JP2010268009A - Device and method for inspecting pattern - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device and method for inspecting a pattern, using information of data used for manufacturing an inspection object pattern. <P>SOLUTION: This device for inspecting a pattern is used for inspecting a pattern using an image of an inspection object pattern and data used for manufacturing the inspection object pattern. The device includes: a generation means 11 to generate a reference pattern represented by segments or curves from the data; a generation means to generate a reference pattern used for region inspection from the data; a generation means 7 to generate the image of the inspection object pattern; a means 12 to detect the edge of the image of the inspection object pattern; and an inspection means 12 to inspect the inspection object pattern by comparing the edge of the image of the inspection object pattern with the reference pattern represented by the segments or the curves. The inspection means inspects the inspection object pattern by comparing the edge of the inspection object pattern with the reference pattern used for the region inspection. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a pattern inspection apparatus and method, and more specifically, inspects fine patterns such as semiconductor integrated circuits (LSIs), liquid crystal panels, and their photomasks (reticles) manufactured based on design data. The present invention relates to a pattern inspection apparatus and method.

  Optical pattern inspection equipment using a die-to-die comparison method is used for pattern inspection of wafers in the manufacturing process of semiconductor integrated circuits or pattern inspection of photomasks for pattern formation. . This die-to-die comparison method is a method of detecting defects by comparing images obtained from the same position of a semiconductor device called a die to be inspected and the adjacent die.

  On the other hand, a method called die-to-database comparison is used for inspection of a photomask called a reticle having no adjacent die. This method is a method in which the mask data is converted into an image to replace the image of the proximity die and the same inspection as described above is performed. The mask data is data obtained by applying photomask correction to the design data. This technique is described, for example, in US Pat. No. 5,563,702 “Automated photomask inspection apparatus and method”.

  However, when the die-to-database comparison method is used for wafer inspection, a corner round of a pattern actually formed on the wafer is detected as a defect. As a workaround for this problem, preprocessing is used in which an image converted from mask data is given a corner round with a smoothing filter. However, the corner round by this pretreatment is not equal to the corner round of all patterns formed on the wafer. Therefore, if the allowable pattern deformation amount is set so as to ignore this difference, there is a problem in that minute defects existing in areas other than corners cannot be detected.

  Because the photomask needs to match the mask data as much as possible, the above problem is not fatal in die-to-database comparison photomask inspection. Therefore, die-to-database comparison photomask inspection is now in practical use. However, the pattern formed on the wafer is allowed to be deformed as long as electrical characteristics are guaranteed. This allowable pattern deformation amount is quite large. Actually, the allowable pattern deformation occurs due to a difference in the exposure conditions of the stepper. For these reasons, the above problem is fatal in die-to-database comparison wafer inspection. Therefore, die-to-database comparative wafer inspection has not been put into practical use.

  Focusing on problems in semiconductor integrated circuit production, systematic defects are more important than random defects caused by dust. Systematic defects (defects that occur repeatedly) are defined as defects that occur repeatedly in all dies on the wafer due to a photomask defect or the like. Because systematic defects occur in both the inspected die and the adjacent die to be compared, they cannot be detected by die-to-database comparison. Therefore, there is a need for wafer inspection using a die-to-database comparison method.

  Therefore, although there is a problem in calculation cost and the like, it has not been put into practical use, but an inspection method using design data and a wafer image has been proposed. For example, NEC Technical Report Vol. 50 No. 6/1997 “Automatic failure point tracing method of logic LSI using electron beam tester”. This document describes a method that uses projection of wiring edges to the X and Y axes, a method that focuses on wiring corners, and a method that applies a genetic algorithm. As a method adopted in this document, a matching method is described in which a closed region is extracted after linearly approximating an edge and this closed region is used. However, none of these methods can realize a speed that can be used for high-speed inspection, and furthermore, matching cannot be performed while detecting the amount of pattern deformation.

  At present, automatic defect classification (ADC) using an image including a defect is used. However, since this method cannot recognize which part of the circuit the defect is destroying, it cannot classify the fatal defect and the defect that is not.

  Furthermore, the position of the defect obtained by the die-to-die comparison inspection has an error due to the stage accuracy and optical system accuracy of the inspection apparatus, and the error is about 10 times or more larger than the wiring line width. Due to these errors, even if the defect position is associated with the design data, it is not known where the defect exists on the design data.

  In recent years, the line width of a semiconductor integrated circuit has become much smaller than the wavelength of a light source used in a lithography process. In such a lithography process, a method of adding an OPC (Optical Proximity Correction) pattern is used. This method is a method of manufacturing a photomask using mask data generated by adding an OPC pattern to design data. The pattern on a wafer manufactured using this photomask matches the design data as much as possible. The purpose is to let you. Adding an OPC pattern is one of the most important corrections for a photomask.

  A systematic defect occurs if the OPC pattern does not act effectively as a correction for the pattern formed on the wafer, but this defect cannot be detected by die-to-die comparative wafer inspection. As a solution to this, a method for comparing and inspecting a pattern formed on a wafer and design data in consideration of an allowable pattern deformation amount is required.

  In addition, a short delivery time is required in a high-mix low-volume production process such as a system-on-a-chip (SoC) production process. In this production process, even if a systematic defect is found by an electrical inspection as a final inspection, it may not be possible to meet a short delivery time. As a countermeasure against this, it is necessary to inspect the difference between the pattern formed on the wafer and the design data for each lithography process. In this inspection method, it is necessary to set a pattern deformation that does not affect the electrical characteristics as an allowable pattern deformation amount and detect a deformation that exceeds the allowable pattern deformation amount.

  At present, as an evaluation of the OPC pattern, design data and the OPC pattern are checked by a lithography simulator. Although the lithography simulator can verify the entire device, it cannot always simulate a defect that actually occurs. Further, the lithography simulator cannot detect defects caused by problems other than the OPC pattern. Such defects include random defects present in the photomask and stepper distortion.

  Furthermore, in order to verify the validity of the simulation, a means for comparing and examining the simulation pattern output from the lithography simulator and the pattern image actually formed on the wafer is required. In addition, it is becoming increasingly important to improve circuit design techniques by strictly setting pattern deformation amounts for design data.

  Currently, a CD-SEM (Critical Dimension Scanning Electron Microscope) is used for wafer line width management in the manufacturing process of a semiconductor integrated circuit. The CD-SEM automatically measures the line width of a linear pattern at a position specified using a line profile. In order to manage the exposure conditions of the stepper using the CD-SEM, the length is measured at several locations in several shots on several wafers for each lot.

  In addition to the line width, shrinkage of the wiring end and the position of the isolated pattern are also important for the management of the circuit pattern. However, the CD-SEM automatic length measurement function can measure only the length such as the line width in one dimension. Therefore, these two-dimensional shapes are manually inspected by an operator using images obtained from a CD-SEM or other microscopes.

  The OPC pattern plays an important role not only to secure the gate line width but also to form corners and isolated patterns. Furthermore, with the improvement of the operating frequency, in addition to the gate line width, it is now important to manage the shape of the end of the gate pattern and the base called the end cap or field extension.

  Such a two-dimensional pattern inspection is important both in the sampling inspection in the manufacturing process and in the trial production stage. In particular, in the trial production stage, the inspection of all patterns formed on the wafer is required. However, the management of the two-dimensional shape is currently performed by human work, and is not completely implemented. To solve this problem, automated die-to-database comparison wafer inspection is required.

The following issues are raised as specific requests for automation.
(1) In order to obtain a defect that repeatedly occurs in each semiconductor device, it is practically difficult to check whether there is a defect at the same location by comparing a large amount of defect information.

(2) Automation of line width inspection, average line width inspection, space width inspection, and average space width inspection of a linear pattern for the purpose of process management of semiconductor devices is required. Moreover, although complicated calculation is required, the line width inspection, the average line width inspection, the space width inspection, and the average space width inspection are also required for the corner portion that is a curved shape pattern.

  In particular, automatic inspection of the gate line width of all semiconductor devices is important for performance improvement, but in order to identify the gate part, the overlapping part of the polysilicon layer and the active layer (the previous process of the polysilicon layer) is extracted. It must be very labor intensive.

In addition, in order to obtain information that a cut or short circuit that is important as a defect type is obtained, an operator needs to reexamine the image. In addition, it is necessary to enhance the recognition of cut or short-circuited defects observed with a thin contrast.
As another requirement, it is necessary to shorten the image acquisition time by using a method in which only the vicinity of the region to be inspected is scanned.

(3) An inspection method for managing the contact area between the end of the wiring layer and the contact hole / via hole is required. In this inspection method, it is necessary to automatically determine the allowable pattern deformation amount as termination shrinkage management by determining whether there is a sufficient margin between the termination and the contact hole / via hole. In addition, an inspection method using the contact area as an evaluation value is also required.

(4) As a kind of OPC pattern, it is added for the purpose of correcting a pattern existing in the vicinity of the pattern, but there is a pattern that is not formed on the wafer. Such a pattern may be formed and cause a defect. Such a defect inspection method is required.

(5) There is a need for a method of displaying an enormous amount of inspection results after converting them into information suitable for the evaluation of the pattern deformation amount of the entire semiconductor device and the distortion of the stepper.

(6) The pattern image to be inspected may be rotated due to the rotation of the sample by moving the stage. In addition, deformation such as rotation including skew and change in magnification may occur due to a charging phenomenon. Due to these influences, it is not possible to detect fine defects below the above-mentioned strain amount. This distortion occurs discontinuously in time and is difficult to predict. As a countermeasure against this phenomenon, a method for detecting and correcting the distortion amount of the pattern image to be inspected every time an image is acquired is necessary.

(7) There is a need for a method for correcting image distortion of an image generation apparatus having a wide field of view. Image distortion includes nonlinear image distortion and line width variation depending on the image position. It is desired that this method can be automatically performed in a short time with high accuracy.

(8) Since the defect information is enormous, there is a method for reducing the number of detected defects by separating the pattern deformation amount into a global pattern deformation amount and a local pattern deformation amount and detecting important defects reliably. Desired.

(9) In the long-time inspection, the beam diameter may change gradually with time. As the beam diameter increases, the measured line width increases by the increased amount. It is necessary to correct the change in the measured value of the line width due to the variation in the beam diameter.

(10) Classification of enormous defects using defect types determined by geometric information of reference patterns, design data information, or data information related to design data in order to easily recognize the tendency of defects to occur There is a need to.

(11) There is a problem that when the number of defects of one type of defect is very large and the number of defects of other defect types is small, the latter image cannot be registered.
(12) There is a problem that defects having the same defect type that are detected a little cannot be sufficiently reinspected compared to defects having the same defect type that is detected a lot.

(13) It is necessary to easily recognize the tendency of the defect to occur by separating the defect detected in the thin and dense part and the defect detected in the sparse part.

(14) A line segment in which the distance between the closest line segments is smaller than a specified value or a line segment larger than a specified value due to a phenomenon caused by a change in the generation rate of secondary charged particles or a capture rate. Etc. require signal strength correction. There is a need for a method for correcting this signal intensity.

(15) The region suitable for image adjustment has been determined by the operator's visual observation, but the image is based on the geometric information of the line segments forming the design data or the relationship between the line segments forming the design data in contact with or close to each other. There is a need for automation using a method that automatically extracts a region suitable for adjustment. Further, when a part of an image is used for image adjustment, automatic adjustment is performed with high accuracy. However, there is a problem that there is only a method of using the entire image conventionally.

(16) Most of the design data is a horizontal line and a vertical line. Using this property, it is necessary to speed up matching using the projection data in the horizontal and vertical directions of the edges obtained from the design data and the projection data in the horizontal and vertical directions of the edges detected from the pattern image to be inspected. .

(17) In simple matching, it is difficult to accurately match a boundary between a portion where the same pattern is periodically arranged and a portion where it is not. There is a need to solve this problem.

(18) The matching of the hole pattern and the island pattern requires more calculation time due to the matching because there are more polygons smaller than the linear pattern. In order to solve this problem, a high-speed calculation method is required.

(19) In the hole pattern and the island pattern, the brightness distribution of the image may be uneven depending on the location due to the influence of the charging phenomenon or the like. A matching method that is resistant to this variation is required.

(20) When the inspection unit area is divided into a plurality of sub-inspection unit areas, a matching method using the sub-inspection unit area most suitable for matching is required for the purpose of speeding up.

(21) When the previous process pattern exists in the lower layer of the inspection target pattern, the portion of the inspection target pattern in which the previous process pattern exists in the lower layer and the inspection target pattern in which the previous process pattern does not exist in the lower layer In the portion, the formation of the inspection target pattern and the observed shape are different. As a countermeasure, an inspection method that uses different inspection parameters for the inspection target pattern affected by the previous process and the inspection target pattern not affected by the previous process is required.

(22) Corresponding to the design data corresponding to the inspection result, the design data to which the correction pattern is added, the shape obtained by the simulation using the design data, or another information related to the design data If the displayed information and the defect shape or defect image as the inspection result are displayed in parallel or overwritten, it becomes easier to understand the tendency of defects to occur. A display method that meets this requirement is needed.

(23) In the lithography process, the pattern formed on the resist film on the silicon substrate is inspected. In this case, when the pattern formed on the resist film is inspected using an electron beam (charged particle beam), the resist is generally an insulator made of a high molecular compound, so that the object to be inspected by the charging phenomenon. The pattern shape of the pattern image is deformed. This is because the electron beam is bent by the upper surface of the partially charged resist film, so that the electron beam is not irradiated to an accurate position of the resist. Therefore, it is necessary to prevent the occurrence of charging phenomenon and obtain a pattern image without deformation.

(24) A method of performing automatic measurement using a scanning electron microscope while keeping all scanning directions of an electron beam (charged particle beam) constant. However, in this method, a measurement error occurs depending on the direction of the line segment. Therefore, a method capable of automatically setting various measurement conditions based on the reference data is required.

  In view of these problems, an object of the present invention is to provide a pattern inspection apparatus and method using information of data used for manufacturing a pattern to be inspected.

  In order to achieve the above object, a first aspect of the pattern inspection apparatus of the present invention is a pattern inspection apparatus that inspects using an inspection target pattern image and data used for manufacturing the inspection target pattern, Generating means for generating a reference pattern expressed by a line segment or a curve from the data; generating means for generating a reference pattern used for region inspection from the data; generating means for generating the inspection target pattern image; Means for detecting an edge of the inspection target pattern image, and inspection means for inspecting the inspection target pattern by comparing the edge of the inspection target pattern image with a reference pattern expressed by the line segment or curve. The inspection unit compares the edge of the pattern image to be inspected with a reference pattern used for the region inspection. By, it is characterized in that to inspect the inspection object pattern.

One aspect of the present invention is the above-described pattern inspection apparatus, wherein the reference pattern used for the region inspection uses the geometric information of the line segment forming the data or the relationship between the line segments forming the data in contact with or close to each other. And extracting.
One aspect of the present invention is the above-described pattern inspection apparatus, wherein the inspection using the reference pattern used for the region inspection includes line width inspection of a linear shape pattern, space width inspection, average line width inspection, average space width inspection, curve shape It is at least one of pattern line width inspection, space width inspection, average line width inspection, average space width inspection, or cutting / short circuit inspection.
According to an aspect of the present invention, in the pattern inspection apparatus, the means for extracting a reference pattern used for the region inspection from the data includes the polygon of the data of the layer related to the inspection target and the data of the layer related thereto. It is characterized by using the result of a logical operation with a polygon.
One aspect of the present invention is characterized in that, in the pattern inspection apparatus, a defect that is cut or short-circuited is detected by detecting the edge in a direction different from an edge of a reference pattern used for the region inspection. To do.

  A second aspect of the pattern inspection apparatus of the present invention is a pattern inspection apparatus that inspects using a pattern image to be inspected and data used to manufacture the pattern to be inspected. Generating means for generating the expressed reference pattern; generating means for generating the inspection target pattern image; means for detecting an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve The inspection unit for inspecting the pattern to be inspected by comparing the reference pattern expressed by the above, and a plurality of semiconductor devices manufactured based on the same data by the inspection unit. The defect position and defect size obtained from the same inspection area expressed by the coordinates of When the region having a size overlap is characterized in further comprising a recognizing defect recognizing means and defects recurring.

  According to one aspect of the present invention, in the pattern inspection apparatus, at least one of the defect positions and defect sizes obtained from the plurality of semiconductor devices is set as a first semiconductor device. The defect position and the defect size are obtained from the inspected area, and for the other second semiconductor device, it is examined whether the defect position obtained by the first semiconductor device and the area having the defect size overlap. Only the area necessary for the inspection is inspected.

  A third aspect of the pattern inspection apparatus of the present invention is a pattern inspection apparatus that inspects using an inspection target pattern image and data used for manufacturing the inspection target pattern, and includes a line segment or a curve from the data. Generating means for generating the expressed reference pattern; generating means for generating the inspection target pattern image; means for detecting an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve And an inspection means for inspecting the inspection target pattern by comparing with the reference pattern expressed by the above, the inspection means being obtained from the information obtained from the geometric information of the reference pattern and the data information A defect type is set for each item obtained from the item or data information related to the data. It is intended.

One aspect of the present invention is characterized in that, in the pattern inspection apparatus, the defect classification is performed using a feature obtained by cutting out the reference pattern in the vicinity of the defect.
One aspect of the present invention is the pattern inspection apparatus, wherein the data information is used to add the correction pattern, the shape data obtained by the simulation using the data, or the data. Any one of the different data is associated with the defect.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspecting means for inspecting the inspection target pattern by comparing with a reference pattern, and defect information obtained from the same inspection area for a plurality of semiconductor devices manufactured based on the same design data It is also possible to recognize a defect that repeatedly occurs from the obtained defect information.

  Among the defect information obtained from the plurality of semiconductor devices, the defect information is obtained from the entire inspection region for at least one of the semiconductor devices, and the other semiconductor devices are obtained from the whole. The defect information may be obtained by inspecting only the vicinity of the defect information.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection means for inspecting the pattern to be inspected by comparing with a reference pattern, and the reference pattern suitable for inspection using a plurality of the edges is contacted with geometric information of line segments constituting the design data or Alternatively, it may be extracted using the relationship between the line segments forming the design data adjacent to each other.

As the inspection using the plurality of edges, line width inspection of a linear shape pattern, space width inspection, average line width inspection, average space width inspection, curve shape pattern line width inspection, space width inspection average line width inspection, average space It may be at least one of width inspection or cutting / short circuit inspection.
The means for extracting the reference pattern from the design data may use a result of a logical operation between the polygon of the design data of the layer related to the inspection object and the polygon of the design data of the layer related thereto. Good.

Moreover, you may test | inspect whether it is the defect cut | disconnected or short-circuited by detecting the said edge in the direction different from the said edge.
Further, a scanning microscope for obtaining the inspection target pattern image may be provided, and the scanning microscope may scan only a portion corresponding to the reference pattern suitable for inspection using the plurality of edges and the vicinity thereof.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspection means for inspecting the inspection target pattern by comparing with a reference pattern, and the inspection parameter or inspection evaluation value is the polygon of the design data of the layer related to the inspection target and the design data of the layer related thereto May be obtained using the result of a logical operation with the polygon.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data Generating means for generating a pattern; generating means for generating from a correction pattern added to the design data; generating means for generating the inspection target pattern image; means for detecting an edge of the inspection target pattern image; An inspection unit that inspects the inspection target pattern by comparing an edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve, and is formed on the sample on which the inspection target pattern is formed. When inspecting the correction pattern that should not be performed, the reference pattern generated from the correction pattern It may be associated with an edge of said object pattern image and.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern may be provided, a statistic may be obtained for each of the divided inspection areas, and the statistic may be displayed as a distribution diagram.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspection means for inspecting the inspection target pattern by comparing with a reference pattern, and detecting the distortion amount of the inspection target pattern image immediately after obtaining the inspection target pattern image, thereby detecting the reference pattern and the inspection target At least one of the pattern images may be corrected.

A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and the generation unit that generates the inspection target pattern image obtains an image distortion amount in advance, and uses the image distortion amount The inspection target pattern image may be corrected.
The image distortion amount may be a nonlinear image distortion amount or a line width variation amount depending on the image position.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit for inspecting the inspection target pattern may be provided by comparing with a reference pattern, and the pattern deformation amount may be separated into a global pattern deformation amount and a local pattern deformation amount.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and the generation unit that generates the inspection target pattern image determines a divided inspection region to be inspected a plurality of times, and After inspecting the divided inspection areas to be inspected multiple times and obtaining the inspection results, the multiple inspections are performed at an appropriate time during the inspection. To obtain the test results again divided examination region to be the may be corrected temporal variation of the inspected pattern image using the pre-test results and test the obtained results again obtained.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection means for inspecting the pattern to be inspected by comparing with a reference pattern, and the defect type includes geometric information of the reference pattern, information on the design data, or information on data related to the design data. You may decide to use.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, has a maximum number of registered defect images for each defect type, and when a new defect image is detected, the new defect New defect images are registered until the maximum registration number of the defect type to which the image belongs is exceeded, and a new one is registered when the maximum registration number is exceeded. Serial defect image may be judged whether or not to register the.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection means for inspecting the pattern to be inspected by comparing with a reference pattern, and having a maximum number of registered defects to be reinspected for each defect type, and using the maximum number of registrations, the reinspection object The defect may be selected.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit for inspecting the inspection target pattern by comparing with a reference pattern may be provided, and the detected defects may be grouped based on the characteristics of the reference pattern in the vicinity of the defects.

A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, extracts a line segment that requires signal intensity correction from the reference pattern, and according to the amount of the signal intensity correction The position of the line segment may be corrected, or an allowable pattern deformation amount may be set.
The line segment that requires signal intensity correction may be a line segment in which the distance between the closest line segments among the opposing line segments is smaller than a specified value or larger than a specified value.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspecting means for inspecting the pattern to be inspected by comparing with a reference pattern, and using the geometric information of the line segment forming the design data or the relationship between the line segments forming the design data in contact with or close to the design data Then, an area suitable for the inspection target pattern image adjustment may be extracted.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit for inspecting the inspection target pattern by comparing with a reference pattern, and matching may be performed using the projection data of the edge of the reference pattern and the projection data of the edge of the inspection target pattern image .

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and the inspection unit recognizes the reference pattern existing at the boundary of the reference pattern having periodicity as a unique pattern, The unique pattern is shifted by one cycle, and the reference pattern is placed near the unique pattern shifted by one cycle. The unique pattern of the shifted one cycle when the over down no confirmed negative pattern may be matched by using the negative pattern and the unique pattern.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspection means for inspecting the inspection target pattern by comparing with a reference pattern, and the inspection means includes geometric information obtained from the reference pattern and a geometry obtained from an edge of the inspection target pattern image. Matching may be performed using academic information.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and the inspection unit calculates an image statistic corresponding to the inside of the reference pattern and an image statistic corresponding to the outside. By comparing, the pattern image to be inspected and the reference pattern may be matched.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve Inspection means for inspecting the inspection target pattern by comparing with a reference pattern, and the inspection means divides the inspection target pattern image to obtain the reference pattern corresponding to the divided image Select the reference pattern that is most suitable for matching and correspond to the selected reference pattern. May be carried matching using said segmented image that.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and the inspection unit is not affected by the part of the inspection target pattern affected by the previous process and the previous process. Different inspection parameters may be used for the target pattern portion.

  A preferred embodiment of the present invention is a pattern inspection apparatus for inspecting using an inspection target pattern image and design data used for manufacturing the inspection target pattern, and a reference expressed by a line segment or a curve from the design data A generating unit that generates a pattern; a generating unit that generates the inspection target pattern image; a unit that detects an edge of the inspection target pattern image; and an edge of the inspection target pattern image and the line segment or curve An inspection unit that inspects the inspection target pattern by comparing with a reference pattern, and has, as the inspection result, information on the design data corresponding to an edge of the reference pattern, and the information on the design data Corresponding to the inspection result, the design data to which the correction pattern is added, and the simulation using the design data Shape obtained in Shon, or may be associated with any of the other data relating to the design data.

  A preferred embodiment of the present invention is a pattern to be inspected by using an image of an inspection target pattern formed on a sample that is easily charged and design data used for manufacturing the inspection target pattern formed on the sample that is easily charged. An inspection method, wherein a reference pattern expressed by a line segment or a curve is generated from the design data, carbon is coated on the inspection object pattern formed on the easily charged sample, and the coated charging An image of the inspection object pattern formed on the coated easily charged sample is obtained by scanning a charged particle beam on the inspection object pattern formed on the easy to carry sample, and the coated easily charged object Edges of the image of the pattern to be inspected formed on the sample are detected, and the coated and easily charged The inspection object pattern formed on the coated easily charged sample by comparing the edge of the image of the inspection object pattern formed on the material with the reference pattern expressed by the line segment or curve May be inspected.

  A preferred embodiment of the present invention is a pattern to be inspected by using an image of an inspection target pattern formed on a sample that is easily charged and design data used for manufacturing the inspection target pattern formed on the sample that is easily charged. An inspection method, wherein a reference pattern expressed by a line segment or a curve is generated from the design data, and a charged particle beam is acquired at a time from the pattern to be inspected formed on the easily charged sample. Scanning over a wide area to obtain an image of the inspection object pattern formed on the easily charged sample, detecting an edge of the image of the inspection object pattern formed on the easily charged sample, and charging the By comparing the edge of the image of the pattern to be inspected formed on the easy sample with the reference pattern expressed by the line segment or curve, Inspecting the pattern to be inspected formed on a sample that is easily charged, performing the inspection on a plurality of semiconductor devices manufactured based on the same design data, obtaining a plurality of inspection results, The test results may be obtained by fusing the test results.

  A preferred aspect of the present invention is a pattern inspection method for inspecting using an inspection object pattern image and design data used for manufacturing the inspection object pattern, and a reference expressed by a line segment or a curve from the design data By generating a pattern, generating the inspection target pattern image, detecting an edge of the inspection target pattern image, and comparing the edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve Inspecting the pattern to be inspected, obtaining defect information obtained from the same inspection region for a plurality of semiconductor devices manufactured based on the same design data, and repeatedly generating from the obtained defect information Defects may be recognized.

  A preferred aspect of the present invention is a pattern inspection method for inspecting using an inspection object pattern image and design data used for manufacturing the inspection object pattern, and a reference expressed by a line segment or a curve from the design data By generating a pattern, generating the inspection target pattern image, detecting an edge of the inspection target pattern image, and comparing the edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve Inspecting the inspection object pattern, the reference pattern suitable for inspection using a plurality of the edges, the geometric information of the line segment forming the design data, or the relationship between the line segments forming the design data in contact with or close to each other May be used to extract.

  A preferred aspect of the present invention is a pattern inspection method for inspecting using an inspection object pattern image and design data used for manufacturing the inspection object pattern, and a reference expressed by a line segment or a curve from the design data By generating a pattern, generating the inspection target pattern image, detecting an edge of the inspection target pattern image, and comparing the edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve The inspection target pattern may be inspected, and the defect type may be determined using geometric information of the reference pattern, information on the design data, or information on data related to the design data.

  A preferred aspect of the present invention is a pattern inspection method for inspecting using an inspection object pattern image and design data used for manufacturing the inspection object pattern, and a reference expressed by a line segment or a curve from the design data By generating a pattern, generating the inspection target pattern image, detecting an edge of the inspection target pattern image, and comparing the edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve The inspection target pattern may be inspected, and the detected defects may be grouped based on the characteristics of the reference pattern in the vicinity of the defects.

  A preferred aspect of the present invention is a pattern inspection method for inspecting using an inspection object pattern image and design data used for manufacturing the inspection object pattern, and a reference expressed by a line segment or a curve from the design data By generating a pattern, generating the inspection target pattern image, detecting an edge of the inspection target pattern image, and comparing the edge of the inspection target pattern image with the reference pattern expressed by the line segment or curve The inspection target pattern is inspected, and the inspection result has the design data information corresponding to the edge of the reference pattern, and the correction pattern corresponding to the inspection result is added from the design data information. Design data, shapes obtained by simulation using the design data, or the design data With any of the other data to be communicated may be associated.

According to the present invention, the following effects can be obtained.
(1) It is possible to recognize automatically and repeatedly generated defects using defect information obtained from a plurality of semiconductor devices. As a result, a large amount of simple labor by the operator is unnecessary, and it becomes possible to prevent a reduction in defect recognition due to operator error. In addition, when a sample is contaminated by a carbon coat or the like to be described later, the contaminant is hardly present at the same location of different dies, so that the contaminant is not recognized as a defect that repeatedly occurs.

(2) Area inspection can be realized. The area inspection method means an inspection method using opposite edges. Specifically, line width inspection, average line width inspection, space width inspection, and average space width inspection of a linear pattern can be automatically performed. Further, the line width inspection, space width inspection, average line width inspection, average space width inspection, and gate line width inspection of the corner portion which is a curved shape pattern can be automatically performed.

  These not only save labor for inspections for the purpose of the above process management, but also enable a wide range of inspections impossible with operator inspections. In particular, the automatic inspection of the gate line width of all semiconductor devices can greatly contribute to improving the performance of semiconductor devices. Since these area inspections use an average value of a plurality of defect information, the defect detection capability and the defect recognition accuracy are greatly improved as compared with the inspection method using a single edge.

  Further, it is possible to realize an inspection method for a portion that is easily cut or short-circuited. According to this method, a cut or short-circuited defect observed with a thin contrast can be recognized. In addition, it is possible to set a defect type having information that it has been cut or short-circuited.

  As another effect, the image acquisition time can be shortened by using a method in which only the vicinity of the region to be subjected to region inspection is scanned. In addition, since the scanning direction and the edge direction can be orthogonal to each other, the edge detection accuracy can be improved.

(3) A reference pattern suitable for gate line width inspection and end cap inspection can be obtained by using a logical product operation of the polysilicon layer and the active layer (pre-process of the polysilicon layer). By this method, the gate portion can be automatically extracted. In addition, since an allowable pattern deformation amount different from a simple end can be automatically set in the end cap of the gate, it becomes possible to strictly inspect the end cap of the gate.
Furthermore, a margin is obtained by using a logical product operation result of the polygon of the wiring layer design data and the polygon of the contact hole / via hole design data, and the connection to the contact hole / via hole is adapted to the obtained margin. It is possible to set the allowable pattern deformation amount at the end used in the above. Moreover, the contact area as an evaluation value can also be obtained by using the AND operation of these polygons.

(4) As a kind of OPC pattern, it is added for the purpose of correcting a pattern existing in the vicinity of the pattern, but there is a pattern that is not formed on the wafer. However, such a pattern may be formed and become a defect. Such a correction pattern inspection method that should not be formed on the wafer can be realized by applying edge detection.
(5) By obtaining a statistic from the inspection result for each divided inspection region and displaying the statistic as a distribution diagram, the pattern deformation amount of the entire semiconductor device can be visually grasped. Thereby, the distortion of the stepper can be recognized.

(6) It is possible to detect and correct a linear amount of distortion of the inspection target pattern image. As a result, the amount of distortion that does not need to be recognized as a defect can be ignored, and the occurrence of a pseudo defect can be prevented.
(7) Image distortion correction of an image generating apparatus having a wide field of view can be automatically performed in a short time with high accuracy. In addition, it is possible to correct line width variation depending on the image position. Therefore, it is possible to extend the field of view to a portion where this correction is possible.

(8) The number of acquired defects can be reduced by separating the pattern deformation amount into a global pattern deformation amount and a local pattern deformation amount. As a result, it becomes possible to sufficiently detect important defects and reduce detection of pseudo defects.

(9) When a pattern made of ArF resist is inspected many times by a scanning electron microscope, the pattern gradually shrinks. However, according to this embodiment, since the same place is inspected only twice, this pattern shrinkage can be ignored. Therefore, even when the pattern is inspected, it is possible to correct the variation in the measured value of the line width due to the gradual variation in the beam diameter.
(10) By classifying the defect using the defect type using the geometric information of the reference pattern, the information of the design data, or the information of the data related to the design data, the tendency to generate the defect becomes easy. Also, the cause of the defect can be easily identified.

(11) Even when there are a large number of defects of one type of defect and few defects of other types of defects, more types of images can be registered.
(12) A defect having the same defect type detected in a small number and a defect having the same defect type detected in a large number can be sufficiently reinspected.

(13) Grouping can be realized based on the characteristics of the reference pattern near the defect. As a result, it is possible to easily grasp the tendency of defects such as “there are many defects in a pattern that is complicated by thin vertical lines”. Further, defects can be classified for each reference pattern having the same shape. Furthermore, it becomes easy to identify the cause of the defect.

(14) Due to a phenomenon such as the occurrence rate of secondary charged particles and the capture rate, the distance between the closest line segments among the opposing line segments is smaller than the specified value or larger than the specified value. Need signal strength correction. For this signal intensity correction, the effect of these phenomena can be reduced by correcting the position of the line segment of the reference pattern or setting the allowable pattern deformation amount.

(15) The region suitable for image adjustment is automatically optimized by extracting the region based on the geometric information of the line segment constituting the design data or the relationship between the line segments constituting the contact data or the adjacent design data. Can be extracted. When this area is a separated area, automatic adjustment is performed with higher accuracy than when the entire image is used.

(16) Most of the design data is a horizontal line and a vertical line. Using this property, matching is possible using projection data in the horizontal and vertical directions of the edges obtained from the design data and projection data in the horizontal and vertical directions of the edges detected from the inspection target image. If this method is used, instead of shifting for each pixel, the calculation time can be greatly shortened because the shift can be performed at intervals of skipping.

(17) Since the negative pattern imposes a large penalty on the deviation of one cycle from the optimal matching position, it is possible to accurately match the boundary between the portion where the same pattern is periodically arranged and the portion where it is not. Here, the negative pattern is recognized as follows.
A reference pattern existing at the boundary of a periodic reference pattern is recognized as a unique pattern, the unique pattern is shifted by one period, and the reference pattern is shifted by one period when there is no reference pattern in the vicinity of the unique pattern shifted by one period. Recognize unique patterns as negative patterns.

(18) The matching of the hole pattern and the island pattern requires more calculation time due to the matching because there are more polygons smaller than the linear pattern. In order to solve this problem, it is possible to realize a matching method using information obtained by collecting a plurality of edges. This method can be executed faster than the method of matching using individual edges. Furthermore, the amount of calculation can be greatly reduced.
(19) By using the difference histogram between the inside and outside of the hole pattern and island pattern as the evaluation value, even if the brightness distribution of the background image fluctuates depending on the location due to the influence of the charging phenomenon or the like, it is not easily affected. A matching method can be realized.

(20) When the inspection unit region is divided into a plurality of sub-inspection unit regions, the sub-inspection unit region most suitable for matching can be obtained. This makes it possible to execute faster than the matching using the entire inspection unit area.

(21) When the previous process pattern exists in the lower layer of the inspection target pattern, the portion of the inspection target pattern in which the previous process pattern exists in the lower layer and the inspection target pattern in which the previous process pattern does not exist in the lower layer In the portion, the formation of the inspection target pattern and the observed shape are different. As a countermeasure, an inspection method that uses different inspection parameters for the inspection target pattern affected by the previous process and the inspection target pattern not affected by the previous process is realized. This can reduce the probability of detecting a pseudo defect.

(22) Corresponding to the design data corresponding to the inspection result, the design data to which the correction pattern is added, the shape obtained by the simulation using the design data, or another information related to the design data If the displayed information and the defect shape or defect image as the inspection result are displayed in parallel or overwritten, it becomes easier to understand the tendency of defects to occur. As a result, it is possible to easily identify the cause of the occurrence of the defect, so that the design change is facilitated.

(23) In the lithography process, before the pattern formed on the resist film on the silicon substrate is inspected using an electron beam (charged particle beam), a carbon film is coated on the pattern formed on the resist film. By coating the carbon film in this way, when the electron beam is irradiated, the electron beam flows to the silicon substrate through the carbon film. As a result, since the silicon substrate flows to the ground, the charging phenomenon can be prevented and a pattern image without deformation can be obtained.

(24) In a lithography process, when a pattern formed on a resist film on a silicon substrate is inspected using an electron beam, it can be inspected without performing a special process. In this inspection method, only the central portion of the uniformly charged image is inspected.

(25) In order to obtain an image of the inspection target pattern, it is only necessary to scan a minimum number of electron beams (charged particle beams). Therefore, an image of the inspection target pattern can be obtained in a minimum time.
(26) A wide range of blocks is realized with the smallest possible number of blocks by making the best use of the scannable area. Furthermore, an image can be acquired under optimum conditions using a reference pattern in order to prevent a decrease in edge detection accuracy depending on the scanning direction.

(27) The difference between the image quality in the X direction and the Y direction can be reduced as much as possible by a method of taking data between scanning lines by changing the scanning path, a method of scanning twice, a method of applying a filter, or the like.

(28) The profile deformation due to the charging phenomenon of the sample can be reduced, and the edge detection accuracy is improved. In addition, it is possible to acquire an image at high speed by scanning only the edge portion that is most important for measurement, instead of performing raster scan of the whole to acquire data.
(29) A rotated image without image quality degradation due to interpolation can be acquired, and degradation in edge detection accuracy can be avoided.

It is the schematic which shows the basic composition of the inspection apparatus of this invention. It is a schematic diagram showing the intensity | strength of the secondary electron detected with the secondary electron detector shown in FIG. FIG. 3 is a schematic diagram when the pattern shown in FIG. 2 is rotated 90 degrees and a profile of this pattern is acquired. It is a schematic diagram which shows the scanning area in the case of performing a pattern inspection by the pattern inspection apparatus of this invention. It is a figure for demonstrating the test | inspection precision at the time of performing the scanning of a horizontal direction. It is a figure for demonstrating the test | inspection precision at the time of scanning in the vertical direction toward the upper direction from the bottom. It is a schematic diagram in the case of performing bidirectional scanning. It is a figure which shows typically the case where the scanning direction of an electron beam is 45 degree | times and -45 degree | times. It is a figure which shows the example of the reference | standard pattern obtained from design data. It is a figure which shows the example of the test object pattern image manufactured based on design data. It is a figure which shows the outline | summary of the inspection process which the pattern inspection apparatus which concerns on embodiment of this invention performs. It is a figure which shows the example of the line segment which should be test | inspected using a 0 degree image or a 90 degree image. FIG. 4 shows a method for obtaining a rotated image only by replacing the pixel positions. FIG. 6 shows another method for obtaining a rotated image only by replacing pixel positions. It is a figure which shows the basic structural example of the pattern inspection apparatus in embodiment of this invention. It is a figure which shows the functional block diagram of the pattern inspection apparatus in embodiment of this invention. It is a figure which shows the other example of the functional block diagram of the pattern inspection apparatus in embodiment of this invention. It is a figure which shows the example of a correction | amendment of a reference pattern. It is a flowchart which shows the example of the recipe registration process in embodiment of this invention. It is a figure for demonstrating a sequential inspection. It is a figure for demonstrating a random test | inspection. It is a figure which shows the example of a reference | standard pattern. It is a figure which shows the example which converted the reference | standard pattern of FIG. 22 into the edge for every pixel. It is a figure which shows the example which converted the reference | standard pattern containing a curve into the edge vector. It is a flowchart which shows the example of the basic test | inspection process in embodiment of this invention. It is a subblock of the flowchart which shows the example of the inspection process in the case of recognizing the defect which generate | occur | produces repeatedly. It is the main block of the flowchart which shows the example of the inspection process in the case of recognizing the defect which generate | occur | produces repeatedly. It is a figure which shows the example of the image (inspection object pattern image) which has the contrast in the inside of a pattern, and the foundation | substrate. It is a figure which shows the edge detected from the image of FIG. It is a figure which shows the example of an image (inspection target pattern image) with a bright outline only. It is a figure which shows the edge detected from the image of FIG. It is a figure which shows the example of the edge strength of a one-dimensional inspection object pattern image. It is a figure which shows the example which expanded the edge of FIG. It is a figure which shows the example of the intensity | strength of the edge of a one-dimensional reference | standard pattern. It is a figure which shows another example which expanded the edge of FIG. It is a figure which shows another example of the intensity | strength of the edge of a one-dimensional reference pattern. It is a figure which shows another example which expanded the edge of FIG. It is a figure which shows the example of a smoothing filter. It is a figure which shows the example of the intensity | strength of the edge of a two-dimensional test object pattern image. It is a figure which shows the example which expanded the edge of FIG. It is a figure which shows another example which expanded the edge of FIG. It is a figure which shows the example of the edge vector of a two-dimensional test object pattern image. It is a figure which shows the example which expanded the edge vector of FIG. It is a figure which shows another example which expanded the edge vector of FIG. It is the figure which represented the reference | standard pattern of FIG. 23 with the edge vector of a pixel unit. It is a figure for demonstrating matching. FIG. 46 is a diagram in which FIG. 43 and FIG. 45 are superimposed. It is another figure which piled up Drawing 43 and Drawing 45. (A) shows an example of a reference pattern, and (b) shows an example of a pattern image to be inspected. It is a figure which shows the example in case line | wire width and space width are the same. (A) shows an example of a reference pattern, and (b) is a diagram showing an example of the relationship between the reference pattern of (a) and an inspection target pattern image. It is a figure which shows typically the calculation method of the matching evaluation value of the pattern in which the rectangle was located in a line. It is a figure which shows typically the calculation method of a matching evaluation value. It is a figure which shows the matching method using the projection data to the horizontal-vertical axis of the edge detected by the 1st edge detection. It is a figure which shows the calculation result of matching error value _Epm . It is a figure which shows the shift amount suitable for the matching chosen from the matching error value _Epm . FIG. 6 is a diagram schematically showing a method for calculating a matching error value _Epm . It is a schematic diagram explaining the 1st method of matching of a hole pattern. It is a schematic diagram explaining the 2nd method of matching of a hole pattern. It is a figure which shows the example of matching with the edge of a test target pattern image, and the edge of a reference | standard pattern. (A) shows an example of an edge of a reference pattern, and (b) shows an example of an edge of an inspection target pattern image. It is another figure which shows the example of the edge of a test object pattern image. It is a figure which shows typically the method of recognizing the abnormal pattern deformation amount defect. It is a figure which shows typically the recognition method of the defect which uses the luminance distribution of a pixel. It is a figure which shows the example of distribution of the frequency with respect to a luminance value. (A) shows an example of an edge of a reference pattern and an edge of an inspection target pattern image, and (b) shows an X component of a vector d (x, y ₀ ) at y = y ₀ between the edges shown in (a). It is a figure which shows the example approximated by the regression line D (x). (A) shows another example of the edge of the reference pattern and the edge of the pattern image to be inspected, and (b) shows the X of the vector d (x, y ₀ ) at y = y ₀ between the edges shown in (a). It is a figure which shows the example which approximated the component with the regression line D (x). It is a figure which shows the example of the attribute of a pattern. It is a figure which shows the positional offset amount of a termination | terminus. It is a figure which shows the positional offset amount of the gravity center of an isolated pattern. (A) shows an example of an edge of a corner of a reference pattern, and (b) shows an example of a corner of an inspection target pattern image. It is a figure which shows the example of a profile acquisition area. It is a figure which shows the curve which forms the external shape of the exposure pattern obtained with the lithography simulator. FIG. 73 is an enlarged view of a part of FIG. 72 (part B). FIG. 75 is an enlarged view of a part (part C) of FIG. 74. It is a figure which shows the example of a profile. It is a figure which shows the example which performed curve approximation (a polygon approximation is included) using the detected 2nd edge, and connected the detected 2nd edge. (A) shows another example of the profile acquisition section, and (b) is a diagram showing an example of the relationship between the first edge and the second reference pattern of the inspection target pattern image. It is a figure which shows typically the case where a test | inspection area | region is divided | segmented into four test | inspection unit area | regions. It is a figure which shows typically the defect information obtained from the 1st semiconductor device, and the defect information obtained from the 2nd semiconductor device. It is a figure which shows typically the defect information obtained from the defect information obtained from the 1st semiconductor device, and the limited part of the 2nd semiconductor device. It is a figure which shows typically the rule which extracts automatically the reference pattern suitable for line width inspection from design data. It is a figure which shows typically the method of isolate | separating the linear shape pattern which has a corner into two rectangles in a corner part. It is a figure which shows typically the rule which extracts automatically the reference pattern suitable for space width inspection from design data. It is a figure which shows typically the test | inspection method which uses the reference pattern suitable for a line width test | inspection, and the reference pattern suitable for a space width test | inspection. It is a figure which shows typically the procedure of obtaining the reference pattern suitable for the line | wire width test | inspection of a corner part. It is a figure which shows typically the procedure of the minimum line | wire width test | inspection of the curve shape pattern which is a corner part. It is a figure which shows typically the procedure of the minimum line | wire width test | inspection of the curve shape pattern which is a corner part which uses Erosion calculation. It is a figure which shows typically the extraction method of the part which is easy to cut | disconnect or short-circuit. It is a figure which shows typically the procedure which performs the test | inspection of the part which is easy to cut | disconnect or short-circuit. It is a figure which shows typically the inspection method using the reference | standard pattern obtained by the logical product operation of the layer design data regarding the test object, and the layer design data regarding the process before and after that. It is a figure which shows typically the adaptive setting method of the allowable pattern deformation amount with respect to the termination | terminus of the wiring layer which touches a contact hole / via hole. It is a figure which shows typically the method of calculating | requiring a contact area. (A) is a figure which shows the example of the correction pattern which must not be formed in a wafer, (b) is a figure which shows typically the inspection method of the correction pattern which must not be formed in a wafer. It is a figure which shows typically the method of extracting a near line segment from a reference | standard pattern. It is a figure which shows typically the method of extracting a separation line segment from a reference | standard pattern. It is a schematic diagram showing an example in which a pattern is formed with a width that is globally different from the line width of design data due to a difference in pattern formation conditions. It is a figure which shows the example of the 1st method of calculating | requiring the global average line width variation | change_quantity of line | wire width using the test | inspection unit area | region which has been test | inspected. FIG. 99 is a diagram showing an example of a second method in which the global average line width deformation obtained by the first method shown in FIG. 98 is used for correcting the line width of design data. It is a figure which shows the example of the method of calculating the deformation | transformation amount of the line | wire width of a 30 degree direction. FIG. 21 is a diagram schematically showing fluctuations in beam diameter in a simplified form of FIG. 20. It is a figure which shows typically the method of determining the test | inspection unit area | region test | inspected twice. It is a figure which shows typically the method of test | inspecting a test | inspection unit area | region twice. It is a figure which shows typically the subdivision item of the defect kind classified for every line segment in the case of periodic patterns, such as a memory. It is a figure which shows typically the division | segmentation item of the defect kind used in combination. It is a figure which shows typically a defect position, a circumscribed rectangle, and the cut-out reference pattern. It is a schematic diagram which shows the example of feature-value space. FIG. 108 is a schematic diagram illustrating another example of feature amounts used in the feature amount space of FIG. 107. It is an example of a distribution chart obtained by converting a line width deformation amount, which is one of pattern deformation amounts obtained from the entire inspection unit region, into information for grayscale display and overwriting defects. It is a figure which shows typically the case where the scanning direction of an electron beam is 18 degree | times. It is a figure which shows typically the scanning method scan of a hexagonal block. It is a schematic diagram which shows the automatic setting method of the scanning conditions based on a reference pattern. It is a schematic diagram explaining the scanning path | route of an electron beam. It is another schematic diagram explaining the scanning path | route of an electron beam. It is a schematic diagram explaining the case where a filter is applied to scanning in the vertical direction. It is a figure which shows typically the case where only the vicinity of an edge is scanned. It is a flowchart which shows the step in the case of scanning only the vicinity of an edge. It is a figure which shows the acquisition data ordering method in the case of scanning only the vicinity of an edge. It is a figure which shows typically the method of obtaining the vicinity part of the area | region used as the object of area | region inspection. It is a figure which shows typically the image of the test | inspection unit area | region which has distortion. It is a figure which shows typically the method of implementing matching in a sub test | inspection unit area | region. It is a figure which shows typically the correction method which correct | amends the distortion amount of a test target pattern image. It is a figure which shows typically the example of nonlinear image distortion. It is a figure which shows typically the structural example of a deflection | deviation control apparatus. It is a figure which shows typically the method by which a distortion correction vector calculation circuit calculates a distortion correction vector using a representative distortion vector. (A) is a figure which shows typically the method of calculating a representative distortion vector from the distortion vector of a rectangular area. (B) is a figure which shows the example of a synthetic | combination distortion vector typically. It is a figure which shows typically the method of calculating a representative distortion vector from the distortion vector of a some rectangular area. It is a figure which shows typically the method by which a distortion correction vector calculation circuit converts a distortion correction vector into a deflection voltage. It is a figure which shows typically the method of correct | amending the line width distribution of the test target pattern image at the time of an inspection using the line width distribution of the test target pattern image obtained beforehand. It is a figure which shows typically the procedure for obtaining the area | region suitable for automatic contrast and brightness adjustment, and automatic focus adjustment. It is a figure which shows the example of the area | region suitable for automatic astigmatism adjustment. It is a figure which shows typically the procedure for obtaining the area | region suitable for automatic astigmatism adjustment. It is a figure which shows two sub test | inspection unit area | regions. It is a figure which shows typically the selection method of the sub test | inspection unit area | region most suitable for matching. It is a figure which shows the example which performs length measurement using a high magnification image and a low magnification image. It is a figure which shows typically the example by which the pattern of a previous process is observed as a foundation | substrate. It is a figure which shows typically the example which overwrite-displays design data, mask data, and a defect image. It is a figure which shows typically the example of the method of displaying the recognized defect as a figure. It is a figure which shows typically the method of displaying after converting a defect into design data. It is a figure which shows typically the lithography process of a semiconductor device. It is a figure which shows typically the state in which the carbon film is coated on the resist film and the part where the resist film is removed by development and the oxide film is exposed. It is the schematic which shows the semiconductor wafer inspection system which provided the pattern inspection apparatus of this invention and the carbon sputter coating apparatus integrally. (A) shows an actual pattern image when the sample has no carbon coating, and (b) shows an actual pattern image when the same sample as shown in (a) is coated with a carbon film. is there. It is a figure which shows typically the inspection method which test | inspects only the center part of an image.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

Content
1. Overview
2. Hardware Configuration 2.1 Basic Configuration of Image Generating Device 2.2 Scanning Method of Image Generating Device 2.2.1 Scanning Method 1
2.2.2 Scanning method 2
2.2.3 Scanning method 3
2.3 Basic configuration of pattern inspection system 2.4 Functional block diagram
3. Explanation of terms 3.1 Edge 3.2 Reference pattern 3.3 Recipe data 3.4 Inspection unit area 3.5 Inspection result
4). Basic Inspection Process 4.1 First Edge Detection 4.1.1 First Edge Detection Method 1
4.1.2 First edge detection method 2
4.2 Matching method of linear pattern 4.2.1 Matching method using unique pattern 4.2.2 Matching method using negative pattern 4.2.3 Matching using projection data of edge to vertical axis and horizontal axis Method 4.3 Hole Pattern / Island Pattern Matching Method Using Geometric Information 4.4 Hole Pattern / Island Pattern Matching Method Using Statistics 4.5 Process After Matching 4.6 First Check 4.6. 1 Abnormal Pattern Deformation Defect Recognition Method 4.6.2 Defect Recognition Method Using Pixel Luminance Distribution 4.7 Defect Type Using Feature Quantity Obtained from Image 4.8 Pattern Deformation Obtained from Entire Inspection Unit Area Amount 4.9 Defect recognition method detected in pattern attribute units 4.9.1 Misalignment defect at the end 4.9.2 Straight line, no corner position Defect 4.9.3 Misalignment defect of isolated pattern 4.9.4 Other isolated pattern defect 4.9.5 Abnormal curvature defect of corner 4.10 Pattern attribute extraction rule 4.11 Second edge detection 4 .12 Second inspection
5. Applied Inspection Processing 5.1 Repetitive Defect Recognition Method 5.2 Area Inspection Method 5.2.1 Line Pattern Inspection Method, Average Line Width Inspection Method, Space Width Inspection Method, and Average Space Width Inspection Method 5.2.2 Curve width pattern line width, average line width, space width, and average space width inspection method 5.2.3 Inspection method of a portion that is easily cut or short-circuited 5.3 Design data of the layer related to the inspection object Inspection Method that Uses Logical Operation of Polygon and Polygon of Layer Design Data Related to it 5.3.1 Gate Line Width Inspection Method 5.3.2 End Cap Inspection Method 5.3.3 Contact Hole / Adaptive pattern deformation setting method for the end of wiring layer in contact with via hole 5.3.4 Contact area inspection method 5.4 Correction that must not be formed on wafer Turn inspection method 5.5 Pattern inspection method requiring signal intensity correction 5.6 Separation of pattern deformation amount into global pattern deformation amount and local pattern deformation amount 5.6.1 Time of line width measurement value 5.7 Defect Type Using Geometric Information of Reference Pattern, Design Data Information, or Data Information Related to Design Data 5.8 Grouping Using Reference Patterns Near Defects 9. Method of selecting defect to be registered as image target 5.10 Method of selecting defect as target of reinspection 5.11 Method of displaying distribution diagram of pattern deformation amount of entire semiconductor device
6). Other Scanning Method of Image Generation Device 6.1 Scanning Method of Electron Beam in 18 Degree Direction, Scanning Method of Hexagonal Block, Scanning Condition Automatic Setting Method Based on Reference Pattern 6.2 Scanning of Electron Beam in Image Generation Device Path 6.3: Scanning method only in the vicinity of the edge 6.4 Scanning method only in the vicinity of the region to be inspected
7). Method of correcting inspection target pattern image 7.1 Correction method of at least one of reference pattern and image based on detection of distortion amount of inspection target pattern image immediately after image acquisition 7.2 Nonlinear image distortion correction method 7.3 Position of pattern image Correction method for line width fluctuations
8). Other methods 8.1 Extraction method of region suitable for image adjustment 8.2 Selection method of sub-inspection unit region most suitable for matching 8.3 Inspection method using high-magnification image and low-magnification image 8.4 8.5 Inspection method of inspection target pattern having pattern influence 8.5 Overwriting display method of defect information and information corresponding to the defect
9. Inspection method for easily charged samples 9.1 Carbon coating method for resist samples 9.2 Inspection method for inspecting only the central part of an image
10. Variant of the invention

1. Outline The pattern inspection apparatus according to this embodiment inspects an inspection target pattern image obtained by the image generation apparatus 7 shown in FIG. 1 in comparison with a reference pattern.

  FIG. 9 is a diagram illustrating an example of a reference pattern obtained from design data. FIG. 10 is a diagram illustrating an example of an inspection target pattern image manufactured based on design data. As shown in FIG. 10, the pattern image to be inspected has a short circuit defect, a particle defect, or a deformation within an allowable deformation amount. There is a big corner round especially in the corner. Thus, the inspection target pattern image is considerably different from the reference pattern.

FIG. 11 is a diagram showing an outline of inspection processing performed by the pattern inspection apparatus according to the present embodiment. In the inspection process, first, a first edge is detected from the inspection target pattern image. Next, the inspection target pattern image and the reference pattern are matched by comparing the detected first edge with the edge of the first reference pattern. As a result of the matching, the shift amount S ₁ is obtained, and the first reference pattern is shifted using this shift amount S ₁ . Next, the pattern to be inspected is inspected by comparing the detected first edge with the edge of the shifted first reference pattern. In the first inspection, the detected first edge and the edge of the first reference pattern are compared to obtain the pattern deformation amount, and the defect is detected from the obtained pattern deformation amount. Shift amount S ₂ is obtained as one of the pattern deformation quantity.

Next, in order to detect the second edge from the inspection target pattern image, the corresponding second reference pattern is shifted by the shift amount S ₁ + S ₂ . Using the shifted second reference pattern, a profile is obtained on the inspection target pattern image, and the second edge is detected. Then, the inspection target pattern is inspected by comparing the detected second edge with the edge of the shifted second reference pattern. Also in the second inspection, the pattern deformation amount is obtained by comparing the detected first edge with the edge of the first reference pattern, and a defect is detected from the obtained pattern deformation amount. Shift amount S ₃ is obtained as one of the pattern deformation quantity.

  With the above method, it is possible to detect short-circuit defects, particle defects, and pattern deformation amounts from the pattern image to be inspected, obtain defect types from the attributes of design data, etc., and classify the defects and pattern deformation amounts into classes. .

2. Hardware configuration

2.1 Basic Configuration of Image Generating Device FIG. 1 is a schematic diagram showing the basic configuration of the image generating device in the pattern inspection apparatus of the present invention. As shown in FIG. 1, the image generation apparatus 7 in the pattern inspection apparatus of the present invention includes an irradiation system device 310, a sample chamber 320, and a secondary electron detector 330.

The irradiation system device 310 includes an electron gun 311, a focusing lens 312 that focuses primary electrons emitted from the electron gun 311, an X deflector 313 and a Y deflector 314 that deflect an electron beam (charged particle beam), And an objective lens 315. The sample chamber 320 includes an XY stage 321. A wafer W, which is a sample, is carried into and out of the sample chamber 320 by a wafer transfer device 340.
In the irradiation system 310, the primary electrons emitted from the electron gun 311 are focused by the focusing lens 312 and then focused by the objective lens 315 while being deflected by the X deflector 313 and the Y deflector 314. The surface of the wafer W is irradiated.

  When primary electrons are irradiated onto the wafer W, secondary electrons are emitted from the wafer W, and the secondary electrons are detected by the secondary electron detector 330. The focusing lens 312 and the objective lens 315 are connected to a lens control device 316, and this lens control device 316 is connected to a control computer 350. The secondary electron detector 330 is connected to an image acquisition device 317, and this image acquisition device 317 is also connected to the control computer 350. The X deflector 313 and the Y deflector 314 are connected to a deflection control device 318, and this deflection control device 318 is also connected to the control computer 350. The XY stage 321 is connected to an XY stage controller 322, and this XY stage controller 322 is connected to a control computer 350. Similarly, the wafer transfer device 340 is connected to the control computer 350. The control computer 350 is connected to the operation computer 360.

2.2 Scanning Method of Image Generating Device FIG. 2 is a schematic diagram showing the intensity of secondary electrons detected by the secondary electron detector 330 shown in FIG. FIG. 2 shows the intensity of the secondary electrons obtained by the secondary electron detector 330 when one electron beam is scanned in the X direction, and the edge portion of the pattern P is strong due to the edge effect. The strength of the central portion of the pattern P is weak. In addition, the left and right sides of the pattern P are not symmetrical, and the edge on the entrance side of the electron beam (left edge in the figure) is observed to have a weaker signal amount than the opposite edge (right edge in the figure). The

  FIG. 3 is a schematic diagram when the pattern P shown in FIG. 2 is rotated 90 degrees and a profile of the pattern P is acquired. FIG. 3 illustrates the intensity of secondary electrons by scanning a plurality of electron beams in the X direction. As shown in FIG. 3, it is difficult to clearly obtain the edge effect in the edge portion parallel to the scanning direction as compared with FIG.

  FIG. 4 is a schematic diagram showing a scanning area when pattern inspection is performed by the pattern inspection apparatus of the present invention. In FIG. 4, a portion written with a solid line indicates a pattern P to be inspected. A square block written with an alternate long and short dash line indicates a range (scanning area) of an area acquired by one scan. In this example, this block is composed of a total of nine blocks B1 to B9, three vertically and three horizontally. The portion written with a dotted line is an observation area OA.

FIG. 5 is a diagram for explaining inspection accuracy when scanning in the horizontal direction (X direction) is performed. As shown in FIG. 5, when scanning in the horizontal direction, the inspection accuracy for the vertical lines is good as in FIG. 2, but good inspection accuracy cannot be obtained for the horizontal lines.
FIG. 6 is a diagram for explaining inspection accuracy when scanning in the vertical direction (Y direction) is performed from the bottom to the top. As shown in FIG. 6, when the vertical scanning is performed, the inspection accuracy with respect to the horizontal line is good, but the good inspection accuracy cannot be obtained with respect to the vertical line.

  In the lower left block B7 in FIG. 4 having the vertical and horizontal patterns, if it is attempted to obtain good inspection accuracy for both the horizontal and vertical lines, the horizontal scanning shown in FIG. 5 and the vertical scanning shown in FIG. 6 are performed. Good inspection accuracy cannot be obtained unless scanning is performed twice. In the block B8 having only the horizontal line shown on the right side, only the vertical scanning shown in FIG. Further, in the block B4 having only the vertical line pattern shown in the leftmost middle row, only the horizontal scanning shown in FIG. 5 needs to be performed. In this way, a desired image is obtained by controlling scanning by a method in which scanning is performed in the horizontal direction and the vertical direction, respectively, or scanning is performed twice in the horizontal direction and the vertical direction.

  When the scanning direction is 0 degree (X direction), the edge detection accuracy for a pattern whose pattern extends in the X direction (lateral direction) is weak, and when the scanning direction is 90 degrees (Y direction), the Y direction ( The edge detection accuracy of the pattern extending in the vertical direction is weak. Therefore, in order to obtain good edge detection accuracy, it is necessary to perform scanning in two directions of 0 degrees and 90 degrees in the scanning direction. Most patterns of semiconductor integrated circuits (LSIs) and liquid crystal panels to be detected are composed of a pattern extending in the horizontal direction (X direction) and a pattern extending in the vertical direction (Y direction). In order to detect the image with high accuracy, it is necessary to perform scanning in two directions of the X direction (0 degree) and the Y direction (90 degrees).

  FIG. 7 is a schematic diagram of a method for performing bidirectional scanning. As described with reference to FIGS. 2 to 6, it has been described that the inspection accuracy is poor because the brightness due to the edge effect cannot be obtained unless the scanning direction and the edge are orthogonal. Further, the edge on the scanning line entry side (left edge in FIG. 2) is less accurate than the opposite edge (right edge in FIG. 2). Therefore, as shown in FIG. 7, the image is acquired by alternately reversing the scanning direction. That is, an image is acquired by performing alternate scanning in the left direction and the right direction. By measuring the edge on the entrance side in the scanning direction in the image scanned in the left direction and measuring the edge on the opposite side in the image scanned in the right direction, good inspection accuracy can be obtained at any edge.

  FIG. 8A, FIG. 8B, and FIG. 8C are schematic diagrams showing the case where the scanning direction is 45 degrees and −45 degrees. In such a scanning direction of 0 degrees and 90 degrees, it was necessary to perform scanning twice almost certainly. However, in this method, the pattern P1 of only the horizontal and vertical lines in FIG. The inspection accuracy of the vertical and horizontal lines can be ensured by performing the scanning at 45 degrees shown in FIG. 8B or −45 degrees shown in FIG. 8C once.

  If there is a 45 degree line P2 in FIG. 8A, it is necessary to perform two scans of 45 degree and -45 degree, but this frequency is higher than the pattern P1 of only the vertical and horizontal lines. It is expected that there is a relatively large number of cases where only one scan is required. Therefore, scanning at 45 degrees or −45 degrees in the scanning direction is effective as a method for obtaining inspection accuracy by one scanning.

  Next, a case where scanning at 45 degrees and −45 degrees in the scanning direction is performed will be described. In FIG. 8B, the accuracy is obtained by scanning in the 45 degree direction for the downward-sloping pattern such as the pattern P2, but the accuracy is obtained for the upward-sloping pattern because the scanning direction and the edge to be detected are in parallel. I can't. This pattern is inspected using an image obtained by scanning at −45 degrees shown in FIG. It is expected that the frequency of scanning at 45 degrees and −45 degrees in the scanning direction is less than the frequency of scanning at 0 degrees and 90 degrees in the scanning direction.

As described with reference to FIGS. 2 to 8, the image generation device 7 obtains an image of the inspection target pattern by one of the following three methods.
2.2.1 Scanning method 1
Scan in one direction at 0 degree, 90 degree, 45 degree or -45 degree
2.2.2 Scanning method 2
Alternate scan of 0 degree and 180 degree
2.2.3 Scanning method 3
0 degree and 90 degree bi-directional scan or 45 degree and -45 degree bi-directional scan

  Here, the coordinate system is a coordinate system in which the X axis is in the right direction, the Y direction is in the upward direction, and the most frequent direction of the pattern to be inspected is in the right direction (0 degree direction). The edge direction is defined as a direction in which the right hand side is inside the pattern. The block B4 in FIG. 4 has two edges that run up and down, but the direction of the left edge is 90 degrees and the direction of the right edge is 270 degrees.

As described later in 4.1 first edge detection , the first edge is an edge detected from a local image. The direction of the first edge is determined when it is detected. Hereinafter, a method for detecting an edge from the image of the inspection target pattern obtained by the scanning method 1 to the scanning method 3 will be described.

  In the scanning in one direction as the scanning method 1 and the alternating scanning as the scanning method 2, an edge is detected from one image. In the two-direction scanning which is the scanning method 3, edges are detected from two images and the edge information is fused. In the case of scanning in two directions of 0 degree and 90 degrees, only edges between 45 degrees to 135 degrees and 225 degrees to -45 degrees are extracted from the 0 degree image, and 135 degrees to 225 are extracted from the 90 degree image. And only edges between −45 degrees and 45 degrees are extracted, and both are combined and treated as edges detected from one image.

  In the case of scanning in two directions of 45 degrees and −45 degrees, only edges between 90 degrees to 180 degrees and 270 degrees to 360 degrees are extracted from the 45 degrees image, and 0 degrees to 90 degrees are extracted from the −45 degrees image. Only the edges between 180 degrees and 180 degrees to 270 degrees are extracted, and both are combined and treated as edges detected from one image.

  As will be described later in the second edge detection, the second edge is an edge detected from a profile (one-dimensional data). The direction of the second edge is determined when the profile is set. Hereinafter, a method for detecting an edge from the profile obtained by the scanning method 1 to the scanning method 3 will be described.

In the scanning in one direction as the scanning method 1, the profile is obtained from the same image.
In the scanning method 2 of 0 degree and 180 degree alternating scanning, the right side edge (180 degree to 360 degree edge) is obtained from the 0 degree image, and the left side edge (0 degree to 180 degree edge) is obtained. A profile to be detected is obtained from an image of 180 degrees.

  In the bi-directional scanning as the scanning method 3, a profile for detecting an edge between 45 degrees to 135 degrees and 225 degrees to -45 degrees is obtained from an image of 0 degrees, 135 degrees to 225 degrees, and -45 degrees. A profile for detecting an edge between 45 degrees and 45 degrees is obtained from an image of 90 degrees.

  FIG. 12 shows an example of a line segment to be inspected using a 0 degree image or a 90 degree image. As shown in FIG. 12, straight lines, corners, and end points in the vertical direction (90 ° or 270 ° direction), left upward (135 °), and right downward (−45 °) direction from the image obtained by scanning in the 0 ° direction. Inspect the line segment. Further, it is only necessary to inspect straight line portions, corners, and end line segments in the horizontal direction (0 degree or 180 degrees), the upward direction to the right (45 degrees), and the downward direction to the left (225 degrees) from the image obtained by scanning in the 90 degree direction.

  In the case of scanning in two directions of 45 degrees and -45 degrees, a profile for detecting an edge between 90 degrees to 180 degrees and 270 degrees to 360 degrees is obtained from an image of 45 degrees, and from 0 degrees to 90 degrees and 180 degrees. A profile for detecting an edge between 270 degrees is obtained from an image of -45 degrees.

  When an image is acquired with a 45 ° or −45 ° inclination, there is a rotation between the reference pattern and the image, and therefore an operation for correcting the rotation is necessary. One method is to rotate the reference pattern. However, when the reference pattern is rotated, the tilted image becomes the final output, which is difficult to see. Therefore, in this embodiment, a method of rotating the image is adopted. However, when scanning is performed to sample evenly in the X and Y directions, if the image is rotated, the interpolated value between pixels must be used as the value of the rotated image. In this case, since the obtained image has a detrimental effect of being blurred by interpolation, this embodiment employs a method of obtaining a rotated image only by replacing pixel positions without using interpolation. When this method is used, it is necessary to employ a special scanning method as described below.

  FIG. 13 is a diagram showing a specific example of an image created based on a method of obtaining a rotated image only by replacing pixel positions. The 45 degree tilt scanning method on the left side and the 45 degree tilt image on the right side are exactly the same and are drawn by rotating 45 degrees. The image to be finally acquired has the right shape. In the figure, the grid points of the grid are the positions of the image to be obtained by sampling evenly in the X and Y directions. The black circles (●) are actually sampled data, and there are no black circles that cannot be obtained by this scanning method. In order to acquire the image of the right side, the left side scanning method is used.

  In this case, the sampling interval S in the X direction is the same for each scanning line, but the sampling interval in the Y direction is half of the sampling interval S in the X direction. Further, the odd and even lines are shifted by half the sampling interval S in the X direction. This sampling interval S is obtained by multiplying the right pixel interval by √2. In this case, a desired image can be obtained simply by laying the left figure sideways. In this case, it is necessary to enter values in an order different from the actually sampled order.

FIG. 13 shows a scanning method with an angle of 45 degrees, and FIG. 14 shows a scanning method with an angle of arctan (2) and a rotated image.
If this embodiment is used, in order to obtain an image of the inspection target pattern, it is only necessary to scan a minimum number of electron beams (charged particle beams), and therefore, an image of the inspection target pattern can be obtained in a minimum time. In addition, the method of scanning the missing data portion of the raster scan twice can reduce the slight difference in image quality between the X direction and the Y direction as much as possible. Furthermore, it is possible to obtain a rotated image without image quality deterioration due to interpolation, and avoid a decrease in edge detection accuracy.

2.3 Basic Configuration of Pattern Inspection Apparatus FIG. 15 is a diagram showing the basic configuration of the pattern inspection apparatus in the present embodiment. The pattern inspection apparatus according to this embodiment includes a main control unit 1, a storage device 2, an input / output control unit 3, an input device 4, a display device 5, a printing device 6, and an image generation device 7 shown in FIG.

  The main control unit 1 is constituted by a CPU (Central Processing Unit) or the like, and controls the entire apparatus in an integrated manner. A storage device 2 is connected to the main control unit 1. The storage device 2 can take the form of a hard disk, flexible disk, optical disk, or the like. Further, the input / output control unit 3 prints an input device 4 such as a keyboard and a mouse, a display device 5 such as a display that displays input data, calculation results, and the like, and calculation results and the like through the input / output control unit 3 A printing device 6 such as a printer is connected.

  The main control unit 1 has an internal memory (internal storage device) for storing a control program such as an OS (Operating System), a program for pattern inspection, and necessary data. Realized. These programs can be stored in a flexible disk, an optical disk or the like, and read into a memory, a hard disk or the like before being executed.

2.4 Functional Block Diagram FIG . 16 is a functional block diagram of the pattern inspection apparatus in the present embodiment. The reference pattern generation unit 11, the inspection unit 12, the output unit 13, and the defect type recognition unit 14 are realized by a program. The basic database 21, the recipe database 22, and the defect type reference database 23 are provided in the storage device 2.
The basic database 21 may be provided outside, and the pattern inspection apparatus may access the basic database 21 via a LAN (Local Area Network).

  FIG. 17 is a diagram showing another example of a functional block diagram of the pattern inspection apparatus in the present embodiment. The example illustrated in FIG. 17 is a diagram illustrating a configuration having a function of recognizing a defect that repeatedly occurs. The function block diagram of FIG. 16 includes a defect information storage unit 24, a defect recognition unit 25 that repeatedly generates, Has been added.

3. Explanation of terms
3.1 Edge Edge means the boundary between the inside of the pattern to be inspected and the ground. As edges, as shown in FIG. 60, the edge of the inspection target image and the edge of the reference pattern are used. The edge of the inspection target image is detected by an edge detection method, and the edge of the reference pattern is obtained by dividing a straight line or a curve into pixels.

  An edge is represented by a vector having start point (subpixel accuracy), direction, and intensity information for each pixel. In the case of the edge of the image to be inspected, the intensity is a value obtained by multiplying the length of the vector and the clarity of the edge. In the case of the edge of the reference pattern, the intensity is a value obtained by multiplying the length of the vector and the degree of contribution to matching.

3.2 Reference Pattern A reference pattern is expressed by a line segment or a curve, and is compared with an image to be inspected. Design data is used as the most suitable for the reference pattern. As this design data, for example, layout data in a GDSII (Graphic Design System II) data stream format, which is a combination of layers and fracturing, can be used.

  First, a shrink process (a process for changing the magnification), a size process (a process for changing the line width), and the like are performed on the design data so as to be most suitable for the position of the edge detected from the inspection target pattern image. In addition, since the edge positions to be detected are generally different between the first edge detection and the second edge detection, two types of reference patterns are prepared for the first edge detection and the second edge detection.

  Next, the polygon obtained by this processing is clipped with a rectangular area having one side of a length obtained by adding a stage error and the maximum deformation amount of the inspection target pattern to the visual field.

  The resulting polygon corner is then rounded. As shown in FIG. 18, the design data is usually a polygon having an acute angle (dotted line in the figure), while the circuit pattern actually formed has rounded corners (solid line in the figure). Therefore, a circle, an ellipse, a straight line, or a curve described by another method is applied to the corner portion, and correction is performed so as to be close to an actual pattern.

  Finally, the result obtained above is used as a reference pattern and stored in the recipe database 22. If the stage error is negligible compared to the maximum deformation amount of the inspection target pattern, the absolute coordinate value of the pattern deformation can be measured. In this embodiment, in consideration of the stage error and the maximum deformation amount of the inspection target pattern, the reference pattern is processed larger than the inspection target pattern image, but instead, the inspection target pattern image is more than the reference pattern. You may make it process large.

  If the design data is used as the reference pattern, a defect inspection in which the pattern formed on the wafer is compared with the design data can be executed. In this case, an allowable amount that does not affect the electrical characteristics is set as the allowable pattern deformation amount. The permissible pattern deformation amount can be set for each wiring attribute, and can be varied between a place where the pattern is complicated and a place where the pattern is not.

If a curve (the solid line in FIG. 73) forming the outer shape of the exposure pattern obtained by the lithography simulator is used as the reference pattern, an inspection for verifying the validity of the simulation can be executed. The output data of the lithography simulator is a light intensity distribution obtained by optical simulation. A contour curve is obtained from this distribution. In this case, an allowable pattern deformation amount sets an error allowed as a simulation.
In the present embodiment, a method of using design data for a reference pattern will be described.

FIG. 22 is a diagram illustrating an example of a reference pattern, and FIG. 23 is a diagram illustrating an example in which the reference pattern S of FIG. 22 is converted into an edge for each pixel. In FIG. 22, the reference pattern S (dotted line) is shown with sub-pixel accuracy. Usually, the edge direction of the reference pattern is parallel to the horizontal direction (X direction) or vertical direction (Y direction) of the pixel. Similarly to the edge of the inspection target pattern image, the edge of the reference pattern also has information on the start point (subpixel accuracy), direction, and intensity for each pixel. In the present embodiment, all the edge strengths of the reference pattern are set to 1 except for a matching method using a unique pattern described later and a matching method using a 4.2.2 negative pattern .

  As shown in FIG. 24, the reference pattern may include a curve. In order to convert the curved portion of the reference pattern to the edge of the reference pattern, a tangent 263 at a point 262 on the reference pattern closest to the pixel center 261 is generated.

3.3 Before the recipe data inspection, the following set of inspection parameters called recipe data is set. Operator input parameters in the recipe data include design data search parameters, image acquisition parameters, and parameters for edge detection and inspection. In addition, there is a reference pattern generated by the reference pattern generation unit 11 as output data in the recipe data.

  Parameters for designating the device name and process name of the inspection target wafer (sample) are set as design data search parameters. As an image acquisition parameter, a slot number for specifying a wafer, a condition setting parameter of the irradiation system 310, a pixel interval, the number of pixels, and an inspection area are set.

  The pixel interval means a distance on the wafer with respect to the pixel interval of the inspection target pattern image. Values such as 1024 × 1024, 8192 × 8192, etc. are used for the number of pixels. The number obtained by multiplying the pixel interval by the number of pixels is the size on the sample with respect to the size of the pattern image to be inspected. Hereinafter, this size is referred to as a field of view (FOV). For example, if the pixel interval is 9 nm and the number of pixels is 8192 × 8192, the field of view is about 70 μm × 70 μm.

The following parameters are set as parameters for edge detection and inspection.
(1) Pattern deformation amount to be inspected Edge position movement amount Line width deformation amount Minimum line width Space width deformation amount Minimum space width Contact area inspection ratio Shift amount and diameter deformation amount for hole pattern and island pattern on wafer Defect determination coefficient of correction pattern that should not be formed

(2) The -side limit value and + side limit value of the allowable pattern deformation amount corresponding to the pattern deformation amount, and the limit value of the allowable direction difference of the edge used for matching. These deformation amounts are determined for each wiring attribute. Set to
(3) First edge detection parameter determined empirically from image quality First edge detection method Filter coefficient for edge expansion Binary threshold value of edge of inspection target pattern image p-tile method of edge of inspection target pattern image Coefficient p

(4) Parameters used by extraction rules to recognize reference pattern attributes (straight lines, corners, ends, isolated patterns, etc.) Reference pattern attributes are used to distinguish some or all of the reference patterns. The There are three types of typical reference pattern attributes. The first is an end (a straight line portion corresponding to the end of the reference pattern) or a straight line portion (a straight line portion not corresponding to the end of the reference pattern). The second is a corner (a vertex portion that does not touch a straight line portion having an end attribute). The third is an isolated pattern (a pattern that is isolated from other patterns).

(5) Parameters used by extraction rules for recognizing parts suitable for area inspection Maximum line width, minimum line length, and unused end length of reference patterns suitable for line width inspection Reference patterns suitable for space inspection Maximum line width, minimum line length, end unused length Maximum line width for easy-to-cut parts Maximum line length for easy-to-cut parts Maximum space width for easy-to-short parts Maximum space length for easy-to-short parts

(6) Second edge detection parameter determined empirically from image quality Length of profile acquisition interval Interval of profile acquisition interval Interval of sampling points in profile acquisition interval Method of recognizing edge from profile (using threshold method) Or the like)
Flag to set profile acquisition section when setting recipe data or after first edge is detected

(7) Minimum and maximum values of hole pattern size and safety factor (8) Number of inspection unit regions for which an average value is obtained in order to obtain a global pattern deformation amount (9) Maximum number of registered defect images

(10) Maximum number of registered defects to be re-inspected (11) Location suitable for automatic contrast / brightness adjustment and automatic focus adjustment (12) Representative distortion vector interval of distortion correction circuit

  Recipe data is managed using the device name, process name, and inspection mode, which are design data search parameters, as keys. The inspection mode is a generic name for image acquisition parameters and parameters for edge detection and inspection.

  FIG. 19 is a flowchart illustrating an example of recipe registration processing in the present embodiment. First, the operator inputs operator input parameters (such as design data search parameters) to the reference pattern generation unit 11 via the input device 4 (step S202).

  The reference pattern generation unit 11 searches the basic database 21 using the design data search parameters (device name and process name) as keys, and retrieves design data (step S204). The basic database 21 is a database that stores design data for the pattern image to be inspected. Next, the reference pattern generation unit 11 generates a reference pattern from the design data (step S206).

  Finally, the reference pattern generation unit 11 registers recipe data (reference pattern, operator input parameters) in the recipe database 22 (step S208).

3.4 Inspection Unit Area Inspection is performed for each inspection unit area obtained by dividing the input inspection area by the visual field, so that a reference pattern is generated for each inspection unit area. The inspection includes a sequential inspection and a random inspection.

  FIG. 20 is a diagram for explaining the sequential inspection. The inspection area is not set in units of the entire wafer as shown in FIG. 20, but as a plurality of areas designated by rectangles (such as an upper short rectangle and a lower long rectangle as shown in FIG. 20). Since it is set, in order to inspect the area at high speed, sequential scanning is performed for each inspection unit area. A reference pattern is created for each inspection unit area.

  FIG. 21 is a diagram for explaining the random inspection. In the random inspection, a certain area is not inspected sequentially but limitedly. In FIG. 21, only the inspection unit areas 301 to 304 are inspected.

3.5 Inspection results As inspection results, there are the following types of basic information.
(1) Information on abnormal pattern deformation amount defect (2) Information on defect detected from luminance distribution of pixel (3) Pattern deformation amount obtained from entire inspection unit area Information on using pattern deformation amount in pattern attribute unit There is the following information.
(4) Information on defects detected in pattern attribute units The following information is used as information using opposing edges.
(5) Information on defects detected by the area inspection method

4). Basic Inspection Processing FIG. 25 is a flowchart showing an example of basic inspection processing in the present embodiment. FIG. 26 and FIG. 27 are flowcharts illustrating another example of the inspection process according to the present embodiment, and are flowcharts illustrating an example of the inspection process when recognizing a defect that repeatedly occurs. The block A in FIG. 27 is the same as the block A in FIG. 26, and shows a process of preparation before the inspection. Block B in FIG. 27 is the same as block B in FIG. 26, and is a flowchart showing the inspection process for each inspection region.

  In the basic inspection process based on the flowchart shown in FIG. 25, first, the operator inputs recipe search parameters (device name, process name, and inspection mode) to the inspection unit 12 via the input device 4 (step S302).

  The inspection unit 12 searches the recipe database 22 using the recipe inspection parameter as a key, and extracts recipe data (step S304). Then, in order to acquire a pattern image to be inspected, image acquisition parameters are set to the image generation device 7 and instructions for wafer conveyance, alignment, and condition setting of the irradiation system device 310 are instructed (step S306).

  Alignment refers to a method for obtaining a conversion coefficient between a coordinate system used by design data and a coordinate system for managing a wafer observation position. This is embodied by CAD (Computer Aided Design) navigation. CAD navigation is a method in which after alignment, coordinate values to be observed on CAD data are converted into coordinate values for managing the wafer observation position, and the field of view of the imaging device is moved to that position to obtain an image at that position. Is well known.

As the image generating apparatus 7, the scanning electron microscope shown in FIG. 1 is most suitable, but various scanning microscopes or various microscopes such as a scanning focus ion beam microscope, a scanning laser microscope, and a scanning probe microscope are used. can do.
The image generation device 7 outputs the inspection target pattern image (and its central position) to the inspection unit 12 for each inspection unit region (step S308).

4.1 First Edge Detection Next, the inspection unit 12 detects a first edge from the inspection target pattern image (step S310). The following two edge detection methods can be used as the first edge detection. The first edge detection method is selected by the above-mentioned 3.3 recipe data “(3) First edge detection method”.

4.1.1 First edge detection method 1
One is a method suitable for an image having a contrast between the inside of the pattern and the background. Many of these images can detect edges by binarization, but cannot clearly detect edges when the contrast is relatively unclear. At this time, [Reference 1]: RMHaralick, “Digital step edges from ZERO crossing of second directional derivatives”, IEEE Trans. Pattern Anal. Machine Intell., Vol. PAMI-6, No.1, pp.58-68, Edges can be detected by applying the method disclosed in 1984. According to this method, the inflection point of the edge portion can be detected with subpixel accuracy.

4.1.2 First edge detection method 2
The other is a method of detecting an edge from an image having a bright edge and no contrast between the inside of the pattern and the background. For example, the method disclosed in [Document 2]: “Cartan Steger. An unbiased detector of curvilinear structures”, IEEE Trans. Pattern Anal. Machine Intell., 20 (2), February 1998 can be used. According to this method, the peak of the edge portion can be detected with subpixel accuracy. However, since this method cannot distinguish between the inside of the pattern and the background, the edge direction has only a value of 0 to 180 degrees.

As another method of the above-mentioned 4.1.1 First edge detection method 1 , the method of the above-mentioned document 2 may be used. In this case, an edge strength image is obtained by applying a differential filter (for example, Sobel filter or bandpass filter) to an image having a contrast between the inside of the pattern and the background, and an edge is detected from the obtained edge strength image. . In this case, the inside of the pattern and the ground can be distinguished.

  Since these methods are processes using a somewhat large window, not only the sub-pixel degree is obtained but also the direction of the edge is stable. Therefore, it is not always necessary to improve the edge detection accuracy by using a method for obtaining line segment information by connecting edges and performing linear approximation.

  In the first edge detection in step S310, the strength and direction of the edge are obtained for each pixel from the inspection target pattern image. The strength increases as the edge becomes clearer. In the case of an image having a contrast between the inside of the pattern and the background described in the first edge detection method 1 described above, the absolute value of the first-order differential value of the image is set as the intensity by using the method described in the above-mentioned literature 1. Then, an edge whose edge position is the zero-cross point of the secondary differential value of the image is recognized.

  On the other hand, in the case of an image in which only the edge is bright in the first edge detection method 2 described above, the sign inversion value (absolute value) of the secondary differential value of the image is set as the intensity using the method of the above-described literature 2. An edge whose edge position is the zero-cross point of the first-order differential value of the image is recognized. In any image, the edge is obtained with sub-pixel accuracy.

  FIG. 28 is a diagram showing an example of an image having a contrast between the inside of the pattern and the background explained in the first edge detection method 1 described above, and FIG. 29 is a diagram showing edges detected from the image of FIG. It is. FIG. 28 shows the luminance value for each pixel. As shown in FIG. 29, an edge is detected for each pixel, and information of a starting point (sub-pixel accuracy), a direction (0 to 360 degrees), and intensity is obtained for each pixel. As described above, the strength takes a larger value as the edge becomes clearer.

  FIG. 30 is a diagram illustrating an example of an image having a bright edge and no contrast between the inside of the pattern and the background described in the first edge detection method 2 described above, and FIG. 31 is detected from the image of FIG. It is a figure which shows an edge. Also in FIG. 30, the luminance value is shown for each pixel. Further, as shown in FIG. 31, an edge is detected for each pixel, and information on a start point (sub-pixel accuracy), a direction (0 to 180 degrees), and intensity is obtained for each pixel.

4.2 Linear Pattern Matching Method Next, the inspection unit 12 expands the edge of the inspection target pattern image. Hereinafter, the obtained result is referred to as an expansion edge (step S312). In the present embodiment, expansion is performed by an allowable pattern deformation amount that does not affect electrical characteristics. At this stage, the allowable pattern deformation amount is a positive integer. This value is a value obtained by converting the maximum value of the above-mentioned 3.3 recipe data “(2) -side limit value and + side limit value of allowable pattern deformation amount” into an integer. By expanding by an allowable pattern deformation amount, it is possible to allow and match pattern deformation that does not affect the electrical characteristics.

  FIG. 32 is a diagram illustrating an example of edge strength of a one-dimensional inspection target pattern image, and FIG. 33 is a diagram illustrating an example in which the edge of FIG. 32 is expanded. In FIG. 32 and FIG. 33, one-dimensional data is used for ease of explanation. In order to ignore the deformation within the allowable pattern deformation amount, a maximum value filter having a window twice as large as the allowable pattern deformation amount is applied to the edge of the inspection target pattern image. In the maximum value filter, the maximum value of each pixel in the window that is in the vicinity of the target pixel is obtained, and that value is used as the value of the pixel after filtering. In FIG. 33, the edge of FIG. 32 is expanded left and right by 2 pixels. This is an example when the allowable pattern deformation amount is 2 pixels.

  Consider the case where the edge of the reference pattern is FIG. First, a diagram obtained by shifting FIG. 34 is created. The shift amount is from 2 pixels in the left direction to 2 pixels in the right direction. Next, when a matching evaluation value to be described later is obtained from FIG. 33 and each shifted diagram, each matching evaluation value becomes the same value. Therefore, the shift amount cannot be uniquely determined.

  In order to solve this problem, as shown in FIG. 35, the edges of FIG. 32 are weighted and expanded. In order to realize the expansion in FIG. 35, smoothing filters having coefficients of 0.5, 0.75, 1.0, 0.75, and 0.5 may be used. In the case of the example shown in FIG. 35, the evaluation value decreases when FIG. 34 (edge of the reference pattern) is shifted by one or more pixels left and right.

  Next, as shown in FIG. 36, consider the edge of the reference pattern that is two pixels wider than the edge of the reference pattern shown in FIG. First, a diagram obtained by shifting FIG. 36 is created. The shift amount is 1 pixel in the left direction and 1 pixel in the right direction. Next, when a matching evaluation value to be described later is obtained from FIG. 35 and each shifted diagram, each matching evaluation value becomes the same value. Therefore, the shift amount cannot be uniquely determined.

  In order to solve this problem, as shown in FIG. 37, the edges in FIG. 32 may be weighted and expanded. In order to realize the expansion of FIG. 37, a smoothing filter (FIG. 38) having coefficients of 0.5, 0.9, 1.0, 0.9, and 0.5 may be used.

From the above consideration, the expansion as shown in FIG. 37 is most suitable. From the viewpoints of processing speed and edge coverage, expansion as shown in FIGS. 33 and 35 may be used.
After determining the smoothing filter coefficient, it is registered and used in the above-mentioned 3.3 recipe data “(3) filter coefficient for edge expansion”.

  39 is a diagram showing an example of edge strength of a two-dimensional inspection target pattern image, and FIGS. 40 and 41 are diagrams showing an example in which the edge of FIG. 39 is expanded. In FIG. 39, the edge strengths are all zero except at 20. FIG. 40 shows the result when the same expansion as in FIG. 33 is performed, and FIG. 41 shows the result when the same expansion as in FIG. 37 is performed.

  FIG. 42 is a diagram illustrating an example of an edge vector of a two-dimensional inspection target pattern image, and FIGS. 43 and 44 are diagrams illustrating an example in which the edge of FIG. 42 is expanded. FIG. 43 shows the result when the same expansion as in FIG. 33 is performed, and FIG. 44 shows the result when the same expansion as in FIG. 37 is performed. Expansion is performed for each of the X and Y components.

  The inspection unit 12 compares the expansion edge and the edge of the reference pattern, and performs matching for each pixel of the inspection target pattern image and the reference pattern (step S314).

In the present embodiment, matching is performed using the shift amount S ₂ with sub-pixel accuracy, as will be described in the pattern deformation amount obtained from the whole 4.8 inspection unit region described later. Therefore, here, matching is performed in units of pixels for the purpose of speeding up. Therefore, as shown in FIG. 45, matching is performed using an edge vector in which the edge vector of the reference pattern in FIG. 23 is expressed in units of pixels.

In matching in the present embodiment, in order to obtain a position where the evaluation value F ₀ is maximized, the reference pattern is shifted vertically and horizontally for each pixel with respect to the inspection target pattern image, and the obtained evaluation value F ₀ is maximized. The position is set as a matching position (FIG. 46). In the present embodiment, as shown by the following expression, the sum of the intensities of the dilated edges in the pixels where the edges of the reference pattern exist is set as the evaluation value F ₀ .

Here, E (x, y) is a vector having the intensity of the expansion edge as its magnitude and the direction of the expansion edge as its direction. The size of E (x, y) is 0 at a place where no edge exists. R (x + x _s , y + y _s ) is a vector having the edge direction of the reference pattern as its direction. However, the magnitude of R (x + x _s , y + y _s ) is the length within the pixel of the reference pattern. Here, (x _s , y _s ) is an edge shift amount S ₁ of the reference pattern.

If only pixels whose R (x, y) is not _{0 in} the calculation of the evaluation value F ₀ are stored, the calculation can be performed at high speed and the storage area can be reduced. Furthermore, the calculation is further speeded up by using the truncation of the calculation used in the Sequential Similarity Detection Algorithm (SSDA).

47 and 48 are diagrams in which FIG. 43 (expanded edge) and FIG. 45 (reference pattern edge) are superimposed. 47, pixel 254 corresponds to pixel 251 in FIG. 43 and pixel 252 in FIG. FIG. 48 shows the positional relationship when FIG. 43 is shifted to the right by one pixel and downward by one pixel from the positional relationship shown in FIG. Therefore, the pixel 255 corresponds to the pixel 251 in FIG. 43 and the pixel 253 in FIG. When the evaluation value F ₀ is used, the evaluation value increases as the degree of overlapping of pixels with edges increases. When the evaluation value F ₀ is used, an expansion process as shown in FIGS. 39 to 41 may be performed. The evaluation value F ₀ can be applied to any of the images described in 4.1.1 First edge detection method 1 described above and 4.1.2 First edge detection method 2 described above.

In the present embodiment, the evaluation value F ₀ is used, but other evaluation values can also be used. For example, in the case of an image having a contrast between the inside of the pattern and ground as described in 4.1.1 first edge detection method 1 described above, it may be used the following evaluation value F _a.

Further, for example, in the case of the above-described 4.1.2 first edge only bright image described by the edge detection method 2 can be used following evaluation value F _b.

When the evaluation value F _a or F _b is used, an expansion process as shown in FIGS. 42 to 44 may be performed.

Here, the evaluation values F ₀ , F _a , and F _b are considered. The evaluation value F ₀ is advantageous for high-speed calculation because the data is a scalar. On the other hand, the evaluation values F _a and F _b are effective in the case shown in FIG. 49, for example. When the evaluation values F _a and F _b are used, the edge (vector) of the vertical line portion of the reference pattern (FIG. 49A) and the edge of the horizontal line portion of the inspection target pattern image (FIG. 49B) ( Since the inner product with the vector) is close to 0, the portion 101 and the portion 102 match well. However, when the evaluation value F ₀ is used, since the determination is based only on the intensity regardless of the direction, the portion 101 and the portion 103 may be matched.

Since the evaluation value F _a can be distinguished from the inside of the pattern and the background, matching is more robust than the evaluation value F _b . For example, as shown in FIG. 50, when the line width 111, 113 and the space width 112, 114 used F _a if the same, which is a desirable result from F _b is obtained so can be distinguished whether the line or space.
In the present embodiment, matching is performed by expanding the edges of the pattern image to be inspected. Alternatively, matching can be performed by expanding the edges of the reference pattern.

4.2.1 Matching Method Using Unique Pattern In the method described above, processing was performed with the same strength of the edges of all the reference patterns. As another method, matching can be made robust by giving different weights to the edge strengths of the reference pattern. This method is implemented by the following procedure using FIG.

  51, (a) shows an example of a reference pattern, and (b) shows an example of a reference pattern (dotted line) of (a) and an inspection target pattern image (solid line) corresponding to the reference pattern. The reference pattern shown in FIG. 51 (a) is a periodic pattern, but has a gap in one place. When matching between this reference pattern and the pattern image to be inspected, as shown in FIG. 51B, even if both patterns are shifted by one cycle, they match except for the gap portion. It will be high. Therefore, the weighting of the edge strength corresponding to the gap portion is increased so that the matching evaluation value is greatly reduced when the gap of the pattern image to be inspected and the gap of the reference pattern do not match.

  As a weighting procedure, first, the cycle of the reference pattern is obtained by the autocorrelation method. Next, even if the original reference pattern is compared with the reference pattern shifted by one cycle, a pattern that is in the original reference pattern but not in the reference pattern shifted by one cycle is obtained. Then, the obtained pattern is recognized as a unique pattern, and the weighting (the degree of contribution) is made stronger than the other patterns. In order to express the degree of contribution, the reference pattern intensity is weighted greater than 1. This value can be a fixed value obtained from experience, or a fixed value divided by a ratio of unique patterns among all patterns.

4.2.2 Matching method using negative pattern As a method of using a unique pattern more efficiently, there is a matching method using a negative pattern which is a pair of unique patterns. FIG. 52A and FIG. 52B are diagrams schematically illustrating a method for calculating a matching evaluation value of a reference pattern in which rectangles are periodically arranged. Although rectangles are also periodically arranged on the right side of the inspection target images in FIGS. 52A and 52B, the end of the right reference pattern is not known because the images are limited. In such a case, if matching is performed using the above-described matching method using the 4.2.1 unique pattern , the matching evaluation values are substantially the same in FIGS. It is not decided uniquely.

As a countermeasure, a negative pattern that is a pair of unique patterns is extracted and used for matching evaluation value calculation according to the following procedure.
53 (a), 53 (b) and 53 (c) are diagrams schematically showing a method of using a negative pattern which is a pair of unique patterns. When there is no reference pattern at a portion shifted by one period in the left direction from the original reference pattern, the position of the original reference pattern is a unique pattern (rectangle indicated by a dotted line). A portion obtained by shifting the unique pattern by one period to the left is defined as a negative pattern (rectangle indicated by a solid line). Similarly, it implements about other directions, such as a right direction, an up direction, and a down direction.

  For the unique pattern, a value greater than 1 is used for the intensity of the reference pattern in order to express the degree of contribution as described above. On the other hand, for the negative pattern, a value obtained by multiplying the intensity of the reference pattern by a value larger than 1 and (−1) is used to express the degree of contribution.

Here, the evaluation value using the negative pattern is considered. The evaluation value when the pattern is present in one unique patterns and F _1. The evaluation value in FIG. 53A is (3 · F ₁ ), and FIG. 53B is (0). FIG. 53C shows (3 · F ₁ ) − (3 · F ₁ ) ≈ (0). From this calculation, FIG. 53A is determined as the matching position.

  According to the present embodiment, since the negative pattern imposes a large penalty on a one-cycle deviation from the optimal matching position, it is possible to accurately match the boundary between the portion where the same pattern is periodically arranged and the portion where it is not.

4.2.3 Matching Method Using Projection Data of Edges on Vertical Axis and Horizontal Axis The above matching method is sufficiently fast, but a method that can be executed at a higher speed is required. In order to increase the speed, the “matching for each pixel” portion in step S314 is improved.
Most of the design data is horizontal and vertical lines. Using this property, it is possible to perform faster matching using the projection data on the horizontal and vertical axes of the edges obtained from the design data and the projection data on the horizontal and vertical axes of the edges detected from the image to be inspected. become.

54 (a) and 54 (b) are diagrams showing a matching method using projection data on the horizontal and vertical axes of the edges detected by the above-described 4.1 first edge detection method . In the present embodiment, description will be made using edge detection suitable for an image having a contrast between the inside of the pattern and the background described in 4.1.1 First edge detection method 1 . The line segment forming the reference pattern has four directions, up, down, left, and right. Here, as an example, a matching method is shown by taking an upward line segment as an example.

(1) The total value L _{rp of the} lengths of all the line segments constituting the reference pattern is obtained. Next, the edges detected by the 4.1.1 first edge detection method 1 are sorted by intensity. _Lrp selected sorted edges are selected from the ones with higher strength, leaving them as edges and deleting other edges. Since the size of the reference pattern and the image to be inspected on the wafer is approximately the same, and the reference pattern is expressed in a coordinate system in units of pixels, the selected edge roughly corresponds to the edge of the reference pattern.

(2) An upward line segment is extracted from the line segments forming the reference pattern. Next, this line segment is projected onto the horizontal axis (X-axis) to create one-dimensional data. This one-dimensional data is in the form of an array, the element is an X coordinate value, and the value is the length of the line segment. Similarly, one-dimensional data is created by projecting this line segment onto the vertical axis (Y-axis). This one-dimensional data is in the form of an array, the element is the Y coordinate value, and the value is the length of the line segment. The result is as shown in FIG.

(3) Extract an upward edge from the selected edge. One-dimensional data is created by projecting this edge onto the horizontal axis (X-axis). This one-dimensional data is in the form of an array, the index is the X coordinate value, and the element value is the Y component of the edge (vector). Similarly, one-dimensional data is created by projecting this edge onto the vertical axis (Y-axis). This one-dimensional data is in the form of an array, the index is the Y coordinate value, and the element value is the Y component of the edge (vector). The result is as shown in FIG.

(4) While matching the projection data on the horizontal axis of the upper edge within the range in the X direction shown in FIG. 46, the matching error with the projection data on the horizontal axis of the upper line segment is calculated. Similarly, a matching error value E _pm with the projection data on the vertical axis of the upper line segment is calculated while shifting the projection data on the vertical axis of the upper edge within the range in the Y direction shown in FIG. The calculation result of the matching error value E _pm is shown in FIG.

(5) The maximum value E _{pm_max} and the minimum value E _{pm_min} of the matching error value E _pm are obtained, and the threshold value is obtained by the following equation.
A shift amount having a matching error value E _pm below this threshold is determined to be suitable for matching. Here, k _mt is an empirically determined value and takes a value between 0 and 1, and the closer to 0, the greater the number of shift amounts determined to be suitable for matching. It is determined that the shift amount indicated by the arrow in FIG. 56 is suitable for matching.

(6) Next, an optimum solution is obtained by the following procedure from the shift amount determined to be suitable for matching. In the above-described 4.2 linear pattern matching method , “in the matching in this embodiment, in order to obtain a position where the evaluation value F ₀ is maximized, the reference pattern is vertically, horizontally , and horizontally for each pixel pattern to be inspected. The position where the obtained evaluation value F ₀ is maximized is set as the matching position (FIG. 46). ” When this method is adopted, this portion is expressed as “the amount of shift obtained in (5) above for the reference pattern with respect to the inspection target pattern image in order to obtain the position where the evaluation value F ₀ is maximum in the matching in this embodiment. It shifted vertically and horizontally each, resulting evaluation value F ₀ is the matching position the position of maximum (Figure 46). "and read so will implement the matching method of the straight line-shaped pattern described above.

The matching error value E _pm is calculated by the method shown in FIG. In the present embodiment, as representative examples, the element R _p [i] of the projection data onto the horizontal axis of the upward line segment, the element E _p [i] of the projection data onto the horizontal axis of the upward edge, and the shift amount S _Shows how to use _p . The simple matching error value E _pmS is _calculated from the element R _p [i] of the projection data to the horizontal axis of the upward line segment and the element E _p [of the projection data to the horizontal axis of the shifted upward edge corresponding thereto. Using i + S _p ], the following equation is obtained.
The sum Σ _i means the sum for all elements E _p [i].

  As described in step S312 (the edge of the inspection target pattern image is expanded to obtain the expanded edge), it is necessary to allow the allowable pattern deformation amount that does not affect the electrical characteristics. Although the same method as step S312 may be used, the following method is used as another method.

First, the following calculated values are executed for all elements E _p [i]. Here, a case where the allowable pattern deformation amount is 1 pixel will be described.
(1) If,
If so, the following calculation is executed.

(2) If,
If the next δR is positive, the following calculation of E _p [i + S _p ] from ρ ₋₁ is executed.

(3) If,
If δR is negative, the following calculation is executed.

After the above calculation is completed, a matching error value E _pmD considering the allowable deformation amount is obtained by the following equation.

The results of this calculation are shown in FIGS. 57 (b) and (c). In FIG. 57 (b), R _p [i] and E _p [i + S _p ] are placed at positions suitable for matching. On the other hand, in FIG. 57 (c), R _p [i] and E _p [i + S _p ] are placed at a position shifted by one pixel from the position suitable for matching. As shown in FIGS. _{57B and 57C} , the matching error value E _pmD considering the allowable deformation amount is _smaller than the simple matching error value E _pmS by the amount associated with the allowable deformation amount. It is a value. Therefore, the matching error value E _pmD considering the allowable deformation amount is suitable as the matching error value E _pm .

If the allowable pattern deformation quantity is larger than one _{pixel, R p [i-1]} , in addition to R _p of _{R p [i + 1] [} i-2], R p [i + 2], using, for example, Can be processed.

The above-described calculation of the matching error value E _pm is also performed on the edge and line segment in the downward left direction and the right direction. Moreover, you may use the line segment of the other direction, for example, the direction of the multiple of 45 degree | times.
In the above example, it is possible to distinguish 180-degree reverse edges such as upward and downward edges, but when the first edge detection method 2 is used, processing is performed without distinguishing 180-degree reverse edges. It will be.

In Figure 46, illustrating a method by shifting vertically and horizontally the reference pattern for each pixel with respect to the inspection object pattern image, the evaluation value F ₀ is the matching position the position of maximum. However, according to the present embodiment, instead of shifting every pixel, it is possible to shift at intervals of skipping, so that the calculation time can be greatly shortened.

4.3 Matching method of hole pattern and island pattern using geometric information The above-mentioned matching method is optimal for a linear pattern. However, another method can be used for matching the hole pattern and the island pattern. The hole pattern and the island pattern are rectangles, and both the long side and the short side are patterns that are not more than a few times the minimum line width. The matching of the hole pattern and the island pattern requires more calculation time due to the matching because there are more polygons smaller than the linear pattern. In order to solve this problem, it is possible to use the following method capable of reducing the amount of calculation and increasing the speed as compared with the above-described 4.2 linear pattern matching method .

  This method can be used when all patterns are isolated patterns. In addition, the hole pattern and the island pattern usually do not exist at the same time. Therefore, in this embodiment, a method in which all patterns are hole patterns will be described. For the island pattern, the above-described matching method can be realized by replacing the hole of this embodiment with an island.

  The first method of matching the hole pattern is a method using geometric information obtained from the edge of the inspection target pattern image. FIG. 58 is a schematic diagram for explaining a first method of hole pattern matching. In FIG. 58 (a), the edges detected from the inspection target image are displayed in bold lines. The center of gravity of the edge is indicated by a black circle (●) point.

As a first step, as shown in FIG. 58A, an edge is detected, and the outermost frame and the center of gravity of the edge are obtained. In the case of an image having a contrast between the inside of the pattern and the background, the edge detection described in the above 4.1.1 First edge detection method 1 can be used.

In the case of the above-mentioned 4.1.2 First edge detection method 2 in the case where the described edge is bright and there is no contrast between the inside of the pattern and the background, the above-mentioned 4.1.2 First edge detection method The edge detection described in 2 can be used. In this case, since the edge is not necessarily recognized as a connected pixel, after expanding the edge, the connected pixels are obtained by labeling, and the outermost frame and the center of gravity of these connected pixels are obtained to obtain the outermost edge of the edge. Frame and center of gravity.

As a second stage, the obtained edges are selected by the following procedure using FIG.
(1) The above-mentioned 3.3 recipe data “(7) hole pattern size minimum value _Shmax and maximum value _Shmin and safety factors k _hmin , k _hmax ” are determined and registered in _advance .
(2) When the size of the outermost frame of the edge is larger than S _hmax × _khmax, it is not regarded as the edge of the hole pattern. Here, k _hmax is a value empirically determined by a value of about 1 to 2.

(3) In addition, (2) when the size of the outermost frame of the edge is smaller than _Shmin × _khmin, it is regarded as noise or dust and is not regarded as the edge of the hole pattern. Here, k _hmin is a value determined empirically with a value of about 0.5 to 1.
(4) The connected edges are not regarded as edges unless they form a ring shape.
(5) In the case of an image having a contrast between the inside of the pattern and the ground, it is possible to determine whether the ring-shaped inside of (4) is a hole or an island. If it is not a hole, it is not considered an edge.

In the present embodiment, matching is performed using the evaluation value F _h instead of F ₀ , F _a and F _b used in the above-mentioned 4.2 linear pattern matching method . Except for using F _h as the evaluation value, the same processing as the above-described linear pattern matching method is used. In this embodiment, the reference pattern is obtained by simply converting the design data. The evaluation value F _h is a value obtained by obtaining the values obtained by the following method for the reference patterns, which are all hole patterns, and taking the sum.

(1) As shown in the first column of FIG. 58 (c), the value is 0 if the center of gravity of the edge does not exist in the reference pattern.
(2) The value is 1 if the center of gravity exists in the reference pattern as shown in the second column of FIG.

In order to use the matching method using the 4.2.1 unique pattern and the matching method using the 4.2.2 negative pattern used in the above-described linear pattern matching method in this example, the following two calculations are performed. to add. The method for recognizing the unique pattern and the negative pattern and the setting of the degree of contribution to the unique pattern and the negative pattern are the same as the above-described linear pattern matching method.
(3) If the center of gravity exists in the unique pattern, the value becomes the weight described above.
(4) If the center of gravity exists in the negative pattern, the value is the aforementioned weight × (−1).

  By using this embodiment, it is possible to realize a matching method using information obtained by collecting a plurality of edges. This method can be executed faster than the method of matching using individual edges. Furthermore, the amount of calculation can be greatly reduced.

Furthermore, it is possible to increase the speed by applying the above-described matching method using the projection data of the 4.2.3 edge onto the vertical axis and the horizontal axis . In this case, projection data of the center of gravity of the edge is used instead of projection data of the edge.

4.4 Hole Pattern / Island Pattern Matching Method Using Statistics The second method of hole pattern matching is a method of comparing the statistics of the image corresponding to the inside of the reference pattern with the statistics of the image corresponding to the outside It is. FIG. 59 is a schematic diagram for explaining a second method of hole pattern matching. FIG. 59A shows a reference pattern used in this embodiment. This reference pattern is obtained by sizing the reference pattern obtained from the design data. The amount to be increased by the size processing is an amount less than half of the value in the direction in which 3.3 the recipe data “(2) Allowable diameter deformation amount in the case of hole pattern and island pattern” increases. FIG. 59B is a typical hole pattern image. The edge of the hole pattern is brighter than the base, and the inside of the hole pattern is darker than the base.

In the present embodiment, matching is performed using the evaluation value F _d instead of F ₀ , F _a and F _b used in the above-described 4.2 linear pattern matching method . Processing other than using _Fd as the evaluation value uses the same processing as the above-described linear pattern matching method. The evaluation value F _d is obtained by the following procedure.

(1) As shown in FIG. _59C , a histogram H _inside is obtained for the inspection pixels corresponding to all the reference patterns. The obtained histogram is normalized.
(2) Histogram H _outside is obtained for pixels existing outside all reference patterns. The obtained histogram is normalized.
(3) Each difference histogram H _difference element is calculated as the difference between the corresponding histogram H _inside element and the corresponding histogram H _outside element. The sum of absolute values of each element of the difference histogram H _difference is defined as an evaluation value F _d .

In order to use the matching method using the 4.2.1 unique pattern and the matching method using the 4.2.2 negative pattern used in the above-described linear pattern matching method in this embodiment, the following two calculations are performed. Add The method for recognizing the unique pattern and the negative pattern and the setting of the degree of contribution to the unique pattern and the negative pattern are the same as the above-described linear pattern matching method.

(4) For the inspection pixel corresponding to the unique pattern, the histogram H _inside is obtained after converting one pixel into the number of pixels corresponding to the aforementioned weight.
(5) For the inspection pixels corresponding to the negative pattern, the histogram H _inside is obtained after converting one pixel into the number of pixels corresponding to the above-mentioned weight × (−1).

What the above calculation (5) means is as follows. If there are holes in the negative pattern, the sum of the elements in the histogram H _inside decreases, but the shape does not change much. Therefore, in this case, the evaluation value F _d is substantially equal to the evaluation value F _d before the calculation of this negative pattern. On the other hand, if there is no hole in the negative pattern, the histogram H _inside resembles the difference histogram H _difference . The evaluation value F _d using the difference histogram H _difference and the histogram H _outside is larger than the evaluation value F _d using the histogram H _inside and the histogram H _outside . Therefore, in this case, the evaluation value F _d is larger than the evaluation value F _d before the calculation of this negative pattern.

In the hole pattern and the island pattern, the brightness distribution of the background image varies depending on the location due to the influence of the charging phenomenon or the like. This means that the histogram H _outside widens. However, the influence of the evaluation value F _d due to the spread of the histogram H _outside is not so great.

  According to this embodiment, since the difference histogram between the inside and outside of the hole pattern and island pattern is used as the evaluation value, the distribution of the brightness of the background image varies depending on the location due to the influence of the charging phenomenon or the like. A matching method that is not easily affected can be realized. Note that this method can also be used as a linear pattern matching.

4.5 Processing matching after matching is performed, and when the shift amount S ₁ = (x _s , y _s ) that takes the maximum evaluation value is obtained, the reference pattern is shifted by S ₁ . Subsequent processing is performed with this shift. The shift amount S ₁ can be output to the display device 5 and the printing device 6 as an inspection result.

After the matching is completed, the edge of the inspection target pattern image is binarized. The binarization is executed by using the above-described 3.3 recipe data “(3) threshold value of edge of the pattern image to be inspected” with respect to the edge strength.

The p-tile method can be used as another method of binarization. In this method, the edge images of the inspection target pattern image in descending order of the edge strength so that the number of pixels of the edge image of the inspection target pattern image having 1 is {number of pixels corresponding to the edge of the reference pattern × p}. So that each pixel has 1. Here, p is usually a number of about 0.9 to 1.1, and is used by being set to 3.3 recipe data “(3) coefficient p of p-tile method of edge of pattern image to be inspected”.

4.6 First Inspection Next, the inspection unit 12 performs a first inspection. Specifically, pattern deformation amount calculation, defect detection, and defect type recognition are performed. The inspection unit 12 first associates the edge of the inspection target pattern image with the edge of the reference pattern (step S318). Edge positions are handled with sub-pixel accuracy. Therefore, the distance between edges can also be obtained with sub-pixel accuracy. The direction is determined as a value of 0 to 360 degrees, for example, with the right direction being 0 degrees.

In this embodiment is performed in the inspection target distance between the pattern image of the edge and the shift amount S ₁ of the shifted reference pattern edge, and the following procedures correspondence considering the direction of both edges. For each edge of the reference pattern, an edge of the pattern image to be inspected within the distance of the above-mentioned 3.3 recipe data “(2) Allowable pattern deformation amount” is searched. Then, among the detected edges, those whose direction difference from the edge of the reference pattern is equal to or smaller than the above-described 3.3 recipe data “(2) allowable edge direction difference” are associated as edges within the allowable deformation amount. The associated vector d (x, y) between both edges can be used to determine the pattern deformation amount. When a plurality of edges are recognized by the above-described procedure, the edge having the shortest distance and the smallest direction difference is adopted.

  FIG. 60 is a diagram illustrating an example of correspondence between the edge of the inspection target pattern image and the edge of the reference pattern. In FIG. 60, edges are indicated by arrows in order to indicate directions. In the example of FIG. 60, in each pixel including the edge of the reference pattern, association is performed by searching for the edge of the inspection target pattern image in the direction perpendicular to the edge direction from the center of the edge of the reference pattern. . If the edge of the inspection target pattern image whose distance from the center of the edge of the reference pattern is equal to or smaller than the allowable pattern deformation amount and whose direction difference is equal to or smaller than the allowable direction difference of the edge is found, both edges are associated with each other. The vector d (x, y) in FIG. 60 is the above example.

  In FIG. 61, (a) shows an example of an edge of a reference pattern, and (b) shows an example of an edge of an inspection target pattern image corresponding to the reference pattern of (a). An example of correspondence between both edges will be described with reference to FIG. In this example, the allowable pattern deformation amount is one pixel. The allowable direction difference of the edge is 60 degrees. For example, when the edge of the inspection target pattern image corresponding to the edge 81 of the reference pattern is searched, the edge 68 is within the distance of the allowable pattern deformation amount of the edge 81 and the direction difference is equal to or less than the allowable edge direction difference. Therefore, the edge corresponding to the edge 81 is recognized. As for the edge 84 of the reference pattern, the edge 70 is recognized as the edge of the corresponding inspection target pattern image.

  Here, the edge 61 is not within the distance of the allowable pattern deformation amount with respect to the edge 82 of the reference pattern. The edge 64 is not within the distance of the allowable pattern deformation amount, and the direction difference is larger than the allowable direction difference of the edge. The edges 66 and 69 are within the distance of the allowable pattern deformation amount, but the direction difference is larger than the allowable direction difference of the edge. Therefore, an edge corresponding to the edge 82 cannot be obtained. Similarly, the edge 83 cannot be obtained.

  The examples in FIGS. 61A and 61B are methods that do not distinguish between the inside and the outside of the pattern, and the direction has only a value of 0 to 180 degrees, but the method that distinguishes the inside and outside of the pattern. It is also possible. For example, if the edge direction is determined so that the inner side of the pattern is always placed on the right hand, FIG. 61 (a) becomes as shown in FIG. 62, and the association can be executed more strictly.

  Next, the inspection unit 12 performs defect detection (step S320). The following two methods can be used for defect detection.

4.6.1 Method for Recognizing Abnormal Pattern Deformation Defects As a first method of defect detection, a defect having an abnormal pattern deformation amount is recognized by the following procedure. FIG. 63 is a diagram schematically illustrating a method of recognizing an abnormal pattern deformation amount defect. The inspection unit 12 recognizes the edges (for example, the edges 61 to 67, 69, and 75 in FIG. 61B) of the inspection target pattern image that could not be associated as defective pixels. A binary bitmap representing the defective pixel is obtained.

Next, the binarized bitmap obtained as shown in FIG. _63A is expanded by an expansion width W _dilation (2 pixels in this figure), and the pixels are connected. The dilated pixel can be obtained as a result of the dilation operation of the binarized bitmap. This Dilation operation is one of typical operations of morphology.

As shown in FIG. 63B, when a defect is detected, the defect may be divided and detected due to factors such as noise. By using the expansion width W _dilation which is an empirical value for a defect that should be one even if it is divided in this way, the defect can be fused.

Here, the Dilation operation and the Erosion operation, which are typical operations of morphology, will be described. Dilation operation δ and Erosion operation ε are operations that output the following results when A is a target image (binarized bitmap) and B is a structural element (binarized bitmap).
(A) _-b means that A is translated by -b. Also, ∪ and ∩ mean bit map sum operation (OR) and product operation (AND) for all b satisfying b∈B.

  Next, the pixels connected by the labeling process are recognized as one area. Here, the labeling process is a method of writing a same value to pixels connected in the vicinity of 4 or 8 to generate a connected pixel group. A group of connected pixels can be distinguished by giving a different value to pixels that are not connected. The connected pixel group is recognized as a defect, and a circumscribed rectangle of the defect is obtained. The circumscribed rectangle of a defect means the smallest rectangle that includes pixels recognized as a defect.

The above procedure is executed as shown in FIG. In FIG. 63 (b), there are discontinuous defects for the line segment in the lower right direction. These defects are originally one defect but are divided. First, an area recognized as a defect is obtained as a binarized image (indicated by black pixels), and the binarized image is expanded by an expansion width W _dilation as indicated by white pixels. Next, a black pixel and a white pixel are obtained as a region connected by a labeling process, and a minimum rectangular region including this region is obtained as a circumscribed rectangle.

  Finally, the center of the circumscribed rectangle is calculated as the defect position, and the size of the circumscribed rectangle is calculated as the defect size. The obtained defect position and defect size are used as defect information.

4.6.2 Defect Recognition Method Using Pixel Luminance Distribution As a second method of defect detection, a defect is recognized using the pixel luminance distribution in the following procedure. First, a region is obtained by connecting edges of the pattern image to be inspected that have been associated with each other. The luminance values of the pixels existing in the inner and outer portions of the obtained area are obtained. Each luminance value distribution can be expected to be a normal distribution if there is no defect. Therefore, it is possible to detect a defective pixel by applying a quality control method.

  Next, pixels having luminance that is not a normal distribution are detected and connected to obtain a region. Finally, the center of the circumscribed rectangle of the obtained region is calculated as the defect position, and the size of the circumscribed rectangle is calculated as the defect size. The obtained defect position and defect size are used as defect information.

  FIG. 64 is a diagram schematically showing a defect recognition method using the luminance distribution of pixels. A broken line 201 indicates the edge of the inspection target pattern image. The solid lines 202 and 203 on both sides of the broken line 201 are line segments obtained by thickening edges by a specified width, and a portion surrounded by the solid lines 202 and 203 is recognized as an edge region. The luminance values of the background 204 and the pattern interior 205 are approximately normally distributed.

  As shown in FIG. 65, the portion D where the luminance value exceeds about ± 3σ is highly likely to be a defect. D includes noise, but the noise exists relatively uniformly in the region, while the defect exists in a solid state. A binarized map is created in which a pixel having a luminance value of D is 1 and a pixel having other luminance values is 0. Erase a block of pixels (for example, a block of pixels 207 in FIG. 64) having a 1 less than or equal to a specified size (for example, 2 × 2 pixels). A median filter or the like can be used for this processing. This size is an empirical value considering the size of the defect to be detected. The remaining block of pixels with 1 (eg, pixel block 206 in FIG. 64) is considered a defect.

The above-described 4.6.1 abnormal pattern deformation amount defect recognition method detects defect detection in the vicinity of an edge. On the other hand, the defect recognizing method using the pixel luminance distribution according to the present method detects a defect at a place other than the vicinity of the edge.

  When a defect is detected, defect information (here, defect position, size information and image) is output to the defect type recognition unit 14 (steps S322 and 324).

4.7 The defect type defect type recognition unit 14 using the feature amount obtained from the image determines the defect type using the defect information and the information in the defect type reference database 23 (step S326). Specifically, a feature amount is obtained from a given image, and is compared with the feature amount of the image stored in the defect type reference database 23 to determine the defect type. The defect type recognition unit 14 outputs the defect information (defect position, size information, and image) and the defect type to the display device 5 and the printing device 6 via the output unit 13 (step S328). Here, the defect type reference database 23 is obtained by registering already acquired images for each defect type.

  The defect type recognition unit 14 can perform automatic defect type classification according to the following procedure. That is, the geometric information of the cluster of pixels recognized as a defect is obtained. Geometric information is a kind of feature quantity. From the obtained geometric information, it is possible to recognize the shape characteristic of the defect such as round or elongated, and it can be recognized as a foreign object if round or a scratch if elongated. A pixel recognized as a defect is divided into three parts: an inside, an outside, and a boundary of the pattern. For each of these portions, a feature amount using the pixel luminance value of the inspection target pattern image is obtained. Based on the characteristic amount obtained here, it is possible to determine whether the foreign substance is a metal piece or an organic substance (for example, a human skin). In other words, the type can be identified by being bright if the foreign material is metal and dark if it is organic.

  Also, if there is a large variation in the brightness of a pixel that is recognized as a foreign object inside the pattern, it is judged that there is a high possibility that the foreign object is present on the pattern, and conversely, the luminance variation is small It is determined that there is a high possibility that a foreign object exists under the pattern. This is a difficult process in the conventional die-to-die method because it cannot be distinguished from the image whether the defect exists inside or outside the pattern. Using these feature quantities, the defect type is determined by a well-known classification method. As the classification method, a method of making a determination by performing comparison with the defect type reference database 23 by the k shortest distance method is effective.

  The automatic defect type classification described above is a method according to the conventional optical method and SEM method ADC (Automatic Defect Classification), but according to the method of the present invention using design data, the inside and the outside of the pattern. Since the distinction is clearly made, the feature amount of each part can be obtained accurately, and the classification accuracy is improved.

4.8 Pattern Deformation Amount Obtained from Entire Inspection Unit Region Next, the inspection unit 12 obtains a pattern deformation amount from the relationship between the edge of the pattern image to be inspected and the edge of the reference pattern (step S330). . The pattern deformation amount is obtained from a portion where no defect is detected as a result of the defect detection. Then, the pattern deformation amount is output to the display device 5 and the printing device 6 via the output unit 13 (step S332).

  There are two types of pattern deformation amounts: a pattern deformation amount obtained from the entire inspection unit region and a defect detected in a pattern attribute unit. As a pattern deformation amount obtained from the entire inspection unit region, for example, a positional deviation amount, a magnification variation amount, and a line width deformation amount can be used.

The displacement amount is obtained as an average value of vectors d (x, y) between the associated edges. This is the shift amount S ₂ with sub-pixel accuracy. A shift amount with subpixel accuracy is obtained by adding the shift amount S ₁ described in the processing after 4.5 matching to the shift amount S ₂ .
When the inspection is performed with sub-pixel accuracy, the reference pattern is shifted using S ₁ + S ₂ as the updated shift amount S ₁ , and steps S318 to S330 are executed again.

  In order to obtain the magnification fluctuation amount in the X direction, a regression line is obtained by approximating the X component of the vector d (x, y) related to the line segment of the reference pattern in the vertical direction by the regression line D (x). Then, the gradient of the regression line is set as the magnification fluctuation amount in the X direction. The same applies to the magnification fluctuation amount in the Y direction.

66A shows an example of the edge (broken line) of the reference pattern and the edge (solid line) of the inspection target pattern image, and FIG. 66B shows a vector d at y = y ₀ between the edges shown in FIG. An example in which the X component of (x, y ₀ ) is approximated by a regression line D (x) is shown. When the X component of the vector d (x, y ₀ ) is approximated by a regression line D (x) = ax + b, the slope a corresponds to the magnification fluctuation amount. In the example of FIG. 66A, it can be seen that the pattern of the inspection target pattern image is larger than the reference pattern as a whole.

In FIG. 67, (a) shows another example of the edge (broken line) of the reference pattern and the edge (solid line) of the inspection target pattern image, and (b) shows y = y ₀ between the edges shown in (a). An example in which the X component of the vector d (x, y ₀ ) is approximated by a regression line D (x) is shown. In the example of FIG. 67A, in addition to the fact that the pattern of the inspection object pattern image is larger than the reference pattern as a whole, the width of the linear pattern is thick. In FIG. 67A, the linear pattern 121, 122, 123 of the reference pattern corresponds to the linear pattern 124, 125, 126 of the inspection target pattern image, respectively.

The amount of deformation of the line width in the X direction can be obtained, for example, as an average value of the X component −D (x)} of sign (x, y ₀ ) · {d (x, y ₀ ). Here, sign (x, y ₀ ) takes −1 if the position of (x, y ₀ ) is at the left end of the line, and takes 1 if the position is at the right end of the line. If the standard deviation of the X component −D (x)} of sign (x, y ₀ ) · {d (x, y ₀ ) is obtained with respect to the deformation amount of the line width, a line width variation index can be obtained.

4.9 Method for Recognizing Defects Detected in Pattern Attribute Units First, pattern attributes will be described. As shown in FIG. 68, as a pattern attribute, a straight line portion 171, a corner 172, a terminal end 173, an isolated pattern 174, or the like can be used. As the pattern deformation amount related to the pattern attribute, for example, in addition to the positional displacement amount, the magnification variation amount, and the line width deformation amount described in the above-mentioned 4.8 inspection unit region , the diameter and area , Deformation amounts of feature quantities such as perimeter, circularity, moment, and radius of curvature can be used.

Pattern attributes can be automatically added to the reference pattern. In order to extract pattern attributes, the above-mentioned 3.3 recipe data “(4) Parameters used by extraction rules for recognizing reference pattern attributes (straight line, corner, end, isolated pattern, etc.)” are set. And use.

4.9.1 End Position Misalignment Defects FIGS. 69 (a) and 69 (b) are diagrams showing end position misalignment amounts. As shown in FIG. 69A, the amount of positional deviation at the end is the minimum distance between the edge 164 constituting the end of the reference pattern and the edge 163 of the inspection target pattern image.

In addition, as shown in FIG. 69B, an average value, maximum value, minimum value, or median value of a plurality of distances corresponding to a section 157 having an arbitrary width may be used as the end position displacement amount. Good.
If the displacement amount is not within the range of the movement amount of the end allowable edge position in 3.3 Recipe data “(2) Limit value on the negative side and the limit value on the + side of the allowable pattern deformation amount”, This termination is recognized as having a defect.

4.9.2 Position Misalignment Defects at Straight Line and Corner In FIGS. 69 (a) and 69 (b), the amount of misalignment at the end has been described. However, the amount of misalignment can also be measured at the straight line and corner. For the straight line portion, the defect is inspected from the positional deviation amount obtained for the section corresponding to the straight line portion. For a corner, a defect is detected by obtaining a positional deviation amount in a direction having a half angle formed by the corner or a specified angle.
In these cases, instead of the movement amount of the allowable edge position at the end, the movement amount of the allowable edge position of the straight line portion and the corner is used.

4.9.3 Isolated Pattern Misalignment Defect FIG. 70 shows the amount of misalignment of an isolated pattern. The misregistration amount is the misregistration amount between the centroid 162 of the edge 160 of the reference pattern (which constitutes an isolated pattern) and the centroid 161 of the edge 159 of the pattern image to be inspected (which constitutes an isolated pattern).
If the positional deviation amount is not within the range of the movement amount of the permissible edge position of the isolated pattern in 3.3 Recipe data “(2) Permissible pattern deformation amount -side limit value and + side limit value”. This isolated pattern is recognized as having a defect.

4.9.4 Other isolated pattern defects It is also possible to inspect the amount of deformation of an isolated pattern feature. As the feature amount, diameter, area, perimeter, circularity, moment, etc. can be used. As shown in FIG. 70, the feature amount is calculated from the edge 160 of the reference pattern and the edge 159 of the inspection target pattern image, and the difference between the feature amounts can be inspected.

4.9.5 Abnormal curvature defect at corner In FIG. 71, (a) shows an example of a corner of a reference pattern, and (b) shows an example of a corner of a pattern image to be inspected. The corner pattern of the edge 166 of the reference pattern shown in FIG. 71A is rounded. As the radius of curvature of the corner, for example, a major axis, a minor axis, or a radius obtained by approximating the corner curve with an ellipse or a circle using a least square approximation can be used. By obtaining the curvature radius of the corner of the edge 166 of the reference pattern and the curvature radius of the corner of the edge 165 of the inspection target pattern image, the deformation amount of the curvature radius of the corner can be obtained and inspected.

The above inspection is simultaneously performed on a plurality of locations in the field of view. The inspection item is selected according to the above-described 3.3 recipe data “(1) pattern deformation amount to be obtained”.

4.10 Pattern Attribute Extraction Rule Example of 3.3 Recipe Data “(4) Parameters Used by Extraction Rule to Recognize Reference Pattern Attributes (Linear Part, Corner, End, Isolated Pattern, etc.)” Will be described with reference to FIG. The straight line portion 171 is extracted as a line segment having a length equal to or longer than the predetermined length L. The corner 172 is extracted as a portion in the vicinity of the contact point of two line segments that make contact at a predetermined angle (90 degrees, 135 degrees, 270 degrees, etc.). The end 173 is a line segment having a length equal to or less than a predetermined length L, and is extracted as a line segment having both ends 173t and 173t that are in contact with the straight line portions 171 and 171 at an angle of 90 degrees. The end 173 and the two straight portions 171 and 171 have a U-shape. The isolated pattern is extracted as a closed figure having a predetermined area or less.

4.11 The second edge detection inspection unit 12 detects an edge (second edge) again from the inspection target pattern image (step S334). The detected second edge is detected from the profile obtained from the inspection target pattern image. As the second reference pattern, a reference pattern having a point Q in FIG. 76 as an edge is used. On the other hand, when only the edge described in the above 4.1.2 First edge detection method 2 is a bright image, a reference pattern in which the point P is an edge is used as the first reference pattern. Therefore, the second reference pattern and the first reference pattern are generally different.

Before performing the second edge detection of the inspection target pattern image, the second reference pattern is shifted by the shift amount S ₁ + S ₂ described above. Subsequent processing is performed with this shift.

  In order to detect an edge from a profile, various methods such as a threshold method and a linear approximation method are disclosed. In this embodiment, a line width performed by a CD-SEM using a threshold method included therein. The length measurement is applied to a two-dimensional pattern (inspection target pattern image). However, similar processing is possible even if the threshold method is replaced with another method such as a linear approximation method. Here, the straight line approximation method is a method of approximating a profile with a straight line and detecting an edge using an intersection.

The following two methods can be used as a method for setting the profile acquisition interval. One of them is a method of presetting the direction and position for acquiring the profile using the second reference pattern. This method is executed when the above-described 3.3 recipe data “(6) Flag for setting the profile acquisition section when setting the recipe data or after detecting the first edge” is off. In this method, the profile acquisition section is uniquely set from the second reference pattern.

FIG. 72 is a diagram illustrating an example of a profile acquisition section. As shown in FIG. 72, the profile acquisition section is set in the vertical direction of the second reference pattern with the second reference pattern as a midpoint (line segment indicated by a double line). The length of the profile acquisition section is the above-mentioned 3.3 recipe data “(6) Length of profile acquisition section”, and the interval of the profile acquisition section is the above-mentioned 3.3 recipe data “(6) Profile acquisition section”. Interval.
Instead of the second reference pattern described above, as shown in FIG. 73, a curve (solid line in the figure) forming the outer shape of the exposure pattern obtained by the lithography simulator may be used.

For the section corresponding to the profile acquisition section from the inspection target pattern image, the profile data is acquired with the above-mentioned 3.3 recipe data “(6) Sampling point interval in profile acquisition section”. The length of the profile acquisition section is longer than the pattern deformation allowable amount. The interval between sampling points is usually a value equal to or less than the pixel interval. For the creation of profile data, methods such as bilinear interpolation, spline interpolation, and Fourier series are used.

  74 is an enlarged view of a part (part B) of FIG. 72, and FIG. 75 is an enlarged view of a part (part C) of FIG. The double line in the figure is the profile acquisition section, the intersection of the grids indicates the position of the pixel, and the black point indicates the position where the luminance value of the inspection target pattern image is acquired.

As shown in the figure, bilinear interpolation means luminance values I (0,0) and I (0,0,1) of pixels indicated by (0,0) (0,1) (1,0) (1,1). 1), I (1,0), I (1,1) are used to calculate the luminance value I (x, y) of the point at position (x, y), (0 <x ≦ 1,0 <y ≦ 1) y) is calculated by the following formula.

  From the profile obtained using this equation, the second edge position is detected using the threshold method. As shown in FIG. 76, the maximum luminance value V and its position P in the obtained profile are obtained. A numerical value obtained by multiplying the maximum luminance value V by a predetermined coefficient k is set as a threshold value T, and an intersection point of the luminance value = threshold value T and the profile curve is obtained. At these intersection points, an intersection point Q that is located outward from the point P and closest to the point P is obtained. For all profiles, the second edge is detected using the method for determining the intersection point Q.

  The cross-sectional shape of the wiring formed on the wafer is trapezoidal. Whether the length measurement is performed at the upper side, the lower side, or the intermediate portion of the cross-sectional shape can be set by a coefficient k.

  When the second edge is detected, curve approximation (including polygon approximation) is performed using the detected second edge, and the detected second edge is connected. The simplest method is simply connecting with a broken line. However, when the following division and fusion method is used, the second edges detected by polygon approximation by the least square method can be connected. T. Pavlidis and S. L. Horowitz: “Segmentation of plane curves”, IEEE Trans. On Computers, vol. C-23, no. 8 Aug., 1974. An example of this method is shown in FIG.

In addition to this, it is also possible to use curve approximation by smoothing plane data using the least square method and a two-dimensional spline function as shown in FIG. The former can be processed at high speed, but is not flexible to those that contain many rounded shapes. On the other hand, the latter has characteristics of satisfying high speed and flexibility. In addition to these, various methods such as a method using a Fourier descriptor are disclosed, and these methods can be replaced.
Note that the curve approximation as described above can be performed after the first edge detection is performed.

As another method of setting the profile acquisition interval, a method of adaptively setting the profile acquisition interval at the time of edge detection can be used. This is executed when the above-mentioned 3.3 Recipe data “(6) Flag to set the profile acquisition section when setting the recipe data or after detecting the first edge” is ON.

  In this method, as shown in FIG. 78A, a profile acquisition section is set in the vertical direction of the first edge of the detected pattern image to be inspected. According to this method, as shown in FIG. 78B, even if the first edge (solid line) of the detected pattern image to be inspected deviates from the second reference pattern (dotted line) described above, the profile The acquisition interval can be made shorter than the method described above. Also, this method is easier to follow the deformation of the pattern than the above-described method. After setting the profile acquisition interval, processing similar to that described above is performed.

4.12 After the second edge detection equal to or higher than the second inspection, the inspection unit 12 performs the second inspection (step S336). This inspection is the same processing as S320 to S332 of the first inspection described above, except that the second edge is used instead of the first edge. In step S318, the edge of the pattern image to be inspected is associated with the edge of the reference pattern. In the second inspection, the edge is associated with the profile acquisition section.

Defect detection is performed in the second inspection, and a pattern deformation amount is obtained. The positional deviation amount (shift amount) S ₃ relating to the entire image obtained here corresponds to the shift amount S ₂ described in the above-described 4.8 pattern deformation amount obtained from the entire inspection unit region . The shift amount S ₃ obtained here, plus the shift amount S ₁ and the shift amount S ₂ described above becomes the total shift amount between the pattern of the inspected pattern image and the second reference pattern.

In the second inspection, the above-described 4.6.1 abnormal pattern deformation amount defect recognition method and the above-described defect recognition method using the luminance distribution of 4.6.2 pixels are set as follows.

In the first inspection 4.6.1 abnormal pattern deformation amount defect recognition method , an edge pixel of an inspection target pattern image that could not be matched is recognized as a defect. However, in the second inspection, the profile acquisition section in which the edge does not exist in the range of the above-described 3.3 recipe data “(2) -side limit value and + side limit value of allowable pattern deformation amount” is treated as a defect. .

In the defect inspecting method using the luminance distribution of the 4.6.2 pixels of the first inspection, the region is obtained by connecting the edges of the inspection target pattern images that have been associated with each other. However, in the second inspection, the region is obtained by connecting the edges of the reference pattern.

  If the above basic inspection process has been performed for all inspection unit areas, the inspection process is terminated, otherwise the process returns to step S308 (step S340).

5. The application inspection process and the above are the basic inspection processes based on the flowchart shown in FIG. In this chapter, the applied inspection process, which is an extension of this basic inspection process, is described.

5.1 Method for recognizing repeated defects 4. As described in the basic process , an example of a process for recognizing a defect that repeatedly occurs is shown in FIG. This process is an extension of the process shown in FIG.

  First, a block A indicating a process to be prepared before the inspection is executed. Next, defects are fused after completion of block B, which is a process of inspecting the inspection area of each semiconductor device (step S402). Block A and block B in FIG. 27 are the same as block A and block B in FIG. 26, respectively. Steps S302 to S306 in block A are the same as steps S302 to S306 in FIG. Further, S308 to S336 in the block B are the same as S308 to S336 in FIG.

  In block B, step S338 for outputting the inspection result to the defect information storage unit 24 is added, which is different from FIG. Step S340 in block B is the same as step S340 in FIG. The inspection unit area in step S340 is an inspection area expressed by coordinates on the design data, and this inspection area is an area to be inspected for a plurality of semiconductor devices.

  When inspecting an inspection area wider than the inspection unit area shown in FIG. 20, a defect existing at the boundary between the inspection unit area and the inspection unit area may be divided into a plurality of locations and detected. By fusing these divided defects at a plurality of locations, it is possible to eliminate the influence of the boundary between the inspection unit regions.

  FIG. 79 schematically shows a case where the inspection area is divided into four inspection unit areas. The defect A exists across the upper right inspection unit region and the lower right inspection unit region. First, the circumscribed rectangle 31 of the defect belonging to the upper right inspection unit region and the circumscribed rectangle 32 of the defect belonging to the lower right inspection unit region are obtained. The circumscribed rectangle 31 and the circumscribed rectangle 32 are obtained by the procedure shown in FIG.

  Next, an overlap inspection is performed on circumscribed rectangles included in all the inspection unit areas constituting the inspection area. If they overlap, the minimum bounding rectangle including all of the overlapping bounding rectangles is merged into a bounding rectangle. In this example, a circumscribed rectangle M fused from the circumscribed rectangle 31 and the circumscribed rectangle 32 is obtained. Note that the circumscribed rectangle M (indicated by a dotted line), the circumscribed rectangle 31 and the circumscribed rectangle 32 should partially overlap, but for convenience of illustration, the circumscribed rectangle M is drawn slightly larger. ing.

  Similarly, the defect B existing over four inspection unit areas is also fused. In this case, four circumscribed rectangles are merged to obtain one merged circumscribed rectangle (step S402). The defect information existing in the obtained circumscribed rectangle is merged, and the merged defect information is stored in the defect information storage unit 24 (step S403).

  After checking whether or not all the inspection target semiconductor devices have been inspected (step S404), if it is determined that the inspection has been completed, a defect that repeatedly occurs is recognized (step S406). These pieces of defect information are obtained from the same inspection region for a plurality of semiconductor devices manufactured based on the same design data, and are expressed in the coordinate system used by the design data. It is stored in the information storage unit 24.

  In FIG. 80, defect information obtained from the first semiconductor device and the second semiconductor device is schematically shown. When the defect information obtained from the first semiconductor device and the defect information obtained from the second semiconductor device are overlapped, it is determined that the circumscribed rectangle 33A and the circumscribed rectangle 33B overlap the circumscribed rectangle 34. This processing is widely known as a graphic logic operation. A common circumscribed rectangle 35 is obtained as a minimum circumscribed rectangle including these three circumscribed rectangles. A defect (not shown) existing in the common circumscribed rectangle 35 is recognized as a common defect, that is, a defect that repeatedly occurs.

  In this case, there is a defect that repeatedly occurs inside the common circumscribed rectangle 35, and when the defect of the first semiconductor device is detected, the defect is detected by being divided into the circumscribed rectangle 33A and the circumscribed rectangle 33B due to factors such as noise. This means that when a defect of the semiconductor device is detected, it is detected as a block of circumscribed rectangles 34. The fact that the circumscribed rectangle 33A, the circumscribed rectangle 33B, and the circumscribed rectangle 34 are displaced means that the defect has been detected at a slightly displaced position.

  The above processing can be similarly executed by a method using defect information obtained from N semiconductor devices which are three or more numbers. In this case, when circumscribed rectangles obtained from M or more semiconductor devices overlap, a defect that repeatedly occurs is recognized. M is a numerical value from 2 to N, and the larger it is, the more precisely the defect that repeatedly occurs can be acquired.

  The defect information of the repeatedly generated defect obtained by the above inspection is output to the defect information storage unit 24 (step S408). The contents of the defect information storage unit 24 are output to the display device 5 and the printing device 6 via the output unit 13 (step S410).

  If this embodiment is used, it is possible to eliminate a large amount of simple labor by the operator and prevent a reduction in defect recognition due to an operator error. In addition, when a sample is contaminated by a carbon coat or the like to be described later, the contaminant is hardly present at the same location of different dies, so that the contaminant is not recognized as a defect that repeatedly occurs.

As another method, out of defect information obtained from a plurality of semiconductor devices, defect information is obtained from the entire inspection region for at least one semiconductor device, and defect positions in the defect information for other semiconductor devices. There is a method of recognizing a defect that occurs repeatedly by inspecting only its vicinity and obtaining defect information.
In this embodiment, a method for obtaining defect information by obtaining defect information from all locations in the inspection region of the first semiconductor device and inspecting only the vicinity of the defect position in the defect information for the second semiconductor device will be described. To do.

  As a first stage, block A, block B, step S402, and step S403 are executed for the first semiconductor device. On the left side of FIG. 81, defect information of the first semiconductor device obtained through step S403 is schematically shown. An image corresponding to the vicinity of the circumscribed rectangle 41 obtained from the first semiconductor device is acquired from the second semiconductor device, and the defect is inspected. On the right side of FIG. 81, the result of such limited inspection is schematically shown as defect information obtained from the second semiconductor device.

  As a second step, it is examined whether the defect information obtained from the first semiconductor device overlaps with the defect information obtained from the second semiconductor device. In FIG. 81, the circumscribed rectangle 41 overlaps the circumscribed rectangle 42A and the circumscribed rectangle 42B. Next, a common circumscribed rectangle 43 is obtained as a minimum circumscribed rectangle including these three circumscribed rectangles. A defect (not shown) existing in the common circumscribed rectangle 43 is recognized as a defect common to both semiconductor devices, that is, a defect that repeatedly occurs. A similar process is performed on the circumscribed rectangle 51. As described above, the second semiconductor device is limitedly inspected for all the defects included in the defect information obtained from the first semiconductor device.

  When the number of defects included in the defect information obtained from the first semiconductor device is small, this method is executed at a higher speed than the previous method.

5.2 Area Inspection Method In the above-described 4.6 first inspection and 4.12 second inspection , design data is simply converted into a reference pattern. As another inspection method, an inspection method for extracting a reference pattern suitable for the region inspection method using the geometric information of the line segments forming the design data or the relationship between the line segments forming the design data that touches or is close to each other can be used. . The area inspection method means an inspection method using opposite edges.

  As the area inspection method, the line width inspection method of the linear shape pattern, the average line width inspection method, the space width inspection method, the average space width inspection method, the line width inspection method of the curved shape pattern, the average line width inspection method, the space width inspection method An average space width inspection method, a cutting / short circuit, a method for inspecting a portion that is easily cut or short-circuited, or a gate line width inspection method can be used.

5.2.1 Line Width Inspection Method, Average Line Width Inspection Method, Space Width Inspection Method, and Average Space Width Inspection Method for Linear Patterns Line width, average line width, space width, and average space width for semiconductor inspection There is a method for managing the process of a semiconductor device by monitoring. According to the present embodiment, a reference pattern suitable for line width, average line width inspection, space width, or average space width inspection is extracted from the design data of the layer corresponding to the inspection target, and for each extracted reference pattern Thus, an inspection method for setting an allowable pattern deformation amount suitable for line width, average line width inspection, space width, or average space width inspection can be realized. These inspection methods are carried out according to the following procedure.

  FIG. 82 schematically shows a rule for automatically extracting a reference pattern suitable for line width inspection from design data. Polygonal object that is a linear shape pattern of design data that is narrower than the specified maximum line width Lw and that is longer than the specified minimum line length Lm to obtain a reference pattern suitable for line width inspection become. As shown in the left part of FIG. 82, there are three linear patterns in the design data of FIG. The left straight line pattern is the target of processing, but the middle straight line pattern is not the target of processing because it is greater than or equal to the maximum line width Lw. Also, the right linear pattern is not subject to processing because it is less than the minimum line length Lm.

  Next, as shown in the right part of FIG. 82, the selected linear pattern is shrunk to the inside of the end unused length Lo specified from the end. This linear pattern is divided into rectangles having a section length Li, and the divided rectangles are registered as line width inspection reference patterns A (indicated by solid lines). Further, a reference pattern B (indicated by a double line) centered on the boundary of the reference pattern A for line width inspection obtained here may be added as the reference pattern B for line width inspection.

  By adding the reference pattern B, the ability to detect defects existing at the boundary of the reference pattern A is improved. The defect detection capability is higher as the ratio of the size of defects existing in the reference pattern is larger. Let R be the ratio of the size of the defect to the size of the reference pattern when the defect exists in one reference pattern. When a defect having the same size is divided into two reference patterns, the ratio of the defect size to the reference pattern size is smaller than R. Therefore, the defect detection capability is higher when the defect exists in one reference pattern.

  As shown in FIG. 83, a linear pattern having corners is processed after being separated into rectangles at the corners. The L-shaped polygon having the corner indicated by the dotted line in FIG. 83 is separated into two rectangles indicated by the solid line.

  The space width inspection can be realized by performing the same process as described above using the inverted design data. Inverted design data is obtained by inverting the inside of the pattern in the design data to the outside and the outside to the inside. FIG. 84 schematically shows a rule for automatically extracting a reference pattern suitable for space width inspection from design data. As shown in FIG. 84, Lm ′, Lw ′, Li ′, and Lo ′ have the same meaning as Lm, Lw, Li, and Lo, but generally use different values. Using these values, the space width inspection may be performed by the same method as described in FIG. The values of Lm, Lw, Li, Lo, Lm ′, Lw ′, Li ′, and Lo ′ used above are the maximum values of the reference pattern suitable for the above-mentioned 3.3 Recipe data “(5) Line width inspection”. Width, minimum line length, end non-use length, maximum line width of reference pattern suitable for space inspection, minimum line length, end non-use length ”.

An inspection using a reference pattern suitable for line width inspection and a reference pattern suitable for space width inspection is performed in the following procedure. The average edge position of the edge of the inspection target image corresponding to the line segment of the obtained reference pattern and existing in the design data is calculated. The distance between the average edge positions is calculated, and the difference between the obtained distance and the line width or space width W of the design data is 3.3 recipe data “(2) Allowable pattern deformation amount of line width” or “(2 When the allowable pattern deformation amount of “space width” is exceeded, the portion corresponding to this reference pattern is recognized as a defect.

  FIG. 85 schematically shows an inspection using a reference pattern suitable for line width inspection and a reference pattern suitable for space width inspection. The reference pattern includes a line segment Ld indicated by a double line existing in the design data and a line segment Le added when separated into rectangles. As shown in FIG. 72, a profile is acquired in the direction perpendicular to the line segment Ld, and an edge is obtained from the profile as shown in FIG. An average of these obtained edge positions is obtained to obtain an average edge position.

In FIG. 85, a left average edge position A and a right average edge position B are obtained. Next, the distance W ′ between the average edge position A on the left side and the average edge position B on the right side is obtained, and the difference from the line width W of the design data is obtained. If this difference is greater than the allowable pattern deformation amount, it is recognized as a defect.
As another method, the edge position may be obtained from a profile obtained by obtaining each profile on the line segment Ld and averaging these profiles.

Although the method of the average line width inspection or the average space width inspection has been described above, each line width or space width may be inspected without averaging.
As will be described later, 5.3.1 Gate line width inspection method is a kind of line width inspection method of linear shape pattern and average line width inspection method of linear shape pattern. The extraction method is different.

5.2.2 Line width, average line width, space width, and average space width inspection method of curve shape pattern Line width, average line width, space width, and curve shape pattern that cannot be performed by the above-described area inspection method An average space width inspection method can be used. A corner portion is typical as a curved shape pattern. Although the curved shape pattern inspection requires complicated calculation, these inspections are important as a method for managing the process of the semiconductor device as well as the straight line portion.

  The procedure of the line width inspection method for the curved shape pattern which is a corner portion will be described below. FIG. 86 is a diagram schematically showing a procedure for obtaining a reference pattern suitable for the line width inspection at the corner portion, and FIG. 87 schematically shows a procedure for the minimum line width inspection for the curved shape pattern as the corner portion. FIG.

  As shown in FIG. 86, a reference pattern (two rectangles indicated by solid lines) suitable for line width inspection of a linear pattern from a reference pattern (L-shaped polygons indicated by dotted lines) obtained from design data. Polygons CP1, CP2, and CP3 obtained by deleting are obtained. The obtained polygon CP2 which is not a pattern including a terminal end is selected as a reference pattern suitable for the line width inspection of the corner portion.

  The line width to be inspected is a reference pattern to be inspected for the line width at the corner portion and is a curve corresponding to a line segment existing in the design data (shown by a thick solid line in FIG. 87). As shown, it is the minimum distance between the two (corresponding to an L-shaped line segment having a corner portion corrected by a curve). A second edge corresponding to these curves is detected (see FIGS. 72 to 76). The double line shown in FIG. 87 represents the profile acquisition section, and the black circle (●) represents the detected second edge.

The following process is performed for all detected second edges corresponding to the lower left curve.
(1) Find the distance between one detected second edge corresponding to the lower left curve and all detected second edges corresponding to the upper right curve.
(2) Obtain the smallest of the obtained distances.
If the obtained minimum value of each distance is less than the above-mentioned 3.3 recipe data “(2) allowable minimum line width”, it is determined that a defect exists. Here, the average line width may be calculated by calculating the average line width instead of the minimum line width.

  The curved pattern is generally composed of a plurality of line segments having different line widths. The curved pattern is used for circuit connection. For the above reasons, the minimum line width inspection is more suitable than the inspection using the allowable pattern deformation amount.

  As another method, there is a method using an Erosion operation on a binarized image. FIG. 88 is a diagram schematically showing a procedure of a minimum line width inspection of a curved shape pattern that is a corner portion using the Erosion calculation. The minimum line width inspection using the Erosion operation is executed in the following procedure. As explained before, the Erosion operation is one of the typical operations of morphology.

(1) A polygon is formed by continuously connecting the detected second edge on the lower left side and the detected second edge on the upper right side clockwise or counterclockwise. In FIG. 88, all detected second edges are continuously connected in the clockwise direction as indicated by arrows CW1 to CW5.
(2) Convert the obtained polygon into a binary bitmap. (Lattice-like portion in FIG. 88)
(3) A rectangle having the width of the radius of the structural element used in the Erosion calculation is added to the line segment Lc added when CP2 is created. (Two rectangular parts indicated by dots in FIG. 88)

(4) The result of Erosion calculation of the obtained binarized bitmap is obtained (two regions Me surrounded by thick lines in FIG. 88). The structural element used in the Erosion calculation uses a circle whose diameter is the above-mentioned 3.3 recipe data "(2) -side limit value of allowable deformation amount of line width".

(5) If the image portion ILc corresponding to the line segment Lc is connected in the region Me, it is determined that there is no defect. However, in this case, since the image portion ILc corresponding to the line segment Lc is not connected in the region Me, it is determined as a defect.
The above processing is the line width inspection, but the space width inspection is similarly inspected.

  If this embodiment related to the above area inspection is used, since these area inspections use the average value of a plurality of defect information, the defect detection ability and the defect recognition accuracy use the single edge inspection. Greatly improved compared to the method.

5.2.3 Method for inspecting a portion that is likely to be cut or short-circuited One of the methods for inspecting the line width or space width of the above-mentioned 5.2.2 linear pattern can be a method for inspecting a portion that is likely to be cut or short-circuited. FIG. 89 is a diagram schematically showing a method for extracting a portion that is easily cut or short-circuited. As shown on the left side of FIG. 89, narrower than 3.3 recipe data of the line width is above "(5) Maximum line width Bw of the scissile moiety", and 3.3 Recipe data "(5) cleavage of the above A rectangle γ that is a portion of the linear shape pattern of the design data, which is shorter than the maximum line length Bl of the portion that is easy to perform, is extracted.

  The rectangle γ that is the extracted portion corresponds to a portion that is easily cut and is registered as a reference pattern. The line segment α and line segment β are inspected by performing edge detection shown in FIG. Here, since the line segment β has rounded corners, each line width is inspected without averaging.

Similarly, as shown on the right side of FIG. 89, for the portion that is easily short-circuited, the above-mentioned 3.3 recipe data “(5) Maximum space width Sw of the portion that is easily short-circuited” and the above-mentioned 3.3 recipe data “( 5) The rectangle ζ obtained by using the “maximum space length Sl of the portion that is easily short-circuited” is registered as the portion that is easily short-circuited, and the space width is inspected.

  As shown in FIG. 90, another inspection method for a portion that is easily cut or short-circuited is performed according to the following procedure. The left side of FIG. 90 schematically shows an inspection method for a portion that is easily cut, and the right side of FIG. 90 schematically shows an inspection method for a portion that is easily short-circuited. The rectangular pattern indicated by the thick black frame in FIG. 90 is the same as the rectangle γ and the rectangle ζ in FIG. In addition, the image portion corresponding to the lattice-shaped portion in FIG. 90 clearly has a background and contrast, but the image portion corresponding to the portion indicated by dots has a thin contrast. A portion indicated by a dot on the left side of FIG. 90 shows a state in which it is cut. Moreover, the part shown with the dot of the right side of FIG. 90 has shown the state short-circuited.

  In such a case, there are three types of edges. One is an edge that exists at the boundary between the base and the grid-like portion, and the other is an edge that exists at the boundary between the base and the portion indicated by dots. The last is an edge that exists at the boundary between the lattice-like portion and the portion indicated by dots. In the case of cutting as shown on the left side of FIG. 90, an edge present at the boundary between the background and the portion indicated by the dot is detected, and no defect is detected. In the case of a short circuit as shown on the right side of FIG. 90, an edge existing at the boundary between the lattice-like portion and the portion indicated by dots is detected, and no defect is detected. Even in such a case, the following inspection is performed as a method for detecting a defect present in a portion indicated by a dot.

It is checked whether an edge exists in the direction of the arrow in the eight sections including γ and ζ indicated by G. There should be no edges in this part as well as parts that are susceptible to cutting or shorting. Therefore, when an edge is detected in these eight sections, γ or ζ is recognized as a defect.
According to this embodiment, it is possible to detect a cut or short-circuited defect observed with a thin contrast. In addition, it is possible to set a defect type having information that it has been cut or shorted.

  According to the present embodiment relating to these area inspections described above, it is possible to perform a wide range of inspections that are impossible with operator inspection.

5.3 Inspection Method Using Logical Operation of Polygon of Layer Design Data Related to Object to be Checked and Polygon of Layer Design Data Related to the same 4.6 First Inspection and 4.12 Second Inspection In the inspection , the inspection is performed using a reference pattern obtained from the design data of the layer related to the inspection object. However, a more advanced inspection can be realized by an inspection method that uses a reference pattern obtained from a logical operation result of the layer design data related to the inspection object and the layer design data related thereto.

5.3.1 Gate Line Width Inspection Method The first method of the inspection method using the above-described logical operation is a method for extracting and inspecting a reference pattern suitable for region inspection using the logical operation. As this method, a gate line width inspection method and an end cap inspection method are used.

  A semiconductor device inspection includes a transistor gate width inspection. The target of the gate width inspection is a portion where the polysilicon layer and the active layer (the previous process of the polysilicon layer) overlap. FIG. 91 shows an inspection using the reference pattern obtained by the logical product operation of the layer design data corresponding to the inspection object and the layer design data related to the process before and after the layer design data.

A reference pattern C (rectangle indicated by a solid line) is obtained by a logical product of the polygon which is the design data of the polysilicon layer and the polygon which is the design data of the active layer. The AND operation used here uses the method used in computational geometry. The inspection is executed using the reference pattern C in the same manner as the above-mentioned 5.2.1 Line width inspection method of the linear pattern or the above-mentioned 5.2.1 Average line width inspection method of the linear pattern .

  According to this embodiment, the gate portion can be automatically extracted. As a result, the gate line widths of all semiconductor devices can be automatically inspected, which can greatly contribute to improving the performance of semiconductor devices.

5.3.2 End cap inspection method As an end inspection method, there is an end cap inspection of a gate portion. First, an end cap recognition method will be described. A polygon formed by removing the reference pattern C from the design data of the polysilicon layer of FIG. 91 is obtained. This corresponds to polygon F and polygon G in FIG. Those patterns that satisfy the following conditions are recognized as end caps.
(1) The line width W (FIG. 91) is a rectangle having a specified value or less. (2) The length D (FIG. 91) to the end is not more than a specified value.

A polygon F satisfies these conditions. Next, the end of the polygon F is inspected in the same manner as in FIGS. 69 (a) and 69 (b). Generally, the value of the allowable pattern deformation amount for managing the contraction of the simple end and the end cap of the gate is different. In order for the latter to secure an effective gate length compared to the former, a more strict allowable pattern deformation amount is set.
If this method is used, the gate end cap can be automatically set with an allowable pattern deformation amount different from that of a simple end, so that the gate end cap can be strictly inspected.

5.3.3 Method for adaptively setting the allowable pattern deformation amount for the end of the wiring layer in contact with the contact hole / via hole The second method of the inspection method using the logical operation described above is the end of the wiring layer in contact with the contact hole / via hole. This is an adaptive setting method of the allowable pattern deformation amount for. In this method, the termination used for connection with the contact hole / via hole is identified with no margin of a certain value or more, and the permissible pattern deformation amount is adaptively set and inspected at the recognized termination. In FIG. 92, this method is schematically shown.

  The contact area between the end of the wiring layer and the contact hole / via hole is inspected. Even if the termination has the same shape, the value of the allowable pattern deformation amount for managing the contraction of the termination differs between the termination used for connection between the simple termination and the contact hole / via hole. Since the latter allowable pattern deformation amount needs to secure a contact area, a strict allowable pattern deformation amount is set as compared with the former.

  The allowable pattern deformation amount for the termination of the wiring layer in contact with the contact hole / via hole is determined in consideration of the alignment error and the margin of termination at the time of exposure of the wiring layer and the contact hole / via hole layer. Many end margins have a certain value or more. However, in the case of complicated wiring, the terminal margin may not be able to ensure a certain value or more.

A method for identifying a terminal used for connection with a contact hole / via hole and having no margin of a certain value or more is performed by the following procedure.
(1) A solid-line rectangle in the upper left frame of FIG. 92 is created that includes the line segment Lea that is the end of the wiring layer and has a length of the allowable pattern deformation amount in the inner direction of the polygon of the wiring layer design data. This rectangle is called an end vicinity pattern.

(2) The result of the logical product of the near-termination pattern and the contact hole / via hole layer design data is obtained as a region. This area is a rectangle indicated by dots. Terminations with no margin above a certain value generate this region.

  The allowable pattern deformation amount of the shrinkage at the end of the wiring layer related to the occurrence of this region is made smaller than the allowable pattern deformation amount of the shrinkage at the other end. The amount to be reduced is the length Δ in FIG. An inspection is performed by adaptively setting an allowable pattern deformation amount reflecting the length Δ at the recognized end.

  According to the present embodiment, the allowable pattern deformation amount can be set in accordance with the margin of the termination used for connection with the contact hole / via hole.

5.3.4 Contact Area Inspection Method The third method of the inspection method using the logical operation described above is a method for calculating and inspecting the contact area of the terminal used for connection with the contact hole / via hole. This method is another method for managing the end of the wiring layer and the contact hole / via hole. 93 (a) and 93 (b) are diagrams showing this method.

First, the reference pattern Rca obtained by the logical product operation of the polygon which is the design data of the wiring layer and the polygon which is the design data of the contact hole / via hole is obtained by the same method as shown in FIG.
Next, an edge with respect to the line segment Ld indicated by the double line existing in the design data is detected by the same method as in FIG. A polygon Pca is obtained by connecting the detected edges. The dotted line of the polygon Pca is a line segment that connects the end of the edge to each line segment Ld.
Finally, the ratio of the area of the polygon Pca and the area of the reference pattern Rca is calculated. If the obtained ratio is smaller than the above-mentioned 3.3 recipe data “(2) Permissible contact area inspection ratio”, it is determined that a defect exists.

5.4 Method for Inspecting Correction Pattern that must not be Formed on Wafer In the above-mentioned 4.6 first inspection and 4.12 second inspection , the pattern to be formed on the wafer is inspected. In addition to such an inspection, there is an inspection of a correction pattern that should not be formed on the wafer. For example, there is a pattern that is added as a kind of OPC pattern for the purpose of correcting a pattern existing in the vicinity of the pattern, but the pattern itself is not formed on the wafer. Such a correction pattern may be formed on the wafer and become a defect. In recent years, such patterns have been used in large quantities, but there is no automatic inspection method.

  FIG. 94A shows an example of a correction pattern that should not be formed on the wafer. FIG. 94B schematically shows a correction pattern inspection method that should not be formed on the wafer. As a solution to the above problem, an inspection method is performed according to the following procedure.

(1) The above OPC pattern is converted into a reference pattern. In this method, an OPC pattern is used instead of design data.
(2) The second edge is detected using the reference pattern obtained as shown in FIG. Even if such an OPC pattern is formed as a pattern, the OPC pattern is considerably different from the reference pattern. Therefore, it is necessary to lengthen the profile acquisition section so as to cover this deformation.

(3) When the ratio of the number of detected edges to the number of profile acquisition sections is larger than the above-mentioned 3.3 recipe data “(2) Defect determination coefficient K _{cp of} correction pattern that should not be formed on wafer” Judge that a defect exists. Here, the defect determination coefficient Kcp is an empirical value of 0.1 or less.

  According to the present embodiment, a correction pattern inspection method that should not be formed on a wafer by applying edge detection can be realized.

5.5 Pattern Inspection Method Requiring Signal Strength Correction In the above-described 4.6 first inspection and 4.12 second inspection , patterns to be inspected are individually inspected. However, due to phenomena caused by fluctuations in the generation rate of secondary charged particles and capture rate, the distance between two edges forming part of the pattern is observed narrower than the substance, or observed more widely There is. These phenomena occur in portions of the pattern corresponding to the proximity line segment and the separation line segment of the reference pattern. The adjacent line segment is a line segment in which the distance between the closest line segments among the opposing line segments is smaller than a specified value. A separation line segment is a line segment in which the distance between the closest line segments among the opposing line segments is greater than a specified value.

  For example, the distance between the edges corresponding to the adjacent line segments may be observed to be widened, and the distance between the edges corresponding to the separated line segments may be observed to be narrowed. According to the present embodiment, this phenomenon is corrected by providing a method of correcting the positions of the proximity / separation line segments and a method of setting an allowable pattern deformation amount different from the normal allowable pattern deformation amount.

  Further, the distance between the two edges corresponding to the separation line segment may be shorter than the distance of the design data due to variations in process conditions, but this does not necessarily affect the characteristics of the semiconductor device. In such a case, the deformation can be ignored by increasing the allowable pattern deformation amount in the pattern separation line segment.

  FIG. 95 schematically shows a method for extracting adjacent line segments from the reference pattern. Let Dp be the maximum distance between adjacent line segments that need to be corrected. First, a line segment close to the line segment indicated by the bold line on the right side of the left rectangle in the left frame in FIG. 95 is obtained. The line segment to be obtained is a line segment that exists in the right direction opposite to the line segment indicated by the bold line and forms the left side of the reference pattern. The dotted line in the left frame in FIG. 95 satisfies this condition. Next, the line segment Lp indicated by the dotted line of the center rectangle whose distance from these line segments and the line segment indicated by the thick line is equal to or less than Dp is selected. Finally, the selected line segment Lp is projected onto the illustrated line segment with a thick line, and the overlapping portion is recognized as the adjacent line segment. This is a line segment indicated by a wavy line in the right frame of FIG.

  For the adjacent line segment, the position of the line segment is corrected, and the allowable pattern deformation amount is set to a value different from that of the other line segments. Here, the position correction amount and the allowable pattern deformation amount may be varied according to the distance.

  FIG. 96 schematically shows a method for extracting a separation line segment from the reference pattern. Let Dt be the minimum distance between line segments that need to be corrected. First, a line segment close to the line segment indicated by the bold line on the right side of the left rectangle in the left frame in FIG. 96 is obtained. A line segment that opposes the line segment indicated by the bold line is a line segment that exists in the right direction and forms the left side of the pattern. The dotted line segment Lt in the left frame in FIG. 96 satisfies this condition.

  Next, the line segment whose distance from the line segment indicated by the bold line is equal to or less than Dt is selected. Finally, the selected line segment is projected onto the line segment indicated by the bold line, and the overlapping portion is recognized as not being the correction target line segment. This is a line segment indicated by a wavy line in the right frame of FIG. As a result, a line segment indicated by a bold line and excluding the above-described wavy line segment is recognized as a separated line segment.

  Also for the separation line segment, the position of the line segment is corrected, and the allowable pattern deformation amount is set to a value different from that of the other line segments. Here, the position correction amount and the allowable pattern deformation amount may be varied according to the proximity distance. For example, it is possible to use the correction amount of the position of the line segment of the H portion smaller than the correction amount of the separation line segment obtained above and not correct the position of the line segment of the J portion.

  According to the present embodiment, it is possible to reduce the effect of the above-described phenomenon by correcting the position of the line segment of the reference pattern or setting the allowable pattern deformation amount.

5.6 Separation Method of Pattern Deformation Amount into Global Pattern Deformation Amount and Local Pattern Deformation Amount In the above-described 4.8 pattern deformation amount obtained from the entire inspection unit region, the pattern deformation amount is obtained for each inspection unit region. According to this method, when a pattern is formed with a width different from the line width of the design data globally due to a difference in pattern formation conditions, the pattern deformation obtained from the entire inspection unit area obtained from all the inspection unit areas The amount will have a large value. However, the local change in the line width often limits the characteristics of the semiconductor device rather than the global change in the average line width. Therefore, there is a demand for evaluating the characteristics of a semiconductor device by separating the pattern deformation amount into a global pattern deformation amount and a local pattern deformation amount.

  97 to 100 are diagrams schematically illustrating a method of separating the pattern deformation amount into the global pattern deformation amount and the local pattern deformation amount. Here, the deformation amount of the line width is used as the pattern deformation amount. FIG. 97 is a schematic diagram showing an example in which a pattern is formed with a width that is globally different from the line width of the design data due to a difference in pattern formation conditions.

  As shown in FIG. 97, the central portion of the semiconductor device has a globally normal line width and a defect K1. On the other hand, the peripheral portion of the semiconductor device is globally thick in the X direction. It is assumed that the amount of deformation and the size of the defect K1 in the X direction are the same amount M. In this case, the defect K1 is desired to be recognized as a defect, but the global line width deformation amount M is not desired to be recognized as a defect. In addition, when the deformation amount M of the global line width is recognized as a defect, the number of defects to be registered becomes enormous.

  In order to solve this problem, a method of correcting the global average line width deformation amount in the line width of the design data before recognizing the defect can be used. In order to correct the line width of the design data, a first method for obtaining a global average line width change amount of the line width using the inspected inspection unit area, and a global obtained by the first method And a second method for correcting the line width of the design data by using the deformation amount of the average line width. Here, the global average line width may be obtained from a sufficiently wide range even if it is not the entire semiconductor device.

98 (a), 98 (b), and 98 (c) are diagrams illustrating an example of a first method for obtaining a global average line width change amount of a line width by using an inspected inspection unit region. . As the global average line width deformation amount, the X-direction line width deformation amount used in the description of the pattern deformation amount obtained from the whole 4.8 inspection unit region is used.
First, when the deformation amount of the line width in the X direction is obtained from each inspection region (see FIGS. 67 (a) and 67 (b)), as shown in FIG. 98 (a), the deformation amount C _X of the line width is obtained. Is obtained. The amount of deformation of the line width in the Y direction can be obtained similarly. If necessary, the deformation amount of the line width in the 45 degree direction and the 135 degree direction may be obtained.

Then, the amount of deformation of the global average line width in the X direction in order to obtain a <C _X>, the average value of the deformation amount C _X of the line width (<> represents the average value). For example, as shown in FIG. 98 (b), the inspection unit area (indicated by a dot in the figure) to be corrected is in the vicinity of the inspection unit area and has already been inspected (in the figure, in a lattice shape). The method of calculating the average value of each _CX of (part) can be used.

Further, in the sequential inspection described in the above-described 3.4 inspection unit region , as shown in FIG. 98 (c), the above-described 3.3 for the inspection unit region to be corrected (the portion indicated by a dot in the figure). Each C _{X of} the inspection unit region (grid-shaped portion in the figure) immediately inspected for the recipe data “(8) Number of inspection unit regions whose average value is to be obtained in order to obtain the global deformation amount of the pattern” The average value can be used. As another method, a moving average value can also be used. The deformation amounts <C _Y >, <C ₄₅ >, and <C ₁₃₅ > of the global average line width in the Y direction, 45 degree direction, and 135 degree direction can be obtained in the same manner. The global average linewidth deformations <C _X >, <C _Y >, <C ₄₅ >, and <C ₁₃₅ > obtained above are the global deformations of the pattern.

  The deformation amount of the global average line width may be distinguished into deformation amounts obtained from the respective groups having the same pattern attribute. This distinction can be distinguished from the proximity line segment and the separation line segment in addition to them, or according to the line width. Further, instead of distinguishing according to the line width, a global average line width deformation amount may be expressed in the form of a function.

FIG. 99 is a diagram showing a second method in which the global average line width deformation obtained by the first method is used for correcting the line width of the design data. Here, the global average line width deformation amounts <C _X >, <C _Y >, <C ₄₅ >, and <C ₁₃₅ > are used as the global average line width deformation amounts.

When using the obtained global average line width deformation <C _X >, <C _Y >, <C ₄₅ >, <C ₁₃₅ > for the correction of the line width of the design data, it is based on the design data. The same method as the size process (process for changing the line width) described in step S206 (see FIG. 19) for generating a pattern is performed. That is, each line segment of the reference pattern is moved by the deformation amount <C _X >, <C _Y >, <C ₄₅ >, <C ₁₃₅ > of the global average line width for each direction. This process is performed after the process described in step S304 (see FIG. 25), searching the recipe database 22 using the recipe inspection parameter as a key, and retrieving the recipe data including the reference pattern.

As an example of this processing, the result of performing size processing for the deformation amount <C _X > of the global average line width in the X direction in the X direction is shown by a double line in FIG. . The deformation amount <C _X > of the global average line width in the X direction in this example is equal to the deformation amount M.

Direction that was not calculated as the deformation amount of the global average line width, for example, the deformation amount of the global average line width in the direction of 30 degrees <C ₃₀ > is the deformation amount of the global average line width in the calculated direction. The amount of deformation of the global average line width in the X direction, the amount of deformation of the global average line width in the Y direction, and the amount of deformation of the global average line width in the 45 degree direction can be obtained by interpolation.

FIG. 100 shows the latter calculation example. A line segment indicated by a broken line in FIG. 100 schematically represents a reference pattern, and a line segment indicated by a solid line schematically represents an edge detected from the inspection target image. Deformation amount of global average line width in the X direction <C _X >, Deformation amount of global average line width in the Y direction <C _Y >, Deformation amount of global average line width in the 45 degree direction <C ₄₅ >, The definition of the direction of the deformation amount <C ₁₃₅ > of the global average line width in the direction of 135 degrees is based on this figure. On the right side of FIG. 100, the global average line width deformation amount <C _X > in the _X direction and the global average line width deformation amount <C ₄₅ > in the ₄₅ degree direction are calculated in the 30 degree direction. The deformation amount <C ₃₀ > of the average line width is shown.

  As a method different from the above method, the global average line width deformation amount is detected once at a specific position before inspection, and this global average line width deformation amount is detected in the design data of each location. The method used for the correction amount of the line width can be used.

  When the processing shown in FIG. 25 after step S304 (see FIG. 25) is performed, defect detection is recognized in step S320 (see FIG. 25). As described above, the semiconductor device of FIG. 97 has a pattern normally formed in the central portion, but has a thick pattern in the periphery. In the present example, when the size processing is performed so as to detect the defect K1, the defect K2 is not detected, and most of the thickly formed pattern is recognized as a defect. However, if this embodiment is used, only the defect K2 can be recognized as a defect as shown in FIG.

  As a result, it is possible to output an inspection result separated into a global average line width deformation amount as a global pattern deformation amount and defect information as a local pattern deformation amount.

When this embodiment is used, it is necessary to cancel the change in defect information due to the correction of the line width of the design data. Specifically, the above-described global average line width deformation amount is added to the line width deformation amount described with respect to the pattern deformation amount obtained from the entire 4.8 inspection unit region .

  According to the present embodiment, the number of acquired defects can be reduced by separating the pattern deformation amount into the global pattern deformation amount and the local pattern deformation amount. As a result, it becomes possible to sufficiently detect important defects and reduce detection of pseudo defects. Here, the pseudo defect is a defect that does not need to be regarded as a defect.

5.6.1 Method for correcting temporal fluctuation of line width measurement value In long-time inspection, the beam diameter may fluctuate gradually over time. As the beam diameter increases, the measured line width increases by the increased amount. This variation is added to the above-described global average line width deformation amount. Therefore, it is necessary to offset the temporal variation to the above-described global average line width deformation amount.

  FIG. 101 is a schematic diagram showing variations in beam diameter in a simplified version of FIG. In FIG. 101, the beam diameter is gradually increased, but the deformation amount of the beam diameter can be ignored during the inspection of the inspection unit area for one row. Such a change in the measurement value of the line width due to the variation in the beam diameter is corrected by the following procedure.

  First, the inspection unit area to be inspected twice is determined by the method shown in FIG. The inspection area to be inspected twice is set for each time range in which the change amount of the beam diameter can be ignored. In FIG. 101, since the inspection time of the inspection unit area for one row corresponds to this time range, the inspection unit area to be inspected twice is set as shown in FIG.

Next, as shown in FIG. 102, the inspection unit area to be inspected twice is inspected, and the above-described global average line width deformation amounts <C _X >, <C _Y >, <C ₄₅ >, and < _Find C ₁₃₅ >. The calculation of the global average line width deformation amounts <C _X >, <C _Y >, <C ₄₅ >, and <C ₁₃₅ > is the same. Therefore, in the present embodiment, description will be made using the deformation amount <C _X > of the global average line width in the X direction. As indicated by <C _X > _1,1 and <C _X > _1,11 in FIG. 102, the first inspection and the number of the inspection unit area are expressed by subscripts. These global average line width deformation amounts express the line width deformation amount depending on the location, and the time-dependent line width deformation amount can be ignored.

After the above first inspection, as shown in FIG. 103, the entire inspection unit region is inspected as the second inspection, and the above-described global average line width deformation amount <C _X >, Y-direction global average line width deformation <C _Y >, 45-degree global average line width deformation <C ₄₅ >, 135-degree global average line-width deformation <C ₄₅ > ₁₃₅ > As indicated by <C _X > _2,1 and <C _X > _2,11 in FIG. 103, the second inspection and the number of the inspection unit area are expressed by subscripts.

Compensation from global average line width deformation <C _X > _2,1 and global average line width deformation <C _X > _1,1 when inspection No. 1 in the inspection unit area is completed The quantity δ <C _X > ₁ is determined by the following formula.
The obtained correction amount Δ <C _X > ₁ is regarded as a correction amount for the amount of fluctuation of the line width depending on time.

Δ <C _X > ₁ is added to the global average line width deformation <C _X > obtained from No. 2 to No. 10 in the inspection unit area to correct the time-dependent line width variation. To do. The number 10 of the inspection unit area means the number immediately before the number 11 of the inspection unit area to be inspected twice next time.
The above procedure is similarly performed for the above-described global average line width deformation amounts <C _Y >, <C ₄₅ >, and <C ₁₃₅ >. In addition, these procedures are performed for all inspection unit areas to be inspected twice.

  When a pattern created with an ArF resist is inspected many times with a scanning electron microscope, the pattern gradually shrinks. However, according to this embodiment, since the same place is inspected only twice, this pattern shrinkage can be ignored. Therefore, even when the pattern is inspected, it is possible to correct the variation in the measured value of the line width due to the gradual variation in the beam diameter.

5.7 Defect Type Using Geometric Information of Reference Pattern, Design Data Information, or Data Information Related to Design Data As described above in 4.7 Defect Types Using Features Obtained from Images The defect type is determined by the defect type recognition unit 14 using the feature amount of the defect image. In addition to this defect type, a defect type determined using the geometric information of the reference pattern, the design data information, or the data information related to the design data can be used.

The following items are used as geometric information of design data.
(1) Pattern attributes (straight line, corner, end, isolated pattern, etc.)
(2) Proximity line segment, separation line segment and other line segments (3) Line width (Example: minimum line width, minimum line width exceeding minimum line width x less than 1.5, minimum line width x 1 .5 or more line width)

The following items are used as design data information.
(4) Location where the defect is detected (eg memory part, logic part, etc.)
(5) The cell name of the design data corresponding to the defect. As additional information, the line number constituting the cell corresponding to the defect, or the position of the defect in the coordinate system describing the cell.
(6) Wiring attributes (such as ground wiring and clock wiring), but can be used when these attributes are defined in the design data.

The following items are used as data information related to design data. Here, mask data is used as data related to design data.
(7) The cell name of the mask data corresponding to the defect. As additional information, the line number constituting the cell corresponding to the defect, or the position of the defect in the coordinate system describing the cell.
In addition, the following items can be used as defect types using the pattern deformation amount.
(8) Defect size information (e.g., large, medium, thin, large, medium, thin)

  FIG. 104 shows an example of a method for classifying each cell name and line segment number shown in (5) above. Two T-shaped patterns A and B represent one memory cell having the same cell name. The end portion surrounded by a circle has the same shape as the reference pattern but a different OPC pattern. In this case, the defect A and the defect B are defects generated by different OPC patterns. However, these defects cannot be classified only by classification by cell name. However, OPC patterns related to the occurrence of defects can be distinguished by classification based on line numbers.

  The aforementioned defect types can be used in combination. FIG. 105 schematically shows defect types used in combination. The defect type used here is a combination of three types (6), (1) and (8).

  If this method is used, the tendency for defects to occur can be easily grasped. In addition, the cause of the defect can be easily identified.

5.8 Grouping Method Using Reference Pattern near Defect Another defect classification is a grouping method using a reference pattern near a defect. 106 to 108 are diagrams showing the above-described method.
When a defect is detected, a reference pattern corresponding to the vicinity of the defect position is cut out and stored. At the end of the inspection, the feature amount is calculated from the extracted reference pattern, and the defects are grouped.

  FIG. 106 shows a defect position, a circumscribed rectangle, and a cut-out reference pattern. The defect position is the center of the circumscribed rectangle, and the cut out reference pattern corresponds to the vicinity of the defect position. As the feature amount, a set of the line width and direction of the linear pattern and the number of the patterns can be used. As others, a combination of the space width and direction and the number of spaces, a corner type and the number of sets, a termination type and the number of sets, an isolated pattern type and the number of sets, and the like can be used. Next, the feature space formed by these features is grouped by cluster analysis. Cluster analysis is one of the well-known classification methods in statistics.

  FIG. 107 is a schematic diagram showing an example of the feature amount space. In FIG. 107, four linear shape patterns in the 100 nm vertical direction, two linear shape patterns in the 200 nm vertical direction, and four linear shape patterns in the 100 nm horizontal direction are used as feature amounts. In this example, the three extracted reference patterns are clearly separated in the feature amount space. However, in actuality, the peripheral pattern is included or not included in the cut-out reference pattern due to the difference in the position where the defect exists, so that it is not always clearly separated. Therefore, cluster analysis to classify similar things is necessary.

  In order to group the cutout reference patterns more finely, it is necessary to subdivide the feature values. For example, in order to distinguish patterns as shown in FIG. 108, it is necessary to distinguish the linear polygons in the vertical direction or the short length.

  According to the present embodiment, grouping can be realized based on the characteristics of the reference pattern near the defect. As a result, it is possible to easily grasp the tendency of defects such as “there are many defects in a pattern that is complicated by thin vertical lines”. Furthermore, the defect can be classified for each reference pattern having the same shape. Furthermore, it becomes easy to identify the cause of the defect.

In the defect selection method step S328 to 5.9 image registered, defect-class determination unit 14 outputs to the display device 5 and the printer 6 through the output unit 13 the image. Here, when the number of defects becomes very large, the number of images to be registered becomes enormous and the amount of storage media increases, which is not suitable for practical use. Therefore, this problem is solved by having the maximum number of registered defect images for each defect type.

The maximum number of registered defect images for each defect type may be statically allocated to the above-mentioned 3.3 recipe data “(9) Maximum number of registered defect images” or dynamically. The number of detected defects may be monitored and made variable according to the number of detected defects. For example, these dynamic maximum registration numbers are determined as numbers proportional to the logarithm of the number of defects detected so far.

When a new defect image is detected, new defect images are registered until the maximum number of registered defect types to which the new defect image belongs is exceeded. When the number of already registered defect images has reached the maximum number of registrations, when registering a new defect image, it is determined whether or not to register a new defect image according to the defect size and other indices. . If it is determined to be registered, the defect image to be discarded is determined. As another method, it may be determined whether to register a new defect image using a random number.
If this method is used, even if there are very many defects of one type of defect and few defects of other types of defects, a larger number of types of images can be registered.

5.10 Method of selecting defect to be re-inspected In some cases, a defect image is re-acquired and re-inspected under high-magnification image acquisition conditions different from conditions such as magnification at the time of inspection. The re-examination is carried out according to the following procedure.

(1) The recipe registration process described in FIG. 19 is executed.
In step S <b> 202 of FIG. 19, the operator inputs operator input parameters to the reference pattern generation unit 11 via the input device 4. When re-inspecting, operator input parameters for re-inspection are input during step S202. The inspection area which is one of the image acquisition parameters among the operator input parameters for re-inspection is not input because it is determined in (4) described later.

(2) The inspection process described in FIG. 25 or 26 is executed.
(3) A defect to be reinspected is automatically selected from the detected defects.

(4) The recipe registration process described in FIG. 19 is executed.
Here, instead of executing step S202, the operator input parameters for reexamination input in the above (1) are input to the reference pattern generation unit 11. The inspection area is automatically set as a square area centered on the defect to be reinspected as an inspection area for random inspection.
(5) As the re-inspection, the inspection process described in FIG. 25 or 26 is executed.

  As described above, it is necessary to automatically select a defect to be reinspected from defects detected before reinspection. A defect to be reinspected may be selected by simply thinning out the detected defect. However, defects with the same defect type detected more frequently are not necessarily more important, and there are cases where it is desirable to sufficiently reinspect defects with the same defect type detected less. In order to meet such requirements, the maximum number of registered defects to be reinspected is determined for each defect type.

These defect numbers are statically used in the above-mentioned 3.3 Recipe data “(”, as used in the distribution of the maximum number of registered defect images described in the above-described defect selection method of 5.9 image registration target. 10) “Maximum number of registered defects to be reinspected” may be allocated, or the number of defects detected so far may be dynamically monitored and varied according to the number of defects already detected. . For example, these dynamic maximum registration numbers are determined as numbers having the same ratio as the logarithm of the number of defects detected so far.

It is determined according to the random number whether it is a defect to be reinspected after inspection or not. That is, random numbers are given to all detected defects, and it is determined that a defect having a larger random number is more important. If a larger defect is to be re-inspected, a random number weighted according to the defect size information may be used. In addition to the defect size information, weighting may be performed using another index.
By using this method, it is possible to sufficiently set a defect having the same defect type detected a little and a defect having the same defect type detected a lot to be reinspected.

5.11 Method for Displaying Distribution of Pattern Deformation of Whole Semiconductor Device As described in steps S328 and 332 in FIG. 25, the inspection result is output to the output unit 13. When the output unit 13 outputs the inspection result as a numerical value, it is difficult to grasp the tendency of occurrence of defects in the entire semiconductor device. As a countermeasure against this, there is a need for a method in which the output unit 13 creates a distribution map represented by the next bitmap and outputs it to the display device 5 and the printing device 6 at the end of the inspection. This distribution map is obtained by converting the pattern deformation amount obtained from the whole 4.8 inspection unit area described above into pixel shading or pseudo color display information and overwriting the defect.

  FIG. 109 is an example of a distribution diagram obtained by converting a line width deformation amount, which is one of the pattern deformation amounts obtained from the entire inspection unit area, into information for grayscale display and overwriting defects. The lattice portion has the maximum line width deformation amount, the dot portion has a larger line width deformation amount, and the space portion has a normal line width deformation amount. The black square represents a defect. From FIG. 109, it can be seen that more defects are generated in the portion having the larger deformation amount of the line width.

  Further, when the deformation amount of the line width is displayed as shown in FIG. 109, it is possible to visually understand the tendency of the deformation of the pattern due to the distortion of the stepper or the position of the wafer. For example, when a distribution diagram is observed by inspecting a semiconductor device having a periodic pattern that is normally formed, a tendency that the line width around the distribution diagram is thicker than the center is shown. From this tendency, it can be seen that there is distortion in the peripheral portion of the stepper. As another example, when a normally formed SoC is inspected and a distribution diagram is observed, it can be seen that the line width differs for each functional block such as a memory and a logic.

Further, the quality of the semiconductor device can be verified by using a line width variation index which is one of the pattern deformation amounts obtained from the entire inspection unit region.
According to the present embodiment, since the tendency of occurrence of defects in the entire semiconductor device can be visually grasped, it can be used for identification of the cause of occurrence of defects and quality verification of semiconductor devices.

6). Other Scanning Method of Image Generating Device In addition to the scanning method described in 2.2 Scanning Method of Image Generating Device, the scanning method described below can be used in image generating device 7.

6.1 Scanning method of electron beam in 18 degree direction, scanning method of hexagonal block, automatic setting method of scanning condition based on reference pattern FIGS. 110A and 110B scan an electron beam in 18 degree direction. It is a schematic diagram which shows the method to do. The patterns P1 and P2 shown in FIG. 110 (a) are the same as those shown in FIG. 8 (a). A pattern of a semiconductor integrated circuit (LSI) or a liquid crystal panel is constituted by 99% or more of a vertical line, a horizontal line, or a 45 degree right-down and 45 degree right-up direction. The scanning direction of 18 degrees shown in FIG. 110 (b) can be used as an optimal scanning direction for crossing the edge direction and scanning direction of the pattern to be inspected with respect to all the directions at a certain angle. By setting the scanning direction to 18 degrees, it is expected that relatively good inspection accuracy can be obtained for horizontal lines, vertical lines, and 45-degree line segments.

  The angle 18 degrees may be any other angle as long as it is perpendicular to all the patterns to be inspected. For example, 22.5 degrees, 63 degrees obtained by adding 45 degrees to 18 degrees, 108 degrees obtained by adding 90 degrees to 18 degrees, and the like can be used.

  111 (a) to 111 (d) are diagrams schematically illustrating scanning of a hexagonal block. In a scanning electron microscope such as a normal CD-SEM, it is common to take a scanning direction in the horizontal direction and take a square image. However, due to the design limitations of the scanning electron microscope, the area that can be scanned without distortion is a perfect circle area. Therefore, scanning is performed using a square block 401 in a round circle 400 as shown in FIG. In this case, there are areas that can be scanned without distortion in the vertical, horizontal, left, and right portions, but there are areas that are not scanned, and a little waste is generated in order to take a larger area at one time. In such a case, if a wide area is taken while being overlapped, a scanning area in which nine rectangular blocks B1 to B9 are overlapped as shown in FIG. 111 (b) is generated.

  On the other hand, as shown in the lower side of FIG. 111 (c), the region acquired by one scan is changed from a rectangular block to a hexagonal block 402, so that the shape can be made closer to a circle and used for scanning. It is possible to take a wider part. As a scanning method, a hexagonal portion is scanned as shown on the left side of FIG. 111 (c), and a rectangle is scanned as shown on the right side of FIG. 111 (c).・ Two methods can be used to avoid using the triangles in the lower left for measurement. With this method, as shown in FIG. 111 (d), it is possible to acquire a wider area with a small number of scans (blocks B1 to B7).

  FIG. 112 is a schematic diagram showing a method for automatically setting scanning conditions based on a reference pattern. As in the description of the scanning directions of 0 degrees and 90 degrees, it is necessary to automatically determine conditions such as whether to scan once or twice according to the reference pattern. The following three methods can be used for automatically determining the electron beam scanning method.

(1) As shown in the block (D), when there is no pattern to be inspected in the scanning area, a method of not scanning the block. (2) A method of determining the scanning condition according to the line width of the pattern. When the pattern Pa in (A) and the pattern Pb in the block (B) are compared, the line width of the pattern Pb is twice that of the pattern Pa. In this example, in order to detect the variation rate of the line width of the pattern, an image is acquired at a magnification of 1/2 in the scanning in the block (B) with respect to the scanning in the block (A). Can do.

(3) A method of determining the scanning direction condition in accordance with the distribution direction of the reference pattern For example, for the block (A), since the pattern Pa has vertical and horizontal line segments, scanning at 45 degrees is performed. It only needs to be done once. For the block (C), since the pattern Pc has two line segments in the 45 degree direction and the 135 degree direction, it is necessary to perform two scans in the 45 degree direction and the 135 degree direction.

  If this embodiment is used, in order to obtain an image of the inspection target pattern, it is only necessary to scan a minimum number of electron beams (charged particle beams), and therefore, an image of the inspection target pattern can be obtained in a minimum time. In addition, a wide range of blocks is realized with the smallest possible number of blocks by making the best use of the scannable area. Furthermore, an image can be acquired under optimum conditions using a reference pattern in order to prevent a decrease in edge detection accuracy depending on the scanning direction.

6.2 Scanning Path of Electron Beam in Image Generating Apparatus FIG. 113 and FIG. 114 are schematic diagrams for explaining the scanning path of the electron beam in the image generating apparatus 7. The oscillator 410, the counter 411, the X deflection generation circuit 412, and the Y deflection generation circuit 413 are circuits constituting the deflection control device 318. The control computer 350 sets the start voltage, end voltage, and step voltage in the X deflection generation circuit 412 and the Y deflection generation circuit 413. Further, the control computer 350 outputs a start signal to the oscillator 410.

  In normal scanning, the pattern is scanned by stepwise deflection in the X direction for each pixel. Further, scanning is performed by deflecting stepwise in the Y direction for each line. However, in such a conventional method, since the information between the scanning lines cannot be acquired, the inspection accuracy tends to be lowered. As shown in FIG. 113, in the present invention, in order to be able to acquire information between scanning lines, a signal having an amplitude like a sine wave is added to the Y deflection so that the distance between the scanning lines can be obtained. Data is taken (see the lower left part of FIG. 113).

  Here, four points of data are sampled as shown (see the lower right portion of FIG. 113). In this case, data spreading about Y deflection can be acquired during one sine cycle. The four points of data are added and transferred to the control computer 350 as information of one pixel.

  As shown in the upper part of FIG. 113, an oscillator 410 having an internal frequency that is four times the output frequency is connected to the counter 411. The counter 411 is connected to the X deflection generation circuit 412 and the Y deflection generation circuit 413. With such a configuration, a stepped right-up waveform is generated for the X deflection and a sine wave is generated for the Y deflection by using the internal frequency clock. Four points of data are sampled at an internal frequency, and the four points of data are added to generate sampling data corresponding to an actual pixel.

  As another method, as shown in FIG. 114, a zigzag scanning path may be formed by generating a Y-deflection waveform and a step-like waveform in the X-deflection according to the above-described method.

  FIG. 115 is a schematic diagram for explaining a case where a filter is applied to scanning in the vertical direction. A is a pixel adjacent in the horizontal direction, and is smoothed by a detector and an amplifier. On the other hand, B is close in the vertical direction but is not smoothed as described above. Therefore, a smoothing filter is applied in the vertical direction to reduce the difference in image quality between the vertical direction and the horizontal direction. FIG. 115 shows the simplest filter coefficient, but it is appropriately selected so as to match the frequency characteristic in the horizontal direction.

  According to the present embodiment, the difference in image quality between the X direction and the Y direction can be reduced as much as possible by a method of obtaining data between scanning lines by changing the scanning path or a method of applying a filter.

6.3 Edge How to shorten the image acquisition time by such scans only vicinity of the scanning method edge only vicinity of is required. Further, there is a need for a method for improving edge detection accuracy by scanning an electron beam perpendicular to the edge direction.

  FIG. 116 is a schematic diagram showing a scanning method only in the vicinity of an edge, and FIG. 117 is a flowchart. In the example shown in FIG. 116, a sub deflection generation circuit 450 is provided.

A method for realizing scanning only in the vicinity of the edge is performed by the following procedure.
(1) The profile acquisition section used for detecting the second edge is obtained from the reference pattern, and the position of information related to the profile acquisition section is registered in advance. This information has the position, direction and length of the center point of the profile acquisition section.
(2) The control computer 350 takes in information related to one profile acquisition section.
(3) The position of the center point of the profile acquisition section is set in the X main deflection generation circuit 452 and the Y main deflection generation circuit 453. Thereby, the center position of the beam moves.

(4) The rotation angle corresponding to the direction of the profile acquisition section is set in the rotation circuit 451, and the amplitude corresponding to the length of the profile acquisition section is set in the sub deflection generation circuit 450.
(5) A start signal is supplied to the oscillator 410, and the counter 411 connected to the oscillator 410 forms scanning waveforms in the X direction and the Y direction. By adding the outputs of the X main deflection generation circuit 452 and the Y main deflection generation circuit 453 to this, a scanning path as shown in the upper center portion of FIG. 116 is created.
(6) In this scanning path, seven points of sampling as shown on the upper right side of FIG. 116 are performed to obtain sampling data.

  FIG. 118 is a diagram showing a method of ordering measurement data acquisition when only the vicinity of an edge is scanned. As shown in FIG. 118 (a), sampling data is sampled while skipping measurement points at a given thinning rate, and as shown in FIG. 118 (b), sampling data is randomly sampled using random numbers or the like. is there. According to the present embodiment, profile deformation due to the charging phenomenon of the sample can be reduced, which is suitable for measurement of an insulator. When the charging phenomenon of the sample can be ignored, the sampling may be sequentially performed so as to make a round of the reference pattern.

  According to the present embodiment, not only high-speed and accurate scanning can be realized, but also the influence of the charging phenomenon of the sample can be reduced.

6.4 Scanning method for only the vicinity of the region to be subjected to the region inspection When the 5.2 region inspection method is used, the image acquisition time is determined by the method for scanning only the vicinity of the region to be subjected to the region inspection. Can be shortened. In addition, since the scanning direction and the edge direction can be orthogonal to each other, the edge detection accuracy can be improved.

For example, when only a reference pattern suitable for line width inspection or space width inspection is inspected, a method of scanning a rectangular portion obtained below can be used. This rectangular portion is the vicinity of the region to be subjected to the region inspection, and is obtained by the following procedure using FIGS. 119 (a), 119 (b), and 119 (c).
(1) A reference pattern K suitable for line width inspection is obtained. The reference pattern K is a rectangle indicated by a solid line and a double line in FIG. 119 (a), and is the same as FIG.
(2) Based on the reference pattern K, a minimum rectangle including profile acquisition sections necessary for all edges to be detected is obtained as a scanning portion.

Specifically, the minimum rectangle R including the reference pattern K is obtained. Profile acquisition sections are set for the right and left line segments of the rectangle R, respectively.
FIG. 119 (b) or (c) can be used as the scanning portion. In FIG. 119 (b), rectangles Sa and Sb in which the right and left line segments of the rectangle R are expanded to both by the length L of the profile acquisition section are scanning portions. The scanning portion has four arrows written in a rectangle, and the arrows indicate the direction of scanning.

In FIG. 119 (c), the rectangle R is inflated to the left and right by the length L of the profile acquisition section is the scanning portion Sc. This method has an advantage that both the left and right edges cannot be scanned from the inside to the outside of the pattern, but only one scanning area is required.
The scanning area can be similarly determined for the space width inspection.

  According to the present embodiment, the image acquisition time can be shortened. In addition, since the scanning direction and the edge direction can be orthogonal to each other, the edge detection accuracy can be improved.

7). Method of correcting pattern image to be inspected
7.1 Immediately after the image is acquired, the inspection target pattern image may rotate due to the rotation of the sample due to movement of at least one correction method stage among the reference pattern and the image by detecting the distortion amount of the inspection target pattern image. In addition, deformation such as rotation including skew and change in magnification may occur due to a charging phenomenon. Due to these influences, it is not possible to detect fine defects below the above-mentioned strain amount. This distortion occurs discontinuously in time and is difficult to predict. As a countermeasure against this phenomenon, a method for detecting and correcting the distortion amount of the pattern image to be inspected every time an image is acquired is necessary.

  FIG. 120 schematically shows an image of the inspection unit region having the above-described distortion. A line segment indicated by a dotted line schematically represents a reference pattern, and an edge where the head of a vector d (x, y) between edges is detected is schematically represented. The reference pattern and the edge are matched. However, matching only deals with translation. In matching, distortions such as rotation and magnification change remain as errors.

First, the matching errors are totaled using the following affine transformation. The affine transformation means a primary transformation using coefficients a to f.

In this conversion formula, (x, y) is the coordinate value of a point in the reference pattern, and (X, Y) is the coordinate value of the detected edge corresponding to the aforementioned point. The coefficients a, b, d, and e express the difference between the rotation including the skew and the magnification. If it is not necessary to correct the skew, a restriction is imposed so that the next matrix becomes an orthogonal matrix.
Further, if there is no need to correct the magnification, a restriction is imposed so that this matrix becomes a rotation matrix.

In this conversion formula, coefficients c and f represent shift amounts. In the example of FIG. 120, this term is zero.
FIG. 121 shows a method for performing matching in the upper right sub-inspection unit region. The sub-inspection unit area is defined as an area obtained by dividing the inspection unit area. When the inspection unit area is large, matching using the sub-inspection unit area is significantly faster than matching using the entire inspection unit area. In this case, the coefficients c and f of the sub-inspection unit areas other than the upper right generally do not become zero.

The coefficients a to f are obtained by the following procedure.
(1) As shown in FIG. 120, a vector d (x, y) representing the sum of the pattern deformation amount and the pattern distortion amount is obtained. The vector d (x, y) is the same as the vector d (x, y) between the edges in FIG.
(2) The coefficient of each vector d (x, y) (d _x (x _i , y _i ), d _y (x _i , y _i )) (where i is the number of data from 1) using the least square method Find f from a. (x _i , y _i ) is the coordinate value of the reference pattern point. The position coordinate value of the detected edge corresponding to this point is (x _i + d _x (x _i , y _i ), y _i + d _y (x _i , y _i )). Accordingly, the square sum E of errors is calculated as follows.
Here, Σ represents the total sum for all i.

The least square method requires that the partial differentiation of the error sum of squares E by the coefficients a, b, c, d, e, and f is zero.
From these equations
Get. The solution can be obtained by solving the above simultaneous linear equations. The above processing is executed between step S314 and step S318 in the flowcharts shown in FIGS. Hereinafter, this process is referred to as step S316.

As shown in FIG. 122, there are three correction methods using the coefficients a, b, c, d, e, and f obtained by this method.
(1) Distortion correction method 1
The reference pattern is corrected using the coefficients a, b, c, d, e, and f. Next, the process after step S314 is performed. However, step 316 is not executed.

(2) Distortion correction method 2
The image is corrected using the coefficients a, b, c, d, e, and f. Next, the process after step S310 is performed. However, step S316 is not executed. In this case, an inverse transformation formula of the formula described in the affine transformation is used.

(3) Distortion correction method 3
(D _x (x _i , y _i ), d _y (x _i , y _i )) representing the components of each vector d (x, y) are expressed below using coefficients a, b, c, d, e, and f. Correct with the following formula.
Next, step S318 is skipped and the processing after step S320 is executed. In step S320, each corrected vector d (x, y) obtained above is used as a vector d (x, y) between edges in FIG.

  Although the distortion correction method 1 and the distortion correction method 2 can accurately correct the distortion amount of the pattern image to be inspected, the calculation amount is large. On the other hand, the distortion correction method 3 is slightly inaccurate in the corner portion, but the calculation amount is small. However, this error is often negligible.

In this embodiment, a method using affine transformation is used, but other transformation formulas may be used. For example, a conversion formula using a quadratic term of x _i and y _i can be used. However, it should be noted that a more complicated conversion formula ignores the actual pattern deformation amount.

  According to the present embodiment, it is possible to detect and correct the linear amount of distortion of the inspection target pattern image. As a result, the amount of distortion that does not need to be recognized as a defect can be ignored, and the occurrence of a pseudo defect can be prevented.

  The above-described rotation and magnification detection method including skew can be used for rotation and magnification adjustment of the image generation device 7. This adjustment is performed at an appropriate time before or during the inspection in the above-described inspection method. In this case, the affine transformation coefficients a, b, d, and e are converted into adjustment values for adjusting the rotation and magnification of the image generating device 7, and these adjustment values are set.

7.2 Nonlinear Image Distortion Correction Method The image generation apparatus 7 having a wide field of view may have image distortion that cannot be adjusted by the affine transformation described above. These image distortions are caused by Seidel's five aberrations. One of the most important image distortions is nonlinear image distortion. In order to correct this nonlinear image distortion, the adjustment method shown in FIGS. 123 to 128 can be used.

  As shown in FIG. 123, the non-linear image distortion can be ignored in the central portion of the image, but is remarkable in the peripheral portion of the image. This method is a method of obtaining a distortion vector from an image, converting the obtained distortion vector into a representative distortion vector, and calculating a distortion correction vector at each scanning point using the obtained representative distortion vector. The obtained distortion correction vector is used in the deflection control device 318 shown in FIG. FIG. 124 is the same as FIG. 113 except that a distortion correction vector calculation circuit 414 is added. The distortion correction vector calculation circuit 414 obtains a representative distortion vector from the control computer 350, calculates a distortion correction vector in synchronization with the signal from the counter 411, and supplies the XY component to the X deflection generation circuit 412 and the Y deflection generation circuit 413. Output.

FIG. 125 shows a method in which the distortion correction vector calculation circuit 414 calculates a distortion correction vector using the representative distortion vector. In FIG. 125, the X deflection voltage and the Y deflection voltage are used as the XY coordinate system. 125 indicate the deflection voltages corresponding to the respective scanning points. The representative distortion vector is the position of the black circle (●) for each of the above-mentioned 3.3 recipe data “(12) Interval of representative distortion vector possessed by distortion correction circuit” (interval in units of step voltage between scanning points) in both XY directions. Is set to Here, for simplicity, a method of using the same interval 8 in the XY directions will be described.

In order to calculate the distortion correction vector C _d (x, y) of the scanning point in FIG. 125, representative distortion vectors R _d ^[0,0] , R _d ^[8,0] , R _d ^[0,8] , R _The following bilinear interpolation formula using _d ^[8,8] is used.
Here, (x, y) is the coordinates of the scanning point determined by the counter 411. Further, the variable with the subscript ^{[x, y]} means that the variable is related to the coordinates (x, y) of the scanning point.

  A method for calculating the representative distortion vector from the distortion vector will be described. First, the reference pattern and the detected edge are matched at the center of the image. Here, the reference pattern suitable for nonlinear image distortion correction is a simple pattern that periodically exists, such as the reference pattern indicated by the dotted line in FIG.

  FIG. 126 (a) shows a method of calculating the representative strain vector of each vertex of the rectangular area from the strain vector of the point inside the rectangular area. The above bilinear interpolation is used as a calculation method. FIG. 126A is the same as FIG. 125 except that a distortion vector d (x, y) is described instead of the distortion correction vector at the scanning point. The distortion vector d (x, y) is defined as a vector whose start point is a point on the reference pattern and whose end point is a detected edge corresponding to the point on the reference pattern. The distortion vector d (x, y) is the same as the vector d (x, y) between the edges in FIG.

The calculation method of the X component and the Y component of the distortion vector d (x, y) at the point (x, y) is the same. Therefore, the X component or Y component of the distortion vector d (x, y) will be described by z (x, y). Similarly, the X component or the Y component of the representative distortion vectors R _d ^[0,0] , R _d ^[8,0] , R _d ^[0,8] , R _d ^[8,8] are represented by r _z ^{[0, 0]} , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] . Therefore, z (x, y) is expressed by the following equation.

r _z ^[0,0] , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] is ^converted to a sufficiently large number of data (x _i , y _i , z _i ) using the least squares method. x _i and y _i represent the coordinate values of the scanning points, and z _i represents the X component or Y component of the obtained distortion vector d (x, y). The square sum E of errors is calculated as follows.
Where Σ ^[0,0] is all data existing in the rectangular region (P _s ^[0,0] , P _s ^[8,0] , P _s ^[0,8] , P _s ^[8,8] ) Is the sum of Subscript ^[0,0] means subscript ^[0,0] with the point of the lower left corner of the rectangular area P _s ^[0,0].

The least squares method has zero partial differentiation of the error sum of squares E by r _z ^[0,0] , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] Request.
From these equations
Get. Here, the following symbols were used.
The solution can be obtained by solving the above simultaneous linear equations.

As shown in FIG. 126 (b), the synthesized distortion vector obtained by synthesizing the distortion vector d (x, y) obtained from the X component and the distortion vector d (x, y) obtained from the Y component. Does not exactly match the distortion vector at this location. In order to reduce this error, the above embodiment may be repeated. The aforementioned bilinear interpolation equation (again shown below) is linear with respect to r _z ^[0,0] , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] . Therefore, the values of r _z ^[0,0] , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] obtained in the second and subsequent iterations are calculated before this calculation. Add to the values of r _z ^[0,0] , r _z ^[8,0] , r _z ^[0,8] , r _z ^[8,8] . The value obtained above is used for correction.

To extend the method to use a plurality of rectangular areas, the above calculation may be applied to a plurality of rectangular areas. This method is represented by four rectangular regions (P _s ^[0,0] , P _s ^[8,0] , P _s ^[16,0] , P _s ^[0,8] , P _s ^[8,8] in FIG. 127 ^{). _{, P s [16,8], P}} s [0,16], P s [8,16], P s [16,16])
First, the previous formula is simplified and expressed.

Here, the following symbols were used.
Here, the subscript ⁰⁰ means the subscript of Σ ^[0,0] .

The square sum E of errors for the data of the four rectangular areas is calculated as follows.

From these equations
Get. The solution can be obtained by solving the above simultaneous linear equations.

Next, the variation of the magnification of the image in the X direction and the Y direction is obtained from (x _i , y _i , z _i ) which is data in the vicinity of the representative distortion vector. Here, a method for _obtaining the variation a _mag of the magnification in the X direction is shown. A portion having a magnification variation a _mag of 1 is observed at the same magnification as the central portion without distortion, and a portion having a magnification variation a _mag or a value greater than 1 is observed at a higher magnification than the central portion without distortion. Yes. X component of z _i distortion vector d (x, y), the following equation is satisfied when the shift amount and S _c.

When the number of data is 3 or more, it is solved by the least square method. The sum of squared error E is
It is. The least squares method requires the following equation:

Solving the above equation gives:

From the above, the following equation is obtained.
Here, N represents the number of data. The obtained magnification variation a _mag is multiplied by the representative strain vector to correct the magnification variation.

  128A and 128B show a method in which the distortion correction vector calculation circuit 414 converts the distortion correction vector into a deflection voltage. 128 (a) and 128 (b) use the position in the X direction where the electron beam is scanned on the vertical axis. The origin of the vertical axis represents the center of the image. The upper part of the vertical axis represents the peripheral part of the image. The horizontal axis uses the voltage generated by the X deflection generating circuit. The scale on the horizontal axis represents the voltage applied to the scanning point.

The step-like waveform indicated by the dotted line in FIG. 128A means an ideal scanning waveform, and the step-like waveform indicated by the solid line means a scanning waveform when there is distortion. A vector whose starting point is the position of the dotted line on the scanning point and whose end point is the position of the corresponding solid line on the scanning point means the X component r _z ^[8,0] of the representative distortion vector R _d ^[8,0]. . In order to simplify the illustration, the representative distortion vector R _d ^[0,0] at the origin position is set to zero.

In order to correct the distortion, the incremental voltage E _dX ^[8,0] is added to the voltage of the step voltage interval applied to the scanning point. The incremental voltage E _dX ^[8,0] is obtained by the following equation.
Here, the coefficient a _mag a magnification a _mag described above, E _s is a step voltage required to move the one scanning point in portion without distortion.

If z _i is used as the Y component of the distortion vector d (x _i , y _i ) and y _i is used instead of x _i , the nonlinear image distortion correction method in the Y direction is similarly performed.
In this embodiment, a distortion correction circuit is added to the image generation apparatus 7. However, the image generation apparatus 7 may be replaced with a method of converting an image to eliminate distortion.

In this example, 8 was used as the interval. As this interval is smaller, the accuracy of the distortion correction vector is improved, but the accuracy of the representative distortion vector is deteriorated. Thus, the interval is a value that should be determined by experience.
If this method is used, the nonlinear image distortion correction of the image generating apparatus 7 having a wide field of view can be automatically performed in a short time with high accuracy. Therefore, it is possible to extend the field of view to a portion where this correction is possible.

7.3 Correction Method of Line Width Variation Depending on Pattern Image Position One of the most important image distortions is line width variation depending on the image position. This variation in line width is caused by the fact that the beam diameter of the electron beam varies depending on the position of the image. In order to correct this variation in line width, a method of correcting the line width distribution of the inspection target pattern image at the time of inspection using the line width distribution of the inspection target pattern image obtained in advance can be used.

  FIG. 129 is a diagram schematically showing this method. As indicated by the circles in FIG. 129, the beam diameter of the electron beam is thicker at the periphery of the image than at the center. As a result, the observed line width is larger in the peripheral portion than in the central portion. If the beam diameter of the electron beam is uniform in the sub-inspection unit region, the non-uniformity of the line width distribution is corrected by the following method.

(1) An inspection target pattern image is acquired in advance from a portion having a pattern with the same line width and uniform density of the reference pattern. This part is suitable for a part having a periodic pattern such as a memory.
(2) The amount of deformation of the line width is calculated from the edge detected from the reference pattern and the inspection target pattern image for each sub-inspection unit area.
(3) The non-uniformity of the line width distribution is corrected using the deformation amount of the line width obtained by the above calculation (2) at the time of inspection.

The calculation method and the correction method of the line width deformation amount in the above (2) and (3) use the method described in the section 5.6 Separating the pattern deformation amount into the global pattern deformation amount and the local pattern deformation amount .

The above correction method may be applied to the above processing for each pattern width. Further, the above correction method may be applied to an area obtained by dividing the sub-inspection unit area.
If this method is used, the variation of the line width depending on the image position of the image generating device 7 having a wide field of view can be automatically performed in a short time with high accuracy. Therefore, it is possible to extend the field of view to a portion where this correction is possible.

8). Other methods
8.1 Extraction Method of Region Suitable for Image Adjustment Image adjustment is necessary for long-term inspection. A region suitable for automatic contrast / brightness adjustment and automatic focus adjustment is a region where there are abundant lines and ends in the horizontal and vertical directions. If an area satisfying this requirement is obtained from the design data and used for automatic contrast / brightness adjustment and automatic focus adjustment, each automatic adjustment can be efficiently performed. Therefore, there is a need for a method for automatically determining a region suitable for this image adjustment.

  A method of extracting a region suitable for image adjustment based on the geometric information of the line segment constituting the design data or the relationship with the adjacent line segment is performed in the following procedure.

FIG. 130 is a diagram for explaining this method.
(1) The size of the rectangular area R used for each automatic adjustment is determined. The size of this region R is determined by experience.
(2) Determine the area A to be used for each automatic adjustment. This area is preferably near the inspection area. Region A is larger than region R.

(3) While moving the region R with respect to the region A, the total length of the vertical line segments existing in the region in the design data corresponding to the region R is obtained. Similarly, the total value of the lengths of the horizontal line segments is obtained. The total value in the vertical direction or the horizontal direction having a smaller value among the obtained total values is used as the evaluation value. Here, the total length of the line segments is used as the geometric information.
(4) A portion having the largest evaluation value obtained in the above (3) is obtained as an optimum region (a region having many vertical and horizontal lines).

The rectangle R (region with many vertical and horizontal lines) obtained by the above procedure is registered in the above-mentioned 3.3 recipe data “(11) Place suitable for automatic contrast / brightness adjustment and automatic focus adjustment”. It is possible to perform automatic contrast / brightness adjustment and automatic focus adjustment at an appropriate timing during inspection.

  The region suitable for automatic astigmatism adjustment has many line segments, and the total lengths of the line segments in the X direction, Y direction, 45 degree direction, and 135 degree direction are approximately the same length. It is an area. In this case, the same procedure as described above is executed except that the total length of the line segments in the 45 degree direction and the 135 degree direction is used in addition to the X direction and the Y direction. When this condition is not satisfied, automatic astigmatism adjustment is performed by the following procedure using a partial region corresponding to the end or corner of the reference pattern.

  First, a region suitable for automatic astigmatism adjustment is obtained. An example of this region is a region including a partial region including the left and right ends as shown in FIG. Another example is an area including a partial area including upper left, lower left, upper right, and lower right corners as shown in FIG. These regions may be regions including partial regions including upper and lower ends. If such a region is obtained, an edge having an inclination in all directions is present, which is suitable for automatic astigmatism adjustment.

  The following procedure similar to the procedure for obtaining a region suitable for automatic contrast / brightness adjustment and automatic focus adjustment is performed. Here, the procedure in the case of using the upper left, lower left, upper right, and lower right corners will be described with reference to FIG.

(1 ′) The size of a rectangular region R ′ used for automatic astigmatism adjustment is determined. The size of this region R ′ is determined by experience.
(2 ′) A region A ′ to be used for automatic astigmatism adjustment is determined. Region A ′ is larger than region R ′.
(3 ′) While moving the region R ′ relative to the region A ′, the number of upper left corners existing in the region of the design data corresponding to the region R ′ is obtained. Similarly, the number of the lower left corner, the number of the upper right corner, and the number of the lower right corner are obtained. The evaluation value is the number of the upper left corner, the number of the lower left corner, the number of the upper right corner, and the number of the lower right corner having the smallest value among the obtained numbers. Here, the number of corners is used as geometric information.

The region R ′ having the largest evaluation value among the evaluation values obtained in (4 ′) and (3 ′) is the optimum region (region having many of the four corners at the upper left, lower left, upper right, and lower right). Get as.
(5 ′) Some corners are thinned out so that the numbers of the upper left, lower left, upper right, and lower right corners are the same.
The vicinity of the corner obtained as described above is used as a partial region.

It is possible to register a rectangular region obtained by the above procedure in a recipe and perform automatic astigmatism adjustment at an appropriate timing at the time of inspection. This automatic astigmatism adjustment is performed by the following procedure.
(1) Carry out automatic focus adjustment.
(2) The image and the region R ′ suitable for automatic astigmatism adjustment are matched.
(3) An astigmatism evaluation value is obtained from the partial image corresponding to the partial region P included in the region R ′.
(4) The above (2) and (3) are executed while changing the value of astigmatism.
(5) The astigmatism value for the best astigmatism evaluation value obtained in (4) is obtained as the best astigmatism value.

  According to the present embodiment, it is possible to automatically and optimally extract a region suitable for image adjustment. When this area is a separated area, automatic adjustment is performed with higher accuracy than when the entire image is used.

8.2 Selection Method of Sub Inspection Unit Area Most Suitable for Matching When the inspection unit area is very large, the inspection unit area is divided into a plurality of sub inspection unit areas and inspected. In this case, when distortion and rotation of the inspection target pattern image can be ignored, matching with one of the sub-inspection unit areas is executed, thereby matching the inspection unit area. Therefore, in this embodiment, a method for selecting the most suitable for matching among these sub-inspection unit areas will be described.

The sub-inspection unit region most suitable for matching is a sub-inspection unit region having the largest evaluation value among the evaluation values obtained by performing the following calculation for all the sub-inspection unit regions.
(1) A unique pattern is obtained by the method shown in FIGS. 51 (a) and 51 (b).
(2) The line segments constituting the unique pattern are classified into four directions of horizontal, vertical, 45 degree direction, and 135 degree direction, and the total line segment length is obtained.

(3) The sum of the second largest line segment length is used as the evaluation value. The reason for using the sum of the second largest line segment length is that it is necessary that there are enough line segments in at least two directions.
FIG. 133 shows two sub-inspection unit areas. Here, the dotted line represents the reference pattern, and the solid line represents the unique pattern. The left sub-inspection unit area in FIG. 133 has many vertical lines but few horizontal lines. On the other hand, the sub-inspection unit area on the right side of FIG. 133 has many horizontal lines and relatively many vertical lines. The total length of the line segments forming the unique pattern in the left sub-inspection unit area of FIG. 133 is longer than the total length of the line segments forming the unique pattern in the right sub-inspection unit area of FIG. The evaluation value of the right sub-inspection unit area is larger than the evaluation value of the left sub-inspection unit area.

  The above calculation was performed by obtaining a unique pattern as two dimensions. However, this method has a drawback of requiring a large amount of calculation. Therefore, a method is used that requires less computation but less accuracy. The present embodiment will be described with reference to FIG.

The sub-inspection unit region most suitable for matching is a sub-inspection unit region having the largest evaluation value among the evaluation values obtained by performing the following calculation for all the sub-inspection unit regions.
(1) Line segments constituting the reference pattern are classified into four directions of horizontal, vertical, 45 degree direction, and 135 degree direction to form one-dimensional data.
(2) A unique pattern is obtained for these one-dimensional data by a method according to the method shown in FIGS. 51 (a) and 51 (b). This corresponds to the one-dimensional data of the horizontal line segment in FIG. 134 and the solid line in the one-dimensional data of the vertical line segment.

(3) The total sum of the lengths of the unique patterns existing in the one-dimensional data of the line segments in the horizontal, vertical, 45 degree direction, and 135 degree direction is obtained.
(4) The sum of the lengths of the second largest unique pattern among the above is used as the evaluation value.
By using this example, when the inspection unit area is divided into a plurality of sub-inspection unit areas, the sub-inspection unit area most suitable for matching can be obtained. This makes it possible to execute faster than the matching using the entire inspection unit area.

8.3 Inspection Method Using High-magnification Image and Low-magnification Image In the case of an SEM having a function capable of electromagnetically observing a part of a low-magnification image as a high-magnification image, it is possible to inspect patterns that do not fit in the high-magnification image. That is, it means that the edge position obtained with the high-magnification image can be accurately converted to the edge position obtained with the low-magnification image. This same relationship may be realized with a high-precision stage. For example, in FIG. 135, positions 182 and 183 on the pattern 181 of the inspection target pattern image are obtained from the high-magnification images 184 and 185, respectively, and then converted to positions on the low-magnification image 187, If the width 186 of 181 is obtained, the length can be measured with higher accuracy than when the width 186 is obtained using only the low-magnification image 187.

8.4 Inspection method of inspection target pattern affected by pattern of previous process The inspection target pattern where the previous process pattern exists in the lower layer is the part of the inspection target pattern where the previous process pattern exists in the lower layer and the lower layer The formation and appearance of the inspection target pattern is different from the inspection target pattern in the portion where the previous process pattern does not exist. As a countermeasure, it is possible to use an inspection method that uses different inspection parameters for an inspection target pattern in a portion where the pattern of the previous process exists and an inspection target pattern in a portion where the pattern of the previous process does not exist.

FIG. 136 is a diagram schematically illustrating an example in which the pattern of the previous process is observed as a base. In such a case, the inspection area is divided into three areas inside, the boundary portion, and the outside of the pattern existing in the previous process. The inside of the pattern existing in the previous process is recognized by the same method as the reference pattern C described in the above-mentioned 5.3.1 Gate line width inspection method . The boundary portion of the pattern existing in the previous process is recognized as a band-shaped portion having a band center line which is a reference pattern of the previous process pattern and has an empirically determined width. The outside of the pattern existing in the previous process is the remaining part.

  The contrast may be different between the inside and outside of the pattern of the previous process due to the influence of the previous process pattern. Also, the width of the pattern to be inspected formed due to the height difference of the lower layer surface may be different.

In order to reduce these effects, the correction amount of the line segment and the allowable pattern deformation amount are set separately for the inside and outside of the pattern in the previous process. If the boundary portion is suitable for edge detection, the correction amount of the position of another line segment and the allowable pattern deformation amount are set. If the boundary portion is not suitable for edge detection, it is removed from the inspection target.
According to this example, it is possible to reduce the probability of detecting a deformation (pseudo defect) that does not need to be regarded as a defect as in the lattice portion of FIG.

8.5 Overwriting display method of defect information and information corresponding to the defect Defect shape or defect image as defect information, design data and mask data corresponding to the defect information (generated by adding an OPC pattern to the design data) ), It is easy to understand the tendency of defects when parallel display or overwriting display is performed on one or more of the shape obtained by the lithography simulator using design data or other information related to the design data Become. A display method that meets this requirement is needed.

The following are examples of the tendency for defects to occur.
(1) Many defects are detected where design data is involved.
(2) Many defects are detected where a specific OPC pattern is attached.
(3) Many defects are detected in the narrowed shape obtained by the simulation using the design data.

In the case of wafer inspection, it is also useful to use a photomask image corresponding to a defect. Compared with the photomask image, it can be determined whether the defect is caused by the photomask or not.
In order to realize these displays, the information related to the design data is associated with the detected defect. This association is performed by the following procedure.

(1) The design data information is added to the edge of the reference pattern. As the information to be added, the polygon cell name, line segment number, which is the design data to which the edge belongs, the coordinates of the start point and end point of the line segment to which the edge belongs, and the position coordinates on the line segment corresponding to the edge are used.
(2) When a defect is detected, design data information added to the edge of the used reference pattern is added to the inspection result.

(3) Using the added design data information, information related to the design data is retrieved. Even if the information related to the design data is described in a coordinate system different from the design data, it can be associated with the polygonal cell name and line segment number as the design data.
FIG. 137 shows an example in which design data, mask data, and a defect image are overwritten and displayed. FIG. 138 (a), FIG. 138 (b), FIG. 138 (c) and FIG. 138 (d) show an example of a method for displaying the recognized defect as a figure. In this example, the following display method is used.

(1) A method of displaying a polygon which is the outline of a defect as shown in FIG.
(2) As shown in FIG. 138 (b), in the case of a dent defect or a convex defect, a method of displaying a rectangle representing the outermost frame of the defect. A short line segment may be put at the corner of the convex defect rectangle to distinguish it from the concave defect.
(3) As shown in FIGS. 138 (c) and 138 (d), in the case of a defect having an abnormal line width, a method of displaying a rectangle having a side representing the detected line width.

  In the above method, the inspection result is directly used and displayed. However, as shown in FIG. 139, a method of displaying defects after converting them into design data can be used. This method is executed in the following procedure.

(1) The polygons representing the defects obtained in FIGS. 138 (a), (b), (c), and (d) are stored in the design data.
(2) When the design data includes a layer that describes a pattern that actually exists and a layer that describes a pattern that does not exist, the layer that describes the pattern that does not exist is displayed in FIGS. 138 (a), (b), The polygon obtained in (c) and (d) is stored.
(3) When a plurality of layers describing a non-existing pattern can be used, each layer may be stored in a separate layer for each of a dent defect, a convex defect, and a defect with abnormal line width. FIG. 139 shows an example in which the design data is stored in the layer 1, the dent defect and the convex defect are stored in the layer 12, and the defect of the line width abnormality is stored in the layer 13.

According to this method, the inspection result can be browsed by an apparatus that handles design data, which is convenient for design change and the like.
With the above method, the overwriting display described above may be processed in the same manner instead of the parallel display.

  According to the present embodiment, it becomes easy to understand the tendency of defects to occur, and the cause of the defect can be easily identified, so that the design change is facilitated.

9. Test method of easily charged samples such as inspection method resist samples easily charged sample will be described. One is a method of applying a carbon coating to the pattern, and the other is a method of inspecting only the central portion of the image. The former method requires a process for carbon coating, but can be inspected using images acquired under high-throughput scanning conditions. On the other hand, the latter method does not require a carbon coating process. However, when the latter method is used, high-throughput inspection cannot be performed.

9.1 Carbon Coating Method for Resist Sample First, a lithography process for manufacturing a semiconductor device to be inspected by the pattern inspection apparatus of the present invention will be described. 140A to 140F are schematic views showing a lithography process. 140 (a) to 140 (f) show typical examples of single-layer manufacturing processes. The semiconductor device is manufactured by forming a multilayer by repeating this single layer process.

Through a thermal oxidation process, an oxide film (SiO ₂ ) 502 is formed on the silicon substrate 501 as shown in FIG. Next, a resist film 503 is formed on the oxide film 502 by a resist coating process, as shown in FIG. Then, as shown in FIG. 140C, in the exposure process, the exposure apparatus (stepper) 505 partially exposes the resist film 503 with the ultraviolet rays that have passed through the photomask 504.

Next, as shown in FIG. 140D, when the resist film 503 is developed by the development process, the exposed portion of the resist film 503 is removed. Thereafter, as shown in FIG. 140E, the oxide film 502 where the resist film 503 has been removed is removed by an etching process. Next, as shown in FIG. 140F, the resist film 503 is removed by a resist peeling (ashing) step.
In the lithography process described above, the pattern formed on the resist film 503 is formed according to the design data by the pattern inspection apparatus of the present invention using the pattern formed on the resist film 503 after the development process shown in FIG. Can be inspected.

  The pattern inspection apparatus inspects the pattern formed on the resist film 503 using an electron beam (charged particle beam). When the resist film 503 is irradiated with an electron beam as it is in the state shown in FIG. 140 (d), the resist is an insulator generally made of a polymer compound. Is deformed. This is because the electron beam is bent by the upper surface of the partially charged resist film 503, so that the electron beam is not irradiated to an accurate position of the resist. Therefore, in this embodiment, before the pattern formed on the resist film 503 is inspected by the electron beam, the carbon film is coated on the pattern formed on the resist film 503 so that the charging phenomenon does not occur.

  That is, as shown in FIG. 141, the carbon film 506 is coated on the resist film 503 and on the portions where the resist film 503 is removed by development and the oxide film 502 is exposed. Thus, by coating the carbon film 506, the electron beam flows to the silicon substrate 501 through the carbon film 506 when the electron beam is irradiated. As a result, since the silicon substrate 501 flows to the ground, the charging phenomenon can be prevented.

  In this case, the thickness of the carbon film 506 varies depending on the line width of the pattern, but is preferably about 5 nm to 20 nm, and more preferably about 10 nm. If the thickness of the carbon film 506 is less than 5 nm, the conductivity is slightly poor, and if it exceeds 20 nm, the carbon coating is formed so as to protrude into the space portion of the pattern. It may be recognized as a defect. Therefore, the thickness of the carbon film 506 is appropriately about 5 nm to 20 nm. For example, a vacuum deposition method or a sputtering method can be suitably used as the carbon coating method. In this embodiment, a carbon coating method using a carbon sputter coating apparatus will be described.

  FIG. 142 is a schematic view showing a semiconductor wafer inspection system in which the pattern inspection apparatus of the present invention and the carbon sputter coating apparatus 510 are integrally provided. As shown in FIG. 142, a carbon sputter coating apparatus 510 is disposed adjacent to the pattern inspection apparatus shown on the right side. A wafer transfer robot 515 and a preliminary exhaust chamber 517 are installed between the carbon sputter coating apparatus 510 and the wafer transfer apparatus 340 of the pattern inspection apparatus. A cassette 516 containing a plurality of semiconductor wafers W is placed on the wafer transfer robot 515. The semiconductor wafer W in the cassette 516 can be taken out and transferred to the carbon sputter coating apparatus 510 by the wafer transfer robot 515. The wafer transfer robot 515 can transfer the semiconductor wafer W after carbon coating by the carbon sputter coating apparatus 510 to the wafer transfer apparatus 340 of the pattern inspection apparatus.

  In the present embodiment, the carbon sputter coating apparatus 510 includes a sputter apparatus, and a holder 511 that holds the semiconductor wafer W, and a carbon rod 512 that is disposed to face the semiconductor wafer W held by the holder 511. And a sputtering control device 513 for controlling the sputtering. The preliminary exhaust chamber 517 maintains the vacuum in the carbon sputter coating apparatus 510 when the semiconductor wafer W is transported and carried out to the carbon sputter coating apparatus 510, and processes the semiconductor wafer W in the carbon sputter coating apparatus 510. It is provided to wait for a new semiconductor wafer W during the period.

  In the configuration of the semiconductor wafer inspection system shown in FIG. 142, the semiconductor wafer W accommodated in the cassette 516 by the wafer transfer robot 515 is transferred to the carbon sputter coating apparatus 510 via the preliminary exhaust chamber 517. At this time, the carbon sputter coating apparatus 510 and the preliminary exhaust chamber 517 are evacuated to the same pressure. When the semiconductor wafer W is transferred into the carbon sputter coating apparatus 510 and is held by the holder 511, a gate valve (not shown) disposed between the preliminary exhaust chamber 517 and the carbon sputter coating apparatus 510. Is closed, and carbon coating of the semiconductor wafer W is performed.

  By this carbon coating process, as shown in FIG. 141, a carbon film 506 is applied on the resist film 503 and on the portion where the resist film 503 is removed by development and the oxide film 502 is exposed. The semiconductor wafer W thus provided with the carbon film 506 is unloaded from the carbon sputter coating apparatus 510 via the preliminary exhaust chamber 517 by the wafer transfer robot 515 and transferred to the wafer transfer apparatus 340. Then, the semiconductor wafer W is transferred to the sample chamber 320 by the wafer transfer device 340 and inspected by the pattern inspection device. In the pattern inspection apparatus, the semiconductor wafer W is irradiated with an electron beam, but the electron beam flows to the silicon substrate 501 through the carbon film 506, and as a result, flows from the silicon substrate 501 to the ground, thereby preventing a charging phenomenon. Can do.

  According to the semiconductor wafer inspection system shown in FIG. 142, the carbon coating process and the pattern inspection process can be performed continuously. Therefore, the inspection process can be speeded up and the throughput is improved. In addition, the coating process and the pattern inspection process can be automated.

  Furthermore, the edge detection accuracy is improved. In the case of FIG. 143 (a) without the carbon coating, when the inspection target pattern image is acquired by the electron beam, the inside of the pattern becomes unevenly or unstablely bright due to a charging phenomenon. This is because the charging phenomenon occurs non-uniformly or unstablely due to the density of the pattern, and the secondary electron emission ratio due to the charging phenomenon varies depending on the inside of the pattern and the material of the base. As a result, an accurate edge cannot be detected. On the other hand, in FIG. 143 (b) in the case where the same sample is coated with a carbon film, the potential of the sample surface is constant, so that the edge effect is uniformly generated and the brightness of the base is also uniform. By this effect, the edge portion is stably and uniformly brightened, and a uniform contrast is formed at the boundary between the edge and the ground. As a result, an accurate edge can be detected.

9.2 Inspection method for inspecting only the central portion of an image When a sample such as a resist sample, which is easily charged, is observed, the peripheral portion of the image is distorted. The central portion of the image is charged to an equipotential. However, in the peripheral part of the image, distortion occurs because the potential is not uniform between the place where the electron beam is not irradiated and the place where it is irradiated. Furthermore, once the sample portion is charged, charging continues for a long time.

  In order to inspect the inspection region of such a sample, an inspection method for inspecting only the central portion of the image shown in FIG. 144 can be used. A rectangle indicated by a dotted line in FIG. 144 is an inspection unit region. In addition, the rectangle indicated by the solid line is an enlarged inspection unit region. The enlarged inspection unit region is scanned, and an image corresponding to the central inspection unit region becomes an inspection target.

  However, in this method, the enlarged inspection unit region is charged and cannot be used for two or more inspections. Therefore, as shown in FIG. 144, after obtaining the inspection results from the four dies of the upper left die, the upper right die, the lower left die, and the lower right die, the inspection results of the obtained four dies are fused. Thus, the inspection result of the inspection area is obtained.

  According to the present embodiment, the pattern formed on the resist film on the silicon substrate can be inspected without performing a special process.

10. Although an example of the present invention has been described above over the variants of the present invention, various other modifications are possible. For example, a scanning electron microscope is used as the image generating device 7 that scans an inspection target pattern with an electron beam (charged particle beam) and obtains an image of the inspection target pattern, but a scanning focus ion beam microscope, a scanning laser microscope, It can be applied to various scanning microscopes such as a scanning probe microscope. Further, the scanning direction is not limited to 0 degrees and 90 degrees, and an arbitrary minute angle such as 5 degrees or 95 degrees may be added.

  The acquired image data may be transformed into an offline input processing type via an external input device such as a magneto-optical disk or a magnetic tape, or via a LAN (Local Area Network) such as Ethernet.

  The image generation method may be another method, and the reference pattern may be converted from other data. The reference pattern may be generated at the time of inspection without being registered in the recipe database 22.

  In the present embodiment, the inspection results and the like are output to the display device 5 and the printing device 6. However, the inspection results may be output to an image database, a simulator, a recording medium, or the like. You may make it transmit (output) to a computer.

  Furthermore, after inspecting a semiconductor device called a typical die in a wafer by the method of the present invention, it is possible to adopt a hybrid method in which other dies are inspected by die-to-die comparison.

DESCRIPTION OF SYMBOLS 1 Main control part 2 Memory | storage device 3 Input / output control part 4 Input device 5 Display apparatus 6 Printing apparatus 7 Image generation apparatus 11 Reference pattern generation part 12 Inspection part 13 Output part 14 Defect type recognition part 21 Core database 22 Recipe database 23 Defect type Reference database 24 Defect information storage unit 25 Defect recognition unit 33A, 33B, 41, 42 circumscribed rectangle 34 circumscribed rectangle 35, 43 common circumscribed rectangle 61-70, 75, 81-84 edge 101-103 portion 111, 113 line width 112, 114 Space width 121 to 126 Linear shape pattern 157 Sections 159, 163, 165 Edge 160, 166 of inspection target pattern image Edge 164 of reference pattern Edge 161 constituting end of reference pattern Edge center of gravity 162 of inspection target pattern image Reference pattern Edge of The center of gravity 171 The straight line portion 172 The corner 173 The end 174 The isolated pattern 181 The pattern image 182 and 183 of the inspection target pattern image The position 184 and 185 The high magnification image 186 The pattern width 187 of the inspection target pattern image The low magnification image 201 The broken line 202 and 203 The solid line 204 The background 205 Pattern interior 206, 207 Pixel cluster 251-255 Pixel 261 Pixel center 262 Point 263 on reference pattern closest to pixel center Tangent 301-304 Inspection unit area 310 Irradiation system device 311 Electron gun 312 Focusing lens 313 X deflector 314 Y deflector 315 Objective lens 316 Lens control device 317 Image acquisition device 318 Deflection control device 320 Sample chamber 321 XY stage 322 XY stage control device 330 Secondary electron detector 340 Wafer transfer device 350 Control computer 360 Operation computer 400 Round circle 401 Block 402 Hexagonal block 410 Oscillator 411 Counter 412 X deflection generation circuit 413 Y deflection generation circuit 414 Distortion correction vector calculation circuit 450 Sub deflection generation circuit 452 X main deflection generation circuit 453 Y main deflection Generation circuit 501 Silicon substrate 502 Oxide film 503 Resist film 504 Photomask 505 Exposure apparatus (stepper)
506 Carbon film 510 Carbon sputter coating device 511 Holder 512 Carbon rod 513 Sputter control device 515 Wafer transfer robot 516 Cassette 517 Pre-exhaust chamber

Claims (10)

A pattern inspection apparatus for inspecting using an inspection target pattern image and data used for manufacturing the inspection target pattern,
Generating means for generating a reference pattern expressed by a line segment or a curve from the data;
Generating means for generating a reference pattern used for area inspection from the data;
Generating means for generating the inspection target pattern image;
Means for detecting an edge of the inspection target pattern image;
An inspection means for inspecting the inspection target pattern by comparing an edge of the inspection target pattern image with a reference pattern represented by the line segment or curve;
The pattern inspection apparatus, wherein the inspection unit inspects the inspection target pattern by comparing an edge of the inspection target pattern image with a reference pattern used for the region inspection.   The pattern inspection apparatus according to claim 1, wherein a reference pattern used for the region inspection is extracted using geometric information of a line segment that forms the data or a relationship between line segments that are in contact with or close to each other. A pattern inspection apparatus.   The pattern inspection apparatus according to claim 2, wherein as the inspection using the reference pattern used for the region inspection, the line width inspection of the linear shape pattern, the space width inspection, the average line width inspection, the average space width inspection, the curve shape pattern A pattern inspection apparatus which is at least one of line width inspection, space width inspection, average line width inspection, average space width inspection, or cutting / short-circuit inspection.   4. The pattern inspection apparatus according to claim 2, wherein means for extracting a reference pattern to be used for the region inspection from the data includes the polygon of the data of the layer related to the inspection object and the layer of the layer related thereto. A pattern inspection apparatus using a result of a logical operation with a polygon of data.   5. The pattern inspection apparatus according to claim 2, wherein the defect is cut or short-circuited by detecting the edge in a direction different from an edge of a reference pattern used for the region inspection. A pattern inspection device characterized by A pattern inspection apparatus for inspecting using an inspection target pattern image and data used for manufacturing the inspection target pattern,
Generating means for generating a reference pattern expressed by a line segment or a curve from the data;
Generating means for generating the inspection target pattern image;
Means for detecting an edge of the inspection target pattern image;
Inspection means for inspecting the inspection target pattern by comparing the edge of the inspection target pattern image with a reference pattern represented by the line segment or curve;
The inspection means obtains a defect position and a defect size obtained from the same inspection area expressed by coordinates on the data for a plurality of semiconductor devices manufactured based on the same data. A pattern inspection apparatus comprising: a defect recognizing unit that recognizes a defect that repeatedly occurs when a region having a defect size and a region having a defect size overlap.   7. The pattern inspection apparatus according to claim 6, wherein at least one first semiconductor device among the defect positions and defect sizes obtained from the plurality of semiconductor devices is set as a first semiconductor device. The defect position and the defect size are obtained from the inspection region, and for the other second semiconductor device, it is examined whether the region having the defect position and the defect size obtained by the first semiconductor device overlaps. A pattern inspection apparatus for inspecting only a necessary area. A pattern inspection apparatus for inspecting using an inspection target pattern image and data used for manufacturing the inspection target pattern,
Generating means for generating a reference pattern expressed by a line segment or a curve from the data;
Generating means for generating the inspection target pattern image;
Means for detecting an edge of the inspection target pattern image;
An inspection means for inspecting the inspection target pattern by comparing an edge of the inspection target pattern image with a reference pattern represented by the line segment or curve;
The inspection unit sets a defect type for each item obtained from geometric information of the reference pattern, an item obtained from the data information, or an item obtained from data information related to the data. Pattern inspection device.   9. The pattern inspection apparatus according to claim 8, wherein the defect classification is performed using a characteristic obtained by cutting out the reference pattern in the vicinity of the defect.   9. The pattern inspection apparatus according to claim 8, wherein the data information is used to add the correction pattern, the shape data obtained by simulation using the data, or another data related to the data. A pattern inspection apparatus, wherein any one of data is associated with the defect.