100, but may be data compatible with the drawing apparatus.

  The format data 403 is input to the magnetic disk device 109 of FIG. That is, the design pattern data used when forming the pattern of the photomask 101 is stored in the magnetic disk device 109.

  The figure included in the design pattern is a basic figure of a rectangle or a triangle. The magnetic disk device 109 includes information such as coordinates (x, y) at the reference position of the graphic, side length, graphic code serving as an identifier for distinguishing graphic types such as a rectangle and a triangle, and each pattern graphic. Graphic data defining the shape, size, position, etc.

  Furthermore, a set of figures existing in a range of about several tens of μm is generally called a cluster or a cell, and data is hierarchized using this. In the cluster or cell, arrangement coordinates and repeated description when various figures are arranged alone or repeatedly at a certain interval are also defined. The cluster or cell data is further arranged in a strip-shaped region called a frame or stripe having a width of several hundreds μm and a length of about 100 mm corresponding to the total length of the photomask in the X direction or Y direction. .

  The input design pattern data is read from the magnetic disk device 109 by the development circuit 111 through the control computer 110.

  The expansion circuit 111 expands the design pattern to data for each graphic, and interprets a graphic code, a graphic dimension, and the like indicating the graphic shape of the graphic data. Then, binary or multivalued design image data is developed as a pattern arranged in a grid having a grid with a predetermined quantization size as a unit. The developed design image data calculates the occupancy ratio of the figure in the design pattern for each area (square) corresponding to the sensor pixel. And the figure occupation rate in each pixel becomes a pixel value.

  The design pattern data converted into binary or multi-value image data (design image data) as described above is then sent to the reference circuit 112. In the reference circuit 112, an appropriate filter process is performed on the design image data which is the image data of the transmitted graphic.

  FIG. 3 is a diagram for explaining the filter processing.

  The mask measurement data 404 as an optical image obtained from the sensor circuit 106, which will be described later, is blurred due to the resolution characteristics of the optical system, the aperture effect of the photodiode array, or the like, in other words, a spatial low-pass filter acts. Is in a state. Therefore, it is possible to match the mask measurement data 404 by filtering the bit pattern data that is the image data on the design side, in which the image intensity (light / dark value) is a digital value. In this way, a reference image to be compared with the mask measurement data 404 is created.

  Next, a method for obtaining the mask measurement data 404 will be described with reference to FIGS.

  In FIG. 1, the optical image acquisition unit A acquires an optical image of the mask 101, that is, mask measurement data 404. Here, the mask measurement data 404 is an image of a mask on which a figure based on the figure data included in the design pattern is drawn. A specific method for acquiring the mask measurement data 404 is as follows, for example.

  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, light is emitted from the light source 103 disposed above the XYθ table 102 to the pattern formed on the photomask 101. More specifically, the photomask 101 is irradiated with the light beam emitted from the light source 103 via the illumination optical system 170. Below the photomask 101, an enlargement optical system 104, a photodiode array 105, and a sensor circuit 106 are arranged. The light that has passed through the photomask 101 forms an optical image on the photodiode array 105 via the magnifying optical system 104. Here, the magnifying optical system 104 may be configured such that the focus is automatically adjusted by an automatic focusing mechanism (not shown). Further, although not shown, the inspection apparatus 100 irradiates light from below the photomask 101, guides the reflected light to the second photodiode array via the magnifying optical system, and collects the transmitted light and the reflected light simultaneously. It may be configured.

  FIG. 4 is a diagram for explaining a procedure for acquiring the mask measurement data 404.

  As shown in FIG. 4, the inspection area is virtually divided into a plurality of strip-like inspection stripes 20 having a scan width W in the Y direction, and each of the divided inspection stripes 20 is continuously scanned. Thus, the operation of the XYθ table 102 is controlled, and an optical image is acquired while moving in the X direction. An image having a scan width W as shown in FIG. 4 is continuously input to the photodiode array 105. When an image in the first inspection stripe 20 is acquired, an image having a scan width W is continuously input to the second inspection stripe 20 in the same manner while the XYθ table 102 is moved in the opposite direction. The third inspection stripe 20 is acquired while moving in the direction opposite to the direction in which the image in the second inspection stripe 20 is acquired, that is, in the direction in which the image in the first inspection stripe 20 is acquired. 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 and then A / D (analog-digital) converted by the sensor circuit 106. The photodiode array 105 is provided with an image sensor. An example of the image sensor is a TDI (Time Delay Integration) sensor. For example, the pattern of the photomask 101 is imaged by the TDI sensor while the XYθ table 102 continuously moves in the X-axis direction.

  The XYθ table 102 is driven by a table control circuit 114 under the control of the control computer 110, and can be moved by a drive system such as a three-axis (XY-θ) motor that drives in the X, Y, and θ directions. It has become. For example, a step motor is used as the X-axis motor, the Y-axis motor, and the θ-axis motor. The movement position of the XYθ table 102 is measured by the laser length measurement system 122 and sent 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.

  Mask measurement data 404 (hereinafter also simply referred to as an 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. Sent. The mask measurement data 404 is, for example, 8-bit unsigned data and represents the brightness gradation of each pixel. Further, the reference image described above is also sent to the comparison circuit 108.

  In the comparison circuit 108, before the optical image and the reference image are compared, pattern alignment in these images is performed.

  As described above, the cause of the decrease in the alignment accuracy of the pattern in the conventional method is that the optimal value of the alignment is (1/16) pixel unit. Therefore, in the present embodiment, the parallel movement of the image data in units of pixels smaller than one pixel is performed by linear interpolation. According to this method, it is possible to perform alignment in units of pixels smaller than (1/16) pixel.

For an optical image, the intensity distribution function of the image on the image plane is Φ (x, y). And, although this was sampled at regular intervals by the image sensor coordinates _(x i, _{y j)} the intensity at _{Φ i, j = Φ (x} i, y j) and. Here, the movement for one pixel is expressed as, for example, Φ _{i, j} → Φ _{i−1, j} .

In this embodiment, the coordinates _(x i, _{y j)} and, and coordinates the coordinates was 1 pixel shifted in the X _{(x i + 1, y j} ), the coordinates obtained by one pixel moves in the Y direction _(x i, y _{j + 1} ) and four types of image data at the coordinates (x _{i + 1} , y _{j + 1} ) moved by one pixel each in the X direction and the Y direction can be linearly interpolated to move in units of pixels smaller than one pixel. Make it possible. When this is expressed by an equation, the equation (1) is obtained. However, 0 ≦ α <1 and 0 ≦ β <1.

A value that minimizes the difference between the optical image data and the reference image data represented by Expression (1) is obtained. That is, the intensity distribution function of the reference image is Ψ (x, y), and the intensity at the coordinates (x _i , y _j ) of the sampled image is Ψ _{i, j} = Ψ (x _i , y _j ). What is necessary is just to obtain | require (alpha) and (beta) which satisfy (2).

  Since Equation (2) is a fourth-order polynomial of α and β, the minimum values of α and β can be obtained mathematically. When this polynomial is P (α, β), the following equation (3) is obtained.

Α _{i (m, n)} is obtained by substituting numerical values of nine different combinations for α and β in equation (3). Then, since the polynomial of equation (3) is determined, a value that minimizes the polynomial can be obtained by obtaining a point at which the value obtained by differentiating this equation with α and β becomes 0 at the same time. Specifically, the simultaneous equations shown in Equation (4) are solved.

  In equation (4), each equation is a differential form of a fourth order polynomial, and therefore becomes a third order polynomial. The solution of this equation is obtained using the Newton method or the like.

  FIG. 5 is an example of an optical image that is centered on a two-dimensional Gaussian function (formula (5)) having a spread of 20 pixels and sampled in units of one pixel.

  FIG. 6 is an example of a reference image corresponding to the optical image of FIG. 5, which is sampled at the same position as in FIG. 5 by translating 0.7 pixels in the X direction and 0.3 pixels in the Y direction. is there. The intensity distribution function of the reference image shown in FIG. 6 is expressed by Expression (6).

  7 and 8 are enlarged views of the central portions of FIGS. 5 and 6, respectively. As can be seen from these figures, the image obtained by translating the reference image by 0.7 pixels in the X direction and 0.3 pixels in the Y direction is substantially coincident with the optical image. That is, there is an optimum position for the alignment of each pattern of the reference image and the optical image near the position where the reference image is translated by 0.7 pixels in the X direction and 0.3 pixels in the Y direction.

With respect to the reference image, there are the following nine combinations of α and β when moving the origin (0, 0) from the origin (0, 0) by (1/2) pixels or 1 pixel respectively in the X direction and the Y direction.
(0,0), (0.5,0), (1,0), (0,0.5), (0.5,0.5), (1,0.5), (0,1 ), (0.5,1), (1,1)

  The difference between the optical image and the moved reference image is obtained for the above nine (α, β) by linear interpolation using the equations (5) and (6). Then, the sum of squares of the difference between the two images is as follows.

(Α, β) Sum of squared differences (0,0) 0.91071
(0.5,0) 0.83211
(1,0) 1.53841
(0,0.5) 0.20426
(0.5, 0.5) 0.12599
(1,0.5) 0.83211
(0,1) 0.28271
(0.5, 1) 0.20426
(1,1) 0.91071

  FIG. 9 is a three-dimensional view of the above relationship. In this figure, the sum of the squares is the minimum in the vicinity of (0.7, 0.3). On the other hand, in the nine combinations of (α, β), the sum of squares is the smallest (0.5, 0.5), that is, the reference image is in the X and Y directions, respectively. This is the position translated by 0.5 pixels. Therefore, in order to obtain the true minimum value, it is necessary to perform a calculation that is finer than the calculations for the above nine patterns.

  The sum of squares of the difference between the optical image and the reference image is a fourth-order polynomial of α and β. Therefore, when this is expressed as a vector as P (α, β), Equation (7) is obtained.

  When the sum of squares of the differences is expressed by an equation for the above nine points, the following fourth-order polynomials are obtained.

  When the above equation is expressed by a determinant, the equation (8) is obtained.

In equation (8),
Then, by obtaining an inverse matrix of A, coefficients a ₁ , a ₂ , a ₃ ,..., A ₉ of a fourth-order polynomial are obtained. Therefore, when the above nine points are substituted, A becomes as shown in Equation (9).

  Therefore, the inverse matrix is as shown in Equation (10).

Therefore, the coefficients of the fourth-order polynomial are obtained by Expression (11), and the coefficients a ₁ , a ₂ , a ₃ ,..., A ₉ are as shown in Expression (12).

  As described above, a specific function form of the fourth-order polynomial P (α, β) of α and β, which is the sum of squares of the difference between the optical image and the reference image, was obtained. FIG. 10 is a plot of this function in the range of 0 ≦ α <1 and 0 ≦ β <1. Further, FIG. 11 is a diagram in which points having the same value of the square sum are connected by the same line in order to make the minimum value of the square sum easy to understand. From FIG. 11, it can be seen that the set of (α, β) that minimizes the sum of squares is in the vicinity of (0.7, 0.3). In order to obtain more accurately, it is necessary to obtain partial derivatives of P (α, β) with respect to α and β. Specifically, the coefficients of the partially differentiated polynomial may be obtained. Here, the coefficient vector of the polynomial and the coefficient vector of the derivative are linear mapping and can be expressed by a matrix.

The matrix expressing the derivative with respect to α is

The matrix expressing the derivative with respect to β is as follows.

  Therefore, the partial differentiation of P (α, β) with respect to α and β is as shown in equations (13) and (14).

When the solution of Equation (13) and Equation (14) is obtained by an iterative method,
α = 0.70026
β = 0.30,000
It becomes. If the set that minimizes the sum of squared differences obtained from FIG. 11 is (0.7, 0.3), the difference between these is about 0.02%.

  As described above, in the present embodiment, the movement amounts of the reference image in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively. An equation representing the sum of squared differences is obtained. If the resulting equation is plotted, the relationship between α, β, and the sum of squares of the differences can be understood, so the pattern alignment in these images can be determined from the set of (α, β) that minimizes the sum of squares of the differences. It is possible to know the amount of movement of the reference image that is most suitable for. Since this movement amount can be set to a value smaller than the (1/16) pixel unit in the conventional method, the alignment accuracy can be improved as compared with the conventional method.

  Further, by solving the partial differential with respect to α and β of the equation representing the square sum of the difference between the optical image and the reference image, the values of α and β can be obtained. Therefore, it is possible to accurately know the amount of movement of the reference image that is optimal for pattern alignment in these images.

  In the above description, the reference image is translated in the X and Y directions in order to align the pattern of the optical image and the reference image. However, the present invention is not limited to this, and the reference image is not moved. It is also possible to move the optical image.

  When a set of (α, β) that is optimal for the alignment of the pattern in the optical image and the reference image, that is, the optimal amount of movement of the image in the X direction and the Y direction is obtained, either one of the images according to this value. Move. These processes are performed by the alignment unit 108a of the comparison circuit 108 for each frame (FIG. 12).

  That is, the alignment unit 108a sets α (0 ≦ α <1) and β (0 ≦ β <1) as movement amounts in the X direction and the Y direction for either one of the optical image and the reference image, respectively. Alignment is performed from a set of a part for obtaining an equation representing the relationship between β and the sum of squares of the difference between the optical image and the reference image, and a set of (α, β) that minimizes the sum of squares of the difference obtained from the equation. And a portion for aligning the optical image and the reference image according to the obtained optimum amount of movement.

  In addition, the alignment unit 108a sets α (0 ≦ α <1) and β (0 ≦ β <1) as movement amounts in the X direction and the Y direction for either one of the optical image and the reference image, respectively. And a part for obtaining an equation representing the relationship between β, the sum of squares of the difference between the optical image and the reference image, and a part for obtaining an optimum movement amount for alignment by solving the partial differentiation of α and β for this equation; The optical image may be configured to have a portion for aligning the reference image according to the obtained optimum movement amount.

  In this embodiment, for example, one stripe can be composed of 3000 to 3500 frames, and one frame can be composed of 512 pixels × 512 pixels. In addition, one pixel can be configured with a size of 70 nm × 70 nm.

  The inspection apparatus according to the present embodiment may be configured to store the optimum alignment amount for each frame, that is, the amount of image movement obtained by the alignment unit 108a.

  After the alignment is completed, the defect detection unit 108b performs pattern defect detection processing after comparing these images.

  In the defect detection unit 108b, the mask measurement data 404 sent from the sensor circuit 106 and the reference image generated by the reference circuit 112 are compared using an appropriate comparison determination algorithm. The comparison is performed using only a transmission image, only a reflection image, or an algorithm that combines transmission and reflection. A plurality of algorithms are selected according to the nature of the defect. A threshold value is set for each algorithm, and an algorithm having a reaction value exceeding the threshold value is detected as a defect. In this case, first, a temporary threshold is set for the algorithm, and a defect inspection result performed based on this threshold is reviewed in a review process described later. When this process is repeated and it is determined that sufficient defect detection sensitivity is obtained, the provisional threshold is determined as the algorithm threshold.

  As a result of the comparison, when the difference between the mask measurement data 404 and the reference image exceeds the threshold value, the portion is determined as a defect. If it is determined as a defect, the coordinates, the mask measurement data 404 and the reference image that are the basis for the defect determination are stored in the magnetic disk device 109 as a mask inspection result 405.

  The mask inspection result 405 is sent to a review device 500 that is an external device of the inspection device 100. The review is an operation in which the operator determines whether the detected defect is a problem. The review device 500 displays an image of a defective portion of the mask while moving the table on which the mask is placed so that the defect coordinates of each defect can be observed. At the same time, these are displayed side by side on the screen so that the determination conditions for defect determination and the optical image and reference image that are the basis for determination can be confirmed. By displaying the defect on the mask and the ripple state on the wafer transfer image side by side in the review process, it becomes easy to determine whether or not the mask pattern should be corrected. In general, a reduction projection of about 1/4 is performed from the mask to the wafer, so this scale is also taken into consideration when displaying them side by side.

  All the defects detected by the inspection apparatus 100 are determined by the review apparatus 500. The determined defect information is returned to the inspection apparatus 100 and stored in the magnetic disk device 109. When at least one defect to be corrected is confirmed by the review apparatus 500, the mask is sent to the correction apparatus 600 that is an external apparatus of the inspection apparatus 100 together with the defect information list 406. In the case of a pattern defect, the correction method differs depending on whether the defect type is a convex defect or a concave defect. Therefore, the defect information list 406 includes defect types including irregularities and defect coordinates. For example, it is attached whether the shading film is to be cut or compensated, and cut out pattern data for recognizing the pattern of the portion to be corrected by the correction device. Here, the above-described mask measurement data 404 can be used as the pattern data.

  Note that the inspection apparatus 100 itself may have a review function. In this case, the mask inspection result 405 is displayed on the CRT 117 of the inspection apparatus 100 or a screen of a separately prepared computer together with the accompanying information for defect determination.

  In the review process, the defect is displayed on the monitor based on the data created from the inspection result, and the operator determines whether this is a really problematic defect and classifies the defect. Specifically, a comparison image is generated from an optical image that is measurement data and a reference image, and defects displayed on the comparison image are reviewed by an operator. Pixel data in these images is expressed by gradation values for each pixel. That is, each pixel is given any value from 0 gradation to 255 gradation from a color palette having 256 gradation values, thereby displaying a drawing pattern or a defect.

  As described above, in the present embodiment, the movement amounts of the reference image in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively. An equation representing the sum of squared differences is obtained. If the resulting equation is plotted, the relationship between α, β, and the sum of squares of the differences can be understood, so the pattern alignment in these images can be determined from the set of (α, β) that minimizes the sum of squares of the differences. It is possible to know the amount of movement of the reference image that is most suitable for. Further, by solving the partial differential with respect to α and β of the equation representing the sum of squares of the difference between the optical image and the reference image, the values of α and β can be obtained. Therefore, it is possible to accurately know the amount of movement of the reference image that is optimal for pattern alignment in these images.

  Since the movement amount can be set to a value smaller than the (1/16) pixel unit in the conventional method, the alignment accuracy can be improved as compared with the conventional method. That is, according to the present embodiment, an inspection apparatus and an inspection method capable of detecting defects at high speed while performing alignment between the pattern of the optical image and the pattern of the reference image with high accuracy are provided.

  The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

  In the above embodiments, descriptions of parts that are not directly required for the description of the present invention, such as the device configuration and control method, are omitted. However, the required device configuration and control method may be appropriately selected and used. Needless to say, you can. In addition, all pattern inspection apparatuses or pattern inspection methods that include elements of the present invention and whose design can be changed as appropriate by those skilled in the art are included in the scope of the present invention.

DESCRIPTION OF SYMBOLS 100 Inspection apparatus 101 Photomask 102 XY (theta) table 103 Light source 104 Expansion optical system 105 Photodiode array 106 Sensor circuit 107 Position circuit 108 Comparison circuit 108a Position alignment part 108b Defect detection part 109 Magnetic disk apparatus 110 Control computer 111 Development circuit 112 Reference circuit 113 Autoloader control unit 114 Table control circuit 115 Magnetic tape device 116 Flexible disk device 117 CRT
118 Pattern Monitor 119 Printer 120 Bus 122 Laser Length Measuring System 130 Autoloader 170 Illumination Optical System 401 CAD Data 402 Design Intermediate Data 403 Format Data 404 Mask Measurement Data 405 Mask Inspection Result 406 Defect Information List 500 Review Device 600 Correction Device

Claims (5)

A portion for forming an image of a sample on which a pattern is formed on an image sensor to obtain an optical image of the sample;
A portion for aligning the optical image with a reference image;
Comparing the optical image with the reference image, and determining a defect when the difference exceeds a threshold,
The portion to be aligned is set such that the movement amount in the X direction and the Y direction is α (0 ≦ α <1) and β (0 ≦ β <1) for either one of the optical image and the reference image, respectively. a part for obtaining an equation representing the relationship between α and β, and the sum of squares of the difference between the optical image and the reference image;
An inspection apparatus comprising: a part for obtaining an optimum movement amount for the alignment from a set of (α, β) that minimizes the sum of squares of the differences obtained from the equation. A portion for forming an image of a sample on which a pattern is formed on an image sensor to obtain an optical image of the sample;
A portion for aligning the optical image with a reference image;
Comparing the optical image with the reference image, and determining a defect when the difference exceeds a threshold,
The portion to be aligned is set such that the movement amount in the X direction and the Y direction is α (0 ≦ α <1) and β (0 ≦ β <1) for either one of the optical image and the reference image, respectively. a part for obtaining an equation representing the relationship between α and β, and the sum of squares of the difference between the optical image and the reference image;
An inspection apparatus comprising: a part for obtaining a movement amount optimum for the alignment by solving partial differentials of α and β with respect to the equation. In an inspection method for illuminating a sample on which a pattern is formed, forming an image of the sample on an image sensor via an optical system, and determining the presence or absence of defects,
Obtaining an optical image of the sample from the image sensor;
With respect to any one of the optical image and the reference image serving as the determination reference, the movement amounts in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively, and α and obtaining an equation representing the relationship between β and the sum of squares of the difference between the optical image and the reference image;
Obtaining an optimal amount of movement for alignment between the optical image and the reference image from a set of (α, β) that minimizes the square sum of the differences obtained from the equation;
Aligning the optical image and the reference image according to the optimal amount of movement;
And a step of comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold value. In an inspection method for illuminating a sample on which a pattern is formed, forming an image of the sample on an image sensor via an optical system, and determining the presence or absence of defects,
Obtaining an optical image of the sample from the image sensor;
With respect to any one of the optical image and the reference image serving as the determination reference, the movement amounts in the X direction and the Y direction are α (0 ≦ α <1) and β (0 ≦ β <1), respectively, and α and obtaining an equation representing the relationship between β and the sum of squares of the difference between the optical image and the reference image;
Obtaining an optimal amount of movement for alignment of the optical image and the reference image by solving partial differentials of α and β for the equation;
Aligning the optical image and the reference image according to the optimal amount of movement;
And a step of comparing the optical image with the reference image and determining a defect when the difference exceeds a threshold value.   The reference image is a reference image created from design data of the pattern or an optical image of the same pattern as the pattern in a region different from the optical image on the sample. Or the inspection method of 4.