JP2005183369A - Method and device for observing sample - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that an observational direction (or an incident direction of electron beams) in an observed image actually obtained has some errors compared to a set value and the errors affect on analysis of the observed image thereafter. <P>SOLUTION: Convergent electron beams are irradiated on a sample with a known shape, electrons emitted from the sample surface are detected, and intensity of the detected electron is formed into an image, and this image is used to estimate an incident direction of the convergent electron beams based on a geometric deformation on the image of the sample with a known shape, and information about the estimated incident direction of the convergent electron beams is used to obtain a three-dimensional shape or a cross directional shape of the sample as an object of observation from a SEM image of the sample as an object of observation. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

The present invention relates to a method and an apparatus for observing a sample using an SEM (Scanning Electron Microscope) in observation or measurement of a sample such as a semiconductor wafer in a semiconductor manufacturing process.

With the miniaturization of semiconductors, it is becoming increasingly difficult to control the semiconductor pre-process manufacturing process. In addition to the height, line width, and side wall inclination angle of the semiconductor pattern, subtle changes in the pattern shape such as corner rounding greatly affect the electrical characteristics of the semiconductor pattern. Therefore, there is a demand for a technique for detecting a process variation by measuring these dimensions or shapes and controlling the process. To perform effective process control, SEM (Scanning
A technique for estimating a three-dimensional profile from observation of a sidewall of a semiconductor pattern with an electron microscope (scanning electron microscope) and an image acquired by SEM (hereinafter referred to as SEM image) is expected. In the side wall observation and the estimation of the three-dimensional profile, it is considered effective to use information of an SEM image obtained by observing the sample from an oblique direction.

As a method for acquiring the SEM image observed from the oblique direction, for example, JP 2000-
As disclosed in Japanese Patent No. 348658, a system in which an electron beam irradiated from an electron optical system is deflected and a tilt image is tilted to irradiate an observation target with an electron beam, and an inclined image is taken. In order to observe the image, a method of inclining and imaging the stage itself for moving the semiconductor wafer and a method of mechanically inclining the SEM electron optical system itself are applied.

JP 2000-348658 A

However, since the observation direction (or the incident direction of the electron beam) in the observation image actually obtained in the prior art is expected to include an error with respect to the set value, the error is used for the analysis of the subsequent observation image. May have an impact. For example, when a process variation is detected by monitoring a dimensional value such as a line width or a contact hole diameter, the dimensional value changes depending on the observation direction, and thus an error in the observation direction becomes a problem (references “Characterisation of 193”).
nm Resist Layers by CD-SEM Sidewall Imaging ”, Proceedings of SPIE Vol.5038, pp.892-900, 2003). In 3D profile reconstruction technology using stereo measurement using observation images from multiple directions. Are the observation direction and the plurality of SEs in a plurality of SEM images.
Since the profile is estimated based on the parallax between the M images, the error in the viewing direction affects the estimated profile error.

An object of the present invention is to accurately estimate the observation direction of a tilt observation image and to calibrate the tilt angle based on the estimated observation direction or to use the estimated observation direction as input information when measuring dimensions or profiles. An object of the present invention is to provide a sample observation method and apparatus that enable accurate measurement or observation.

In order to achieve the above object, in the present invention, a sample having a known shape is irradiated with a focused electron beam, electrons emitted from the sample surface are detected, and an image obtained by imaging the intensity of the detected electrons is used.
Estimate the incident direction of the convergent electron beam based on the geometric deformation on the image of the sample with the known shape, and use the estimated observation direction for the subsequent observation image analysis for high-accuracy measurement. I made it possible. The sample having a known shape includes a sample prepared using a material having a crystal structure (
The angle between the crystal planes that make up the surface is known), or the sample surface shape is measured by a measuring means such as a scanning probe microscope (SPM) or by using laser interference exposure. A sample with a guaranteed pitch accuracy in the pattern was used.

Based on the estimated observation direction, for example, adjusting the deflection of the electron beam (hereinafter referred to as beam tilt) or the tilt angle of the stage or the optical system itself (hereinafter referred to as stage tilt and lens barrel tilt, respectively) for actual observation The direction can be adjusted to the set value. In the three-dimensional profile estimation technique, the estimation accuracy can be improved by using the estimated incident direction as an input value of the observation direction of the observation image.

The estimated observation direction is displayed on a GUI (Graphic User Interface) and is referred to when the user analyzes the observation image. In addition, based on the observation direction, a function of converting an image obtained by deflecting an electron beam into an image obtained by tilting the stage and displaying the image is provided. Since the beam tilt image obtained by the deflection of the electron beam is geometrically different from the image that is seen when a human observes from an oblique angle, it is difficult to interpret sensuously. In view of this, it is effective in visual analysis to convert and display a stage tilt image obtained by tilting a stage closer to human observation.

That is, in order to achieve the above object, in the present invention, an SEM image obtained by irradiating and scanning a focused electron beam on a sample and detecting secondary electrons or reflected electrons generated from the sample by the irradiation is obtained. In this method of observing a sample, a calibration with a pattern with a known shape formed by irradiating and scanning an electron beam focused from an oblique direction on the surface of a calibration substrate with a pattern with a known shape formed. An SEM image of the surface of the substrate is acquired, and an angle of the oblique direction irradiated with the electron beam is obtained using information on the shape of the pattern having a known shape and information on the obtained SEM image, and the obtained angle in the oblique direction is obtained. Is adjusted to a desired angle, and the sample substrate on which the pattern is formed is irradiated with an electron beam from a desired oblique angle. Get the SEM image, and to obtain the cross-sectional shape of the three-dimensional image or pattern of the sample on the substrate of the pattern on the sample substrate by processing the SEM image of a sample substrate by using the information of a desired angle.

In order to achieve the above object, according to the present invention, an SEM obtained by irradiating and scanning a sample with a focused electron beam and detecting secondary electrons or reflected electrons generated from the sample by irradiation.
In a method of observing a sample using an image, a pattern with a known shape is formed by irradiating and scanning an electron beam focused on the surface of a calibration substrate on which a pattern with a known shape is formed. The SEM image of the surface of the calibration substrate obtained was obtained, and the angle of the oblique direction irradiated with the electron beam was obtained using the information on the shape of the pattern having a known shape and the information on the obtained SEM image, and the pattern was formed. An SEM image of the sample substrate is acquired by irradiating the sample substrate with an electron beam from an oblique angle obtained, and the SEM image of the sample substrate is processed using information on the angle obtained from the SEM image of a pattern having a known shape. Thus, a three-dimensional image of the pattern on the sample substrate or a cross-sectional shape of the pattern is obtained.

According to the present invention, the observation direction can be accurately estimated by using a sample on which a pattern having a known shape is formed as a calibrator, and more accurate and reproducible shape measurement is possible.

In addition, according to the present invention, it is possible to accurately estimate the observation direction by using a sample in which a pattern with a known shape is formed as a calibrator. Results came to be obtained.

The present invention will be described with reference to FIGS.

0. Inclination observation method by SEM
FIG. 1 shows an example of a system for acquiring and processing SEM images according to the present invention. Reference numeral 1303 denotes an electron beam source which generates an electron beam 1304. The deflector 1306 deflects the electron beam 1304 to control the position of the electron beam on the semiconductor wafer 1301 that is a sample. Secondary electrons and reflected electrons are emitted from the semiconductor wafer 1301 irradiated with the electron beam, and the secondary electrons are detected by a secondary electron detector 1309. On the other hand, the reflected electrons are detected by the reflected electron detectors 1310 and 1311. The backscattered electron detectors 1310 and 1311 are installed in different directions. The secondary electrons and reflected electrons detected by the secondary electron detector 1309 and the reflected electron detectors 1310 and 1311 are converted into digital signals by the A / D converters 1312, 1313 and 1314, stored in the image memory 13151, and the CPU 13152. The image processing according to the purpose is performed.

FIG. 3 shows a method of imaging a signal amount of electrons emitted from a semiconductor wafer when the semiconductor wafer is irradiated with an electron beam. For example, as shown in FIG. 3A, the electron beam is irradiated by scanning in the x and y directions as 1501 to 1503 or 1504 to 1506. The scanning direction can be changed by changing the deflection direction of the electron beam. electron beam 1501 to 1503 that is scanned in the x direction indicates the location on the semiconductor wafer irradiated with G ₁ ~G ₃ respectively.

Similarly, locations on the semiconductor wafer irradiated with the electron beams 1504 to 1506 scanned in the y direction are indicated by G _{4 to} G ₆ , respectively. The signal amounts of the electrons emitted from G _{1 to} G ₆ are the lightness values of the pixels H _{1 to} H ₆ in the image 1509 shown in FIG. 3B (lower right subscripts in G and H). 1 to 6 correspond to each other). Reference numeral 1508 denotes a coordinate system indicating the x and y directions on the image.

  Reference numeral 1315 in FIG. 1 denotes a computer system that estimates the incident direction of the convergent electron beam from the observation image of the calibration sample 1319 by image processing, or estimates the three-dimensional profile from the observation image of the target pattern on the semiconductor wafer 1301. Or processing and control such as sending a control signal to the stage controller 1320 and the deflection controller 1321. The processing / control unit 1315 is connected to the display 1316 and includes a GUI (Graphic User Interface) that displays an image to the user. Reference numeral 1317 denotes an XY stage, which moves the semiconductor wafer 1301 and enables imaging of an arbitrary position of the semiconductor wafer. Although FIG. 1 shows an embodiment having two reflected electron image detectors, the number of the reflected electron image detectors can be reduced or increased.

  As a method of observing an object with tilting in SEM, (1) a system in which an electron beam irradiated from an electron optical system is deflected, and a tilted image is taken by tilting the irradiation angle of the electron beam (for example, JP-A-2000-348658) (Hereinafter referred to as a beam tilt method, and the obtained image is referred to as a beam tilt image.) (2) A method in which the stage 317 itself for moving the semiconductor wafer is tilted (in FIG. 1, the stage is tilted at a tilt angle 1318. ) (Hereinafter referred to as the stage tilt method, and the resulting image is referred to as the stage tilt image), (3) the method of mechanically tilting the electron optical system itself (hereinafter referred to as the lens barrel tilt method, and the obtained image is mirrored) Called a cylinder tilt image). In order to perform high-precision analysis such as line width and contact hole diameter measurement and 3D profile estimation from the SEM images acquired by the above method, and to detect process variations and process control, convergence is required. The incident direction of the electron beam must be known correctly. However, the conventional technique has no means for confirming how much the actual incident direction is deviated from the set value of the incident direction of the convergent electron beam, and a method for accurately estimating the actual incident direction is required. It is done.

1. Invention basic idea
In order to solve this problem in the present invention, a sample having a known shape is irradiated with a focused electron beam, electrons emitted from the sample surface are detected, and an image obtained by imaging the intensity of the detected electrons is used. We found a method for estimating the incident direction of a convergent electron beam based on geometrical deformation of a sample with a known shape. That is, the incident direction (tilt angle) of the convergent electron beam is estimated based on the posture of the sample (calibrator) whose shape is known. When the attitude of the calibrator with respect to the absolute coordinate system of the electron optical system can be measured, the incident direction of the electron beam with respect to the absolute coordinate system of the electron optical system can be measured.

  In the absolute coordinate system of the electron optical system, the electron beam incident direction (central axis of the optical system) at a beam tilt angle of 0 ° is the z axis, and the electron beams 1501 to 1501 scanned in the x direction shown in FIG. A coordinate system in which the direction orthogonal to the z-axis on the plane including 1503 is the x-axis, and the direction orthogonal to the z-axis on the plane including the electron beams 1504 to 1506 scanned in the y-direction is the y-axis. is there.

  If the attitude of the calibrator with respect to the absolute coordinate system of the electron optical system cannot be measured, some reference (for example, the actual incident direction of the electron beam when the set value of the beam tilt angle is 0 ° (may be slightly different from the set value) It is possible to measure the incident direction of the electron beam with respect to the coordinate system defined by the coordinate system with z) as the z axis.

FIG. 2 shows a sequence for estimating the tilt angle in the present invention. First, in step 1401, the beam tilt angle, stage tilt angle, or lens barrel tilt angle of the apparatus is set (hereinafter, the beam tilt angle, stage tilt angle, and lens barrel tilt angle are collectively referred to as the tilt angle), and in step 1402 Observe the calibrator. As a calibrator, (1) a sample using a crystal plane (the angle between crystal planes constituting the surface is known), or (2) a sample whose shape is measured by a measuring means such as SPM, or (3) A sample in which the accuracy of pitch in the pattern formed by using laser interference exposure is guaranteed is used.
In step 1403, the tilt angle is estimated based on the geometric deformation of the calibrator on the observed image, and a deviation from the set value of the incident direction of the convergent electron beam in step 1401 is detected. Step 1403 is composed of steps 1404 and 1405. Characteristic differences in the appearance of the calibrator include places where the gradient of the sample shape changes greatly, such as ridges and valleys. This is hereinafter referred to as “edge”.

  In step 1404, an edge is detected from the calibrator observation image, and in step 1405, the tilt angle is estimated based on the geometric deformation on the calibrator observation image. At this time, some or all of the information such as the shape / posture / position 1406 of the calibrator, the posture / position information 1407 of the wafer surface, or the image distortion 1408 is necessary for estimating the tilt angle. Therefore, in step 1405, a part or all of the information 1406 to 1408 is estimated based on the geometrical deformation on the observed image of the calibrator as well as the tilt angle, or image processing, SPM, or objective lens in-focus position A part or all of the information 1406 to 1408 is estimated by combining distance measuring means using the above. In order to estimate the tilt angle, at least the attitude of the calibrator is known or needs to be estimated in the information 1406 to 1408.

  The estimated tilt angle is measured by (1) calibration of the tilt angle (matching the tilt angle setting value with the actual tilt angle) or (2) using the estimated tilt angle as an input value for stereo measurement or the like. It is used for observation or measurement of observation images, such as improving accuracy or (3) displaying the estimated tilt angle on a GUI (Graphic User Interface) and using it as a reference value during user observation and analysis. Next, details of each step will be described.

2. About calibrators with known shapes
2.1 Types of calibrator
As described above, the sample having a known shape used as a calibrator in the present invention is either (1) a sample using a crystal plane (the angle between crystal planes constituting the surface is known), or (2) the shape of the sample surface is SPM. A sample measured by a measuring means such as (3) a sample whose dimensional accuracy is guaranteed by pitch formation by laser interference exposure, or (4) a sample that can be used as a reference device can be used.

面で構成されている。そのため，結晶面同士の傾斜角は既知となる。また，同図（ｂ）１０２は四角錐形状の凸部を有する試料（以下，凹ピラミッド型の試料と呼ぶ），同図（ｃ）１０７は山型の試料（傾斜角tan^-1（√2）°），同図（ｄ）１０８はライン・アンド・スペース型の試料（傾斜角９０°）である。この他にも，Ｓｉ結晶のみならずGaAs等様々な結晶を利用した様々な結晶面を利用した試料を生成することができる。 The sample of type (1) will be described. One method of forming the sample surface with crystal planes is crystal anisotropic etching. This is an etching technique that uses the property that the etching rate differs depending on the crystal plane. For example, in Si crystal, the etching rate for the Miller index (111) plane is much higher than the etching rate for the (100) plane or (110) plane. slow. Therefore, the etching proceeds so that the (111) plane having a low etching rate appears on the surface. 4 (a) to 4 (d) show examples of samples constituted by Si crystal planes. A sample having a quadrangular pyramid-shaped recess (hereinafter referred to as a concave pyramid sample) shown in FIG.

It is composed of planes. Therefore, the tilt angle between crystal planes is known. FIG. 10B shows a sample having a quadrangular pyramidal projection (hereinafter referred to as a concave pyramid type sample), and FIG. 10C shows a mountain-shaped sample (inclination angle tan ⁻¹ (√2 (D) 108 is a line and space type sample (inclination angle 90 °). In addition to this, it is possible to generate samples using various crystal planes using not only Si crystals but also various crystals such as GaAs.

As a modification of the pyramid type sample, the upper part of the pyramid as shown in FIG. 5A (corresponding to the vicinity of the vertex P ₀ in FIG. 4B) is flat and formed by the vertices Q _{5 to} Q ₈ . A pyramid sample in which such a surface exists, a modified pyramid type sample whose upper part of the pyramid is rounded, or a region where the surfaces constituting the pyramid as shown in FIG. 5C intersect (FIG. 4B) (Corresponding to the vicinity of line segments Q ₀ -Q ₁ , Q ₀ -Q ₂ , Q ₀ -Q ₃ , and Q ₀ -Q ₄ ) in FIG.

  The sample of type (2) will be described. The shape of the sample surface is measured by measuring means such as SPM, and distance data is obtained. By making a polyhedral approximation to the distance data by local plane fitting, it is possible to measure the angle formed between each plane constituting the sample surface. According to the measurement method, any sample suitable for SEM image observation can be used as a calibrator.

  The sample of type (3) will be described. In exposure using laser interference, a highly accurate pitch pattern is formed according to the wavelength of the laser, and therefore the pitch in the pitch pattern is known.

  The sample of type (4) will be described. For example, the pole-type sample and the hole-type sample shown in FIGS. 18A and 18B, respectively, can be used as a reference unit for the observation direction (the cut surface has a pole or a hole as in the illustrated example). Including shapes that gradually increase or decrease in the direction of the central axis). 18 (c1) and (c2) are examples of observation images when FIG. 18 (a) or (b) is observed from above. For example, in the case of the hole in FIG. 18B, when the hole is perpendicular to the wafer upper surface 1809, the center position of the upper surface 1804 and the lower surface 1803 of the hole in the observation image is one as shown in FIG. 18C1. If so, the observation direction can be determined to be perpendicular to the wafer. That is, it can be used as a reference device for the observation direction. Further, as the hole is deeper, the center position of the upper surface 1804 and the lower surface 1803 of the hole is greatly shifted as shown in FIG. 18 (c2) with only a slight shift of the observation direction from the vertical direction. Is possible.

  A sample containing a steep shape represented by a pole type or a hall type is advantageous in that the sample fits in the field of view and the change in the height of the calibrator in the z direction can be greatly taken for high resolution. is there. Further, it is possible to estimate the amount of deviation from the vertical direction of the observation direction from the deviation of the center position by a geometrical relationship. Furthermore, the observation direction can be calibrated in the vertical direction by adjusting the center position. Further, when the direction in which the holes are vacant is an arbitrary direction with respect to the upper surface of the wafer, similarly, it is determined whether the observation direction is the arbitrary direction, or the amount of deviation between the observation direction and the arbitrary direction is estimated. Alternatively, the observation direction can be calibrated in the arbitrary direction.

  Once the hole vacancy direction is clarified by measurement or the like, it can be used as a reference unit for the observation direction with respect to the hole vacancy direction. Even in the case of a pole-type sample, it is possible to determine, estimate, or calibrate the observation direction in the same manner as the hole-type sample described above.

  In addition, the observation directions that can be judged, estimated, or calibrated by an arbitrary reference device represented by a pole type or a hall type can be measured by an arbitrary calibrator. That is, as shown in FIG. 18D, for example, a pyramid type calibrator 1808 is observed in an observation direction in which the center positions of the upper surface 1805 and the lower surface 1806 of the hole 1807 coincide with each other as shown in FIG. By estimating the observation direction, the inclination direction of the hole 1807 (observation direction that can be determined, estimated, or calibrated by the reference device) can be obtained.

  The definition of 0 ° of the beam tilt angle can be defined as 0 ° in the z-axis direction in the absolute coordinate system in which the central axis of the electron optical system described above is the z-axis. Alternatively, an arbitrary observation direction defined by a reference device or the like can be set to 0 °.

The observation direction is estimated based on the difference in the appearance of the sample due to the difference in the observation direction. In order to estimate the observation direction, which will be described later in detail, the characteristic shape of the sample on the observation image, for example, the concave pyramid segments 301 to 308 shown in FIG. It is necessary to detect and use part or all of the length or inclination angle of the line segment or the coordinate values of the vertices P _{0 to} P ₄ shown in FIG. 6A (vertex P in FIGS. 4 and 6). ₀ -P ₄ and R ₁ -R ₆ are corresponding). The line segments correspond to “edges” in which the gradient of the sample shape changes greatly. In the line and space type sample in FIG. 6B, the line segments are 309 to 311 and the like.

  Next, a method for detecting an edge from an observed image by image processing will be described. As the observation image, either a secondary electron image obtained by imaging the signal quantity detected by the secondary electron detector 1309 or a reflected electron image obtained by imaging the signal quantity detected by the reflected electron detectors 1310 and 1311 is used. .

  In the secondary electron image, the difference in the gradient of the sample shape is reflected in the difference in the signal amount. In addition, a phenomenon called an edge effect is observed at the edge portion where the gradient of the sample shape changes greatly. A large amount of signal is detected in the mountain shape, and a small amount of signal is detected in the valley shape. Thus, an edge can be detected by detecting a place where the signal amount changes or a place where the signal amount is large or small on the image.

  Also in the reflected electron image, the difference in the gradient of the sample shape is reflected in the difference in the signal amount, so that the edge can be detected by detecting the location where the signal amount changes from the image. In the backscattered electron image, the above-mentioned edge effect does not occur. Therefore, when the edge is observed to be greatly expanded due to the edge effect, it may be desirable to use the backscattered electron image to improve the accuracy of edge detection. is there.

  Each line segment of the calibrator can be detected by detecting some locations on the line segment to be detected where there are edges where the gradient of the sample shape greatly changes and applying a straight line to these. For example, the detection of the line segment 304 in the pyramid-shaped sample in FIG. 6A will be described as an example with reference to FIG. 6A2 in which the region 324 surrounded by the dotted line in FIG. Several edges (corresponding to locations 314 in FIG. 6C) are detected on or near the line segment 304 in FIG. 6A2.

  Black dots 321 indicate a plurality of edge positions detected at arbitrary intervals along the line segment 304. The detection positions of the plurality of edge positions 321 vary somewhat for reasons such as image noise and may not exist on a straight line. However, noise may be caused by applying a straight line to the plurality of edge positions 321 using means such as the least square method. Thus, the straight line (that is, the line segment 304) can be detected with high accuracy. However, when the edge position 321 includes a large exceptional value having a tendency different from that of other point groups (for example, the edge position 322 far away from the straight line position), the estimated value by the least square method has low reliability. Therefore, exceptional edge positions that are far away from the detection line are excluded as exceptional values, or each edge position can be exceptionally handled by weighting it inversely proportional to the distance from the edge position detection line. A process that does not take into account the edge position can be selected.

For example, in detecting the edge position 321, it is convenient to know the approximate position of the line segment 304 in advance since the rough detection range or detection direction can be known. Therefore, before the edge position 321 is detected, the approximate position of the line segment 304 can be obtained by the detection process shown as an example below. First, a differential filter is applied to the entire image to roughly extract candidate positions of edge positions from the image. The approximate position of a square line segment is estimated by Hough transform, which is a generally known voting-type straight line detection method, from a point group obtained by binarizing the output value of the differential filter.
Further, some calibrator geometries include a circular pattern as shown in FIG. 18, but a circular or curved pattern is applied to the detected edge in the same manner as the straight line detection, or the geometric shape is obtained by directly using the edge coordinates. Can be expressed quantitatively.

2.2 Advantages of using pyramid samples
When these line segments are detected in an SEM image using image processing, there is an advantage in a pyramid type (quadrangular pyramid shape) sample. That is,
(1) Since the shape of the sample is symmetric with respect to the edge of the sample, the distribution of the signal amount in the observation image is also symmetric. Therefore, it is possible to detect the edge from the observation image with high accuracy.
(2) It is possible to estimate a two-dimensional (x, y direction) observation direction by observing one pyramid.

  FIG. 6A shows the SEM signal amount distribution 312 and the sample surface shape distribution 313 of the straight line AB in the concave pyramid for (1). Since the left and right shapes are approximately equal with respect to the position 314 of the straight line 304 to be detected, the distribution of the signal amount is also approximately equal on the left and right, and the degree of deviation of the signal amount is less, so that a linear position close to the true value can be detected. This tendency is the same for the mountain-shaped sample shown in FIG. 4C where the left and right sample surface shapes are similar to each other at the straight line to be detected. There is no.

  On the other hand, FIG. 6E shows the SEM signal amount distribution 318 and the sample surface shape distribution 319 of the straight line EF portion in the line and space type sample. Since the left and right shapes are different from each other at the position 320 of the straight line 310 to be detected, the distributions of the left and right signal amounts are also different, and the signal amounts tend to be biased to either one. For this reason, there is a risk that the detected linear position is also deviated from the true value. However, when the signal amount deviation is large in this way, it is possible to improve the detection accuracy by correcting the signal amount deviation by estimating the generated signal distribution predicted from the sample shape by Monte Carlo simulation or the like. .

Further, when the coordinate values of the vertices P _{0 to} P ₄ shown in FIG. 6A are detected and used for estimating the observation direction, for example, it is conceivable to obtain the vertices from the detected intersections of the respective straight lines. For example, the vertex P ₀ is obtained from the intersection of some or all of the straight lines 301 to 304, and the vertex P ₁ is obtained from the intersection of some or all of the straight lines 301, 305, and 306. A detection example of the vertex P ₀ is shown in FIG. 6 (a 1) in which a region 323 surrounded by a dotted line in FIG. 6 (a) is enlarged. In the case of estimating a vertex from three or more straight lines in FIG. 6 (a1), there may be cases where the intersection points P _{0a to} P _0d respectively detected from the intersection points of a combination of two arbitrarily selected straight lines do not match. In that case, the variation in straight line detection can be reduced by setting the center of gravity or median of the plurality of intersections as the vertex P ₀ .

FIG. 4B shows a modification of the pyramid type sample in which the upper part of the pyramid in FIG. 5B (the sample in FIG. 5A is observed from the direction perpendicular to the wafer surface) is flat. Vertices Q _{5 to} Q ₈ or line segments 1605 to 1608 that do not correspond to the convex pyramid shown in b) can be detected and used to estimate the geometric deformation of the pyramid on the observed image. On the other hand, it is also possible to express the geometric shape of this modification by using the same apex or line segment as a normal pyramid type sample. That is, each of the line segments 1601 to 1604 is extended, and a virtual vertex Q ′ ₀ corresponding to the vertex P ₀ in FIG. 6A is detected from the intersection, and can be used for estimation of the geometric deformation of the pyramid ( When the vertex Q ′ ₀ is detected, the same centroid calculation as in FIG. 6 (a1) can be performed. When the virtual vertex Q ′ ₀ is used, an analysis similar to the normal pyramid shown in FIG. 6A is possible.

As for the sample etc. in which the region where the faces composing the pyramid shown in FIG. 5 (c) intersect is cut and flattened, there is a method in which a line segment etc. that does not exist in a normal pyramid type sample is characterized. For example, the virtual vertex Q ′ ₁ corresponding to the vertex P ₁ in the normal pyramid-shaped sample in FIG. 6A can be detected by extending the line segments 1613 to 1614 and calculating intersection points, for example. it can. Further, the line segments 1617 to 1618 are respectively extended to detect a virtual vertex Q ′ ₅ (corresponding to the vertex Q ₅ in FIG. 6A), and by connecting the virtual vertices Q ′ ₁ and Q ′ ₅ , FIG. A virtual straight line 1609 corresponding to the straight line 301 in a) can be detected. Furthermore, the virtual vertex Q ″ ₀ corresponding to the vertex P ₀ in FIG. 6A can be detected from the intersections of some or all of the detected virtual straight lines 1610 to 1612 in the same manner as the virtual straight line 1609. When the virtual straight lines 1609 to 1612 or the virtual vertex Q ″ ₀ are used, the same analysis as the normal pyramid shown in FIG. 6A is possible.

  Regarding (2) above, in the pyramid type sample, since the inclinations in the x and y directions of the observation direction both appear in the displacement of the apex of the pyramid, the two-dimensional observation direction is estimated by observing one pyramid. Is possible. On the other hand, in the mountain-shaped and line-and-space-type samples shown in FIGS. 4C and 4D, the direction in which there is no edge (for example, the direction connecting the vertices R1 and R2 in FIG. 4D). Although the change in the observation direction cannot be estimated, the two-dimensional observation direction is estimated by arranging a plurality of the same samples so that the edge of the shape has two directions (for example, arrangement example 709 described later). It is possible.

2.3 Sample size and layout
Next, taking a pyramid type sample as an example, the size and arrangement of the sample will be described.
The size of the sample is (1) the size that fits within the field of view at the time of imaging, (2) the size that can observe the length of a sufficient line segment to ensure the detection accuracy of each line segment of the pyramid, and (3) the observation direction It is necessary to have sufficient height to increase the resolution of displacement on the image of the pyramid detected when changes occur. Depending on the imaging magnification of the SEM and the focal depth of the SEM, for example, if the imaging magnification is 30k to 150k and the focal depth is about several um, the size of the pyramid is suitably about 0.1 um to several um.

  Explain where to place the sample and how to place it. Reference numerals 704 and 705 in FIG. 7 indicate candidate locations for attaching the sample. The sample is attached to 704 on a semiconductor wafer 702 (corresponding to 1301 in FIG. 1) mounted on a stage 701 (corresponding to 1317 in FIG. 1) or 705 on a holder 703 on the stage 701. In the case of mounting on 704, a calibration-specific wafer in which a calibrator is formed instead of the semiconductor wafer 702 is created and mounted instead of the semiconductor wafer 702, or the calibration pattern is added together when forming the semiconductor pattern. The generated semiconductor wafer 702 is attached.

  Although it is generally difficult to generate a calibration pattern such as the pyramid when forming the semiconductor pattern, the shape of the semiconductor pattern generated on the semiconductor wafer 702 is measured by a measuring means such as SPM. , Can also be used as a calibrator whose shape is known. In addition, as shown by 704, samples are arranged at intervals of several locations (in the figure, 9 locations as an example), and the observation direction is estimated at each location, so that the electric field changes due to stage distortion or stage movement. It is possible to measure the in-plane distribution of the resulting change in the viewing direction. Next, an example of the arrangement of calibrators in 704 or 705 will be described with reference to FIG.

  In an arrangement example 706 in FIG. 7B, pyramids 710 of the same size (corresponding to 101 or 102 in FIG. 4) are arranged at regular intervals. By arranging multiple pyramids, (1) Estimating the observation direction in each pyramid, and using the average value or median (median) of the observation directions, observations due to individual differences in the pyramid, image noise, etc. Variations in direction estimates can be reduced. (2) By observing multiple pyramids in the same field, the distribution in the field of view can be measured. (3) Sample contamination In such a case, there is an advantage that good calibration can be continued by observing another sample while avoiding contamination.

  Further, as shown in FIG. 7 (c), image distortion in the field of view is estimated and corrected by observing a plurality of pyramids whose arrangement is known in the same field of view (in the figure, arranged in a grid). Can do. That is, (1) the scanning direction of the convergent electron beam (x, y on the image) so that the line A-B connecting the vertices of the pyramids originally aligned on the straight line in FIG. (2) Correct image distortion by means such as reconstructing an image without distortion by geometric deformation from the obtained image. Such estimation and correction of the image distortion in the visual field can be performed by one pyramid observation. That is, for example, if it is guaranteed that the line segments 301 to 308 of the concave pyramid in FIG. 6A are straight lines, the line segments 301 to 308 are observed as curves (for example, as shown in FIG. 7D). An observation image of a concave pyramid), and a distortion correction method that corrects this to a straight line is conceivable. When the distortion of the observation image is large, it is necessary to correct the distortion of the observation image in order to estimate the tilt angle with high accuracy.

  In order to perform distortion correction between images having different tilt angles in a beam tilt image, a method of correcting distortion by matching the shape of the bottom surface of the pyramid between the images can be considered. In other words, in the beam tilt image, focusing on the unique property of the beam tilt image that the shape of the geometric pattern on the horizontal plane is constant regardless of the beam tilt angle with respect to the xy plane of the electron optical system. When the plane is horizontal with respect to the xy plane of the optical system, the shape of the bottom surface of the pyramid on the wafer surface is constant regardless of the tilt angle. It is thought to be due to distortion. Therefore, corrections that match the shape of the bottom surface of the pyramid between images are performed on the image, or information on the detected line segment, vertex, or parameter (for example, the inclination angle of the line segment) of the pyramid. It is possible to obtain information on the above-mentioned image obtained without any distortion, or information on parameters representing the line segment, vertex, or shape. That is, it is possible to obtain an observation image with less distortion and to estimate the observation direction with higher accuracy.

  As an example, FIG. 8 shows a distortion correction process for matching the shape of the bottom surface of the beam tilt image with the shape of the bottom surface of the top-down image on the basis of the shape of the bottom surface of the top-down image. The shape of the bottom surface used as a reference is a bottom surface shape obtained from a top-down image, a bottom surface shape obtained from a beam tilt image observed at an arbitrary beam tilt angle, or a plurality of images having the same or different observation directions. The obtained average bottom shape can be used, but there are many cases where the image distortion is less than that of the top-down image group beam tilt image, and it can be obtained from the bottom shape obtained from the top-down image or a plurality of top-down images. In many cases, it is effective to use the average bottom surface shape as a reference.

A bottom shape 1701 (a line connecting the vertices P _{1 to} P ₄ and P ₁ in this order) in FIG. 8 is a bottom shape in the top-down image 1704, and the bottom shape 1702 in the beam tilt image 1705 is corrected so as to match the shape of 1701. The deformation parameter to be obtained is obtained. A distortion-corrected beam tilt image 1706 is obtained by correcting the distortion of the beam tilt image 1702 using the deformation parameter. Incidentally, the base shape 1703 in the beam tilt image 1706 after distortion correction is more similar to the base shape 1701 than the base shape 1702.

  Arrangement example 707 is an arrangement in which pyramids 710 having the same size are arranged with no interval between pyramids. In addition to the same measurement as in the arrangement example 706, the distribution of the SEM signal amount on the bottom side of the pyramid (the line segments 305 to 308 in FIG. 6A) indicates the left and right signals with the position of the bottom side to be detected as a boundary. There is an advantage that the distribution of quantities is approximately equal. That is, the distribution of the SEM signal amount in the straight line CD portion shown in FIG. 6A tends not to be symmetric with respect to the position 315 of the base 306 as shown in FIG. In the arrangement example 707, a distribution close to symmetry and a straight line position close to the true value can be detected.

  In the arrangement example 708, a plurality of pyramids having different sizes (for example, 710 and 711 as pyramids having different sizes) are arranged. When the size of the field of view changes due to changes in the imaging conditions, it is possible to select and observe a pyramid having an appropriate size from among pyramids of different sizes in order to ensure the estimation accuracy of the observation direction.

  The above arrangement example can be applied to the mountain-shaped sample 107 or the line-and-space-type sample 108 shown in FIGS. 4C and 4D, respectively. Further, in the case of the mountain-shaped sample 107 or the line-and-space-type sample 108, as shown in the arrangement example 709 in FIG. ) Can be used to estimate a two-dimensional (x, y direction) observation direction like the pyramid type sample described above.

3. About the tilt angle estimation method
3.1 Definition of positional relationship of wafer surface, calibrator, and electron beam incident direction
Next, the beam tilt SEM observation will be taken as an example, and the beam tilt angle estimation method will be described using a concave pyramid type sample as a calibrator as an example. FIG. 9A schematically shows the positional relationship among the wafer surface 203, the pyramid 205 (calibrator), and the incident direction 207 of the electron beam when a concave pyramid sample is observed. The wafer surface refers to a place where a pyramid is not formed on the surface of the sample on which the pyramid is created, and the word is distinguished from a semiconductor wafer that forms a semiconductor pattern to be actually analyzed. An example of a description method of the positional relationship between the wafer surface, the pyramid, and the incident direction of the electron beam will be described. The above-described absolute coordinate system 201 of the electron optical system is used as a reference for the positional relationship.

  First, the inclination of the wafer surface indicated by 203 in FIG. 9A can be expressed using the unit surface normal vector 204 of the wafer surface. The unit surface normal vector 204 can be expressed as shown in FIG. 9B by using rotations ψy and ψx around the x and y axes of the unit surface normal vector 204 with respect to the absolute coordinate system 201.

Next, the inclination of the concave pyramid indicated by 205 in FIG. 9A can be expressed by using a unit direction vector 206 of the central axis of the concave pyramid. The unit direction vector 206 can be expressed as shown in FIG. 9C2 using the rotations φy and φx around the x and y axes of the unit direction vector 206 with respect to the absolute coordinate system 201. The central axis of the concave pyramid is a reference axis provided for uniquely defining the posture of the pyramid. For example, as shown in FIG. 9 (c1), each side P ₀ -P ₁ , P ₀ -P ₂ of the pyramid , P ₀ −P ₃ , P ₀ −P _4, and a straight line 206 passing through a point having the same distance. The relative angle between the inclination of the wafer surface and the inclination of the pyramid is known if the wafer surface 203 is a (100) crystal plane. 9C3, the rotation of the concave pyramid indicated by 205 (rotation of crystal orientation) represents, for example, the angle between the projection vector of the crystal orientation (100) vector of the wafer on the xy plane and the x axis as φz. be able to.

  Next, the incident direction of the electron beam 207 indicated by 207 in FIG. 9A can be expressed using a unit direction vector 208 of the incident direction of the electron beam 207. The unit direction vector 208 can be expressed as shown in FIG. 9D by using the rotations θy and θx around the x and y axes of the unit direction vector 208 with respect to the absolute coordinate system 201.

In order to estimate the incident direction of the electron beam (for example, expressed by θx and θy), some or all of the parameters ψx, ψy, φx, φy, φz, θx, and θy must be estimated. In other words, the equation representing the geometric relationship between the observation amount (for example, the tilt angle of the calibrator edge) and the output variable (electron beam incident direction) is a parameter such as ψx, ψy, φx, φy, φz, θx, θy, etc. This is the case when it cannot be described without using. However, in estimating the tilt angle, at least the shape and the tilt direction of the calibrator need to be estimated or known.
The method of defining the positional relationship of the incident direction of the wafer surface, pyramid, and electron beam is not limited to that using the parameters ψx, ψy, φx, φy, φz, θx, θy as described above. Although the case where a pyramid type sample is used as the calibrator has been described as an example here, the tilt direction can be uniquely defined in a calibrator other than the pyramid type sample.

3.2 Tilt angle estimation method 1 (basic principle: single image, single image)
A method for estimating the beam tilt angle from the captured beam tilt image by image processing will be described. FIG. 6A is a schematic diagram of an SEM image when the concave pyramid is observed from above, but the change in the beam tilt angle is reflected in the appearance of the pyramid. That is, the length and inclination angle of each of the line segments 301 to 308 of the pyramid in FIG. 6A (the inclination angle is defined by, for example, an angle formed by the x axis of the image and the line segment) changes depending on the beam tilt angle. .

3.2.1 Output parameter definition
Therefore, a parameter representing the beam tilt angle can be estimated by detecting a part or all of the line segments 301 to 308 from the captured image and measuring the length and the tilt angle. Hereinafter, the parameter to be finally estimated is called an output parameter. Here, it is a parameter representing the beam tilt angle, and is given by, for example, the aforementioned parameters θx and θy. The output parameter group is expressed as an output parameter vector U = (θx, θy) ^T.

3.2.2 Reference parameter definition
The length and inclination angle of the line segment of the pyramid also change depending on changes in parameters representing the wafer surface, the inclination angle of the pyramid, image distortion, and the like. That is, in order to obtain an output parameter, parameters representing the wafer surface, the inclination angle of the pyramid, image distortion, and the like must be selectively estimated as necessary. Hereinafter, these parameters that must be estimated in order to obtain the output parameters are referred to as reference parameters. Here, parameters represent the wafer surface and pyramid tilt angles and image distortion. For example, the wafer surface and pyramid tilt angles are given by the aforementioned parameters ψx, ψy, φx, φy, and φz. The reference parameter group is expressed as a reference parameter vector V = (ψx, ψy, φx, φy, φz) ^T. Reference parameters are added and deleted as necessary.

  That is, if it is not necessary to estimate the wafer surface gradients ψx and ψy in order to estimate the output parameters, ψx and ψy are deleted from the reference parameters, and image distortion is similarly estimated to estimate the output parameters. If necessary, a parameter describing the image distortion is added to the reference parameter. Here, the case where it is not necessary to estimate the gradients ψx and ψy of the wafer surface is a case where, for example, an estimation parameter is obtained using only information on the gradients of the line segments 301 to 304 on the observation image in FIG. is there. This is because the change in the gradient of the wafer surface affects the gradient of the line segments 305 to 308 in FIG. 6A, but does not affect the gradient of the line segments 301 to 304.

The output parameter and the reference parameter are collectively referred to as an estimation parameter, and the estimation parameter group is expressed as an estimation parameter vector T = (U ^T , V ^T ) ^T.

3.2.3 Estimated parameter estimation method
FIG. 10 shows a conceptual diagram of the estimation parameter estimation method. Reference numeral 401 denotes a processed image obtained by extracting a line segment by image processing from the imaged SEM image or reflected electron image of the calibrator or the SEM image or reflected electron image. In order to obtain the estimated parameter vector T (405 in FIG. 10) when the calibrator 401 is imaged, a combination of various values is substituted for each element of the estimated parameter vector, and what kind of calibrator is observed in each combination. Simulate. The combination of the various values of each element in the estimated parameter vector is denoted by T ⁽¹⁾ , T ²⁾ , ..., T ^(n-1) , T ⁽ⁿ⁾ , T ^{(n +1)} ,..., For example, calibrators observed when the estimated parameter vector values are T ⁽ⁿ⁻¹⁾ , T ⁽ⁿ⁾ , T ^{(n + 1)} (406 to 408 in FIG. 10 respectively ⁾ . These shapes are given by 402 to 404, respectively, from a geometrical calculation formula.

The correlation between the calibrator shape simulated in this way and the calibrator shape in the captured image is calculated, and the estimated parameter vector T ⁽ⁿ⁾ (407) in the most similar simulated shape (403 in FIG. 10) is the estimated parameter in the captured image. This corresponds to the vector 405. The method of calculating the estimated parameter vector that maximizes the correlation between the captured image and the simulation image by changing the estimated parameter vector in various ways is the method of applying numerical values in order as described above, or the analytical method using the least square method Etc. As a correlation calculation method, the length or inclination angle of a part or all of the line segments 301 to 308 shown in FIG. 6A in the captured image and the simulation image or the angle shown in FIG. There is a method based on the degree of similarity of some or all of the coordinate values of the vertices P _{0 to} P ₄ .

  In the above description, the beam tilt angle is the output parameter to be finally estimated. However, when the stage tilt angle and the lens barrel tilt angle are estimated from the tilt image obtained by the stage tilt method or the lens barrel tilt method, respectively. Is the same. That is, in each tilt method, if the geometric shape change on the calibrator image with respect to an arbitrary estimated parameter can be simulated, it is possible to calculate an estimated parameter vector that most closely matches the captured image and the simulated image. is there. For example, in the case of the stage tilt method, the calibrator attached to the stage is tilted together with the tilt of the stage, so the output parameter that is the observation direction (the incident direction of the electron beam on the sample) may be the tilt angle of the calibrator.

The tilt angle of the calibrator is obtained by determining the estimation parameter vector T as shown in FIG. 10 in the same manner as the estimation of the beam tilt angle. Here, the estimated parameter vector T includes an inclination angle of the calibrator that is an output parameter. When the observation direction for observing the semiconductor wafer surface vertically is defined as a stage tilt angle of 0 °, the pyramid with respect to the observation direction at any tilt angle is defined as φ ₀ with respect to the observation direction for observing the semiconductor wafer surface vertically. When the tilt angle is φ, the stage tilt angle is given by φ−φ ₀ .

  In addition, in the “sequence for estimating the beam tilt angle using two images of the top-down image and the beam tilt image” described later, a step of estimating the tilt angle of the calibrator using the top-down image is included. In this step, the inclination angle of the calibrator is estimated as an output parameter.

3.3 Tilt angle estimation method 2 (Variation: Combination with two images or other measuring means)
Various sequences can be considered as a method of estimating the estimation parameter. That is, in addition to the method of estimating all the estimation parameters from one observation image at a time as described above, (1) a sequence in which estimation parameters are divided and partially estimated or (2) a plurality of observation images A sequence to be used or (3) a sequence in which some or all estimation parameters are estimated by means other than an image and the estimation result is used, or a sequence in which (1) to (3) are combined. As an example,
(A) Sequence for estimating beam tilt angle using two images of top-down image and beam tilt image (b) Measuring wafer surface gradient and calibrator gradient by measuring means other than image processing, and said measurement result A sequence for estimating the beam tilt angle using the will be described in order.

3.3.1 Estimation Parameter Estimation Method-Variation 1
(A) Sequence for estimating beam tilt angle using two images of top-down image and beam tilt image It may be difficult to uniquely determine an estimation parameter from one beam tilt image. For example, when a tilted calibrator is observed on the image, whether it is due to the tilt angle 209 of the calibrator or the gradient 210 in the incident direction of the electron beam, the proportion is determined by the image quality or image resolution problem or the estimation parameter. There are cases where it is difficult to determine uniquely due to a problem that information obtained from an image is insufficient for the number.

  FIG. 11 shows an estimation parameter estimation sequence using two images.

  The point of this sequence is not only an image (referred to as a beam tilt image) in which the tilt angle is set in the direction to be observed (referred to as a beam tilt image), but also an image obtained by observing the calibrator with a beam tilt angle set value of 0 ° (hereinafter referred to as “beam tilt image”). The top-down image is used to determine the tilt angle of the calibrator by assuming that the actual tilt angle in the top-down image is 0 °. The beam tilt angle may be estimated independently using the beam tilt image.

  First, a top-down image is captured (step 501), and the tilt angle of the calibrator, the rotation angle of the calibrator, and other reference parameters as necessary are calculated using the top-down image (step 502). In step 502, the tilt angle of the calibrator is calculated assuming that the beam tilt angle is known (0 °). Therefore, the uniqueness of the solution becomes higher and the calculation is easier than when all the estimated parameters are calculated simultaneously. Next, after setting the beam tilt angle in a direction to be observed, a beam tilt image is captured (step 503), and the actual value of the beam tilt angle and other reference parameters as necessary are estimated (step 504). In step 504, since the tilt angle of the calibrator is known (using the value calculated in step 502), the tilt angle is calculated. There are several variations on which estimation parameters are calculated and in what order, but by dividing and estimating the parameters as described above, the number of estimation parameters is reduced and the uniqueness of the solution is increased.

  Here, the beam tilt angle in the top-down image was set to 0 °, and the subsequent calculations were performed. However, the actual beam tilt angle in the top-down image captured with the beam tilt angle set to 0 ° may deviate from 0 °. is there. In that case, the magnitude of the deviation may affect the accuracy of the subsequent estimation of the beam tilt angle. However, an optical system originally designed to acquire a top-down image is expected to have high setting accuracy at a beam tilt angle close to top-down. Therefore, assuming that the actual beam tilt angle in the top-down image is 0 °, the tilt angle is estimated by tilting from 0 ° by using the calibrator tilt angle estimated from the top-down image as an input value. It is possible to reduce an error in estimating the tilt angle rather than using the tilt angle of the calibrator estimated from the observation image by the value as an input value.

Further, in the stereo vision later described in detail, the beam tilt angle set value theta _L, theta _R in tilt image I _L obtained respectively, in the case of three-dimensional reconstituted with I _R, the beam tilt angle in the top-down image It is expected that the influence of the deviation Δθ ₀ from 0 ° on the three-dimensional reconstruction shape error is small. This is because when the tilt angle of each of the tilt images I _L and I _R is estimated using the tilt angle of the calibrator obtained from the top-down image by the sequence as an input value (estimated values are Est [θ _L ] and Est [θ _R ], respectively) The deviation from the true values of Est [θ _L ] and Est [θ _R ] is about the same as Δθ ₀ . Therefore, the relative angle error of Est [θ _L ] and Est [θ _R ] is small, and the error to the stereoscopic reconstruction shape by stereo vision is also small.

  Here, the subsequent calculation was performed with the beam tilt angle in the top-down image set to 0 °. Similarly, the estimated parameter is obtained by substituting an approximate value for the parameter that is considered to have no significant error. Can do. For example, the wafer surface gradient is approximately 0 °.

3.3.2 Estimated parameter estimation method-Variation 2
(B) Measuring the wafer surface gradient and calibrator gradient using measurement means other than image processing, and estimating the beam tilt angle using the measurement results. Estimating the parameters or the relationship (relative value, etc.) between the estimated parameters. In addition to the above-described image processing, there is a distance measuring means using SPM or an objective lens in-focus position. Here, an example of a method for estimating the beam tilt angle using image processing or other measurement means selectively or in combination will be described.

  FIG. 12 is a diagram showing a beam tilt angle estimation method using both SPM, objective lens in-focus position and image processing. First, before SEM observation, a sample is attached to the SPM, and the sample surface in the region including the wafer surface 602 and the calibrator 604 is measured. The obtained distance data 606 on the sample surface is described with reference to the coordinate system 601 set in the SPM. By approximating the distance data 606 by multi-plane approximation by local plane fitting, the angle formed between each plane constituting the sample surface can be measured, and the tilt direction 603 of the wafer surface and the tilt direction 605 of the calibrator can be measured. The relative angle can be calculated.

  Next, the sample is attached to the SEM, and the inclination direction 608 of the wafer surface 602 is estimated based on the distance measurement result by the in-focus position of the objective lens. The distance measurement based on the focal position of the objective lens means that the actual observation target is observed by calculating in advance the relationship between the control current of the objective lens and the distance to the observation target when focusing on the observation target. In this case, the distance to the observation object is estimated using the relationship. Further, by applying a plane or the like to the distance data based on the objective lens in-focus position, the tilt direction 608 is estimated from the surface normal. The tilt direction 608 of the obtained wafer surface 602 is described with reference to the coordinate system 607 set in the SEM.

  The relative angle between the wafer surface tilt direction 608 and the calibrator tilt direction 609 in the coordinate system 607 is equal to the relative angle between the wafer surface tilt direction 603 and the calibrator tilt direction 605 in the already measured coordinate system 601. The tilt direction 609 of the calibrator in the coordinate system 607 can be calculated using the tilt direction 608 of the wafer surface and the relative angle. Assuming that the tilt direction of the calibrator, which is one of the reference parameters, has already been calculated, the beam tilt angle 612 with respect to the coordinate system 607, which is the output parameter, is calculated by image processing. The depth estimation accuracy based on the in-focus position of the objective lens is about a fraction of the focal depth. Therefore, when measuring the gradient of the wafer surface, the measurement accuracy is ensured by calculating based on the distance data at a location where the distance from the end of the sample on the wafer surface is far away.

  In this way, not all the estimation parameters are measured by image processing, but each measurement means using SPM or objective lens in-focus position is selected in consideration of differences in measurement accuracy and usability of each measurement means. Alternatively, there is a method of measuring using a combination, and in addition to the above-described combinations, there are various variations such as estimating other than the relative angle between the wafer tilt direction 608 and the calibrator tilt direction 609 by image processing. It is done.

  Further, for example, the positional relationship between the wafer surface and the calibrator is not changed once it is measured, so it is not necessary to measure it every time the estimation parameter is obtained.

4). About the imaging sequence (including target imaging)-including the estimated value correction method that accompanies changing imaging conditions
Next, an imaging sequence of a calibrator including imaging of a semiconductor pattern (hereinafter referred to as a target pattern) at an arbitrary coordinate desired to be observed on a semiconductor wafer by SEM will be described. As an imaging sequence of calibrator and target pattern,
(A) Sequence for imaging a target pattern after calibrator imaging (FIG. 13A)
(B) Sequence for imaging with calibrator after target pattern imaging (FIG. 13B)
(C) Sequence for imaging the calibrator offline (FIG. 14)
Are described in order. In (a) and (b), since the target pattern and the calibrator are alternately imaged, the calibrator sample 1319 is attached at a position 705 in FIG. In (c), a calibrator sample 1319 is attached at a position 704 or 705 in FIG.

4.1 Imaging sequence-Variation 1
(A) Sequence for imaging a target pattern after calibrator imaging (FIG. 13A)
After the calibrator is imaged with an arbitrary tilt angle set, the setting is basically not changed (may be changed if tilt angle calibration is involved), and a sequence for imaging the target pattern can be considered. This is because if there is a hysteresis characteristic between the tilt angle setting value and the actual tilt angle, even if the arbitrary tilt angle setting value is changed to a different setting value and then returned to the original tilt angle setting value, This is to avoid the risk that the tilt angle estimated from the calibrator cannot be used in the observation image analysis of the target pattern because the actual tilt angle between the front and rear tilt angle setting values may not be the same.

  This sequence will be described with reference to FIG. First, the semiconductor wafer is attached to the SEM (step 800), and then the imaging position is set on the sample on which the calibrator is formed (step 801), and the tilt angle is changed to an arbitrary set value (step 802). However, there is no problem if the order of steps 801 and 802 is reversed. After setting the tilt angle, the calibrator is imaged at step 804. Before the calibrator is imaged, a plurality of calibrators are imaged under a plurality of imaging conditions as required by the loop 810 while changing the imaging conditions as necessary in step 803.

  When imaging a plurality of calibrators while changing imaging conditions (loop 810), a combination of imaging conditions required for estimating a tilt angle in an image group of a target pattern to be imaged later in loop 811 In this case, the calibrator is imaged. Furthermore, the effect of improving the reliability of the estimated value can be expected by estimating the tilt angle from a plurality of calibrator images. Therefore, if it is possible to estimate the tilt angle in the image group of the target pattern only with the observation image of the calibrator obtained under one imaging condition, the loop 810 is not passed. The change of the imaging condition refers to stage movement, image shift (field of view movement by changing electron beam irradiation position), magnification change, focus change, astigmatism correction, and the like. At the time of imaging, the imaging position may be changed or enlarged / reduced observation may be performed, and in that case, focus change or astigmatism correction may be accompanied in order to obtain a good image. Further, if necessary, an estimated value of the tilt angle is calculated from the calibrator image obtained in step 804, and the tilt angle is calibrated in step 805 so as to eliminate the deviation between the estimated value of the tilt angle and the set value of the tilt angle. To do.

  By performing the calibration, an observation image from the set observation direction can be obtained in step 808. The details of the tilt angle calibration method will be described later (FIG. 15). If necessary, the calibrator imaging (step 804) is performed by the loop 810 until the deviation between the estimated tilt angle value and the set tilt angle value is eliminated. The estimation and tilt angle calibration (step 805) are repeated. After correcting the tilt angle (step 805), re-capturing the calibrator (step 804) and confirming that there is really no deviation between the estimated tilt angle value and the set tilt angle value, skip step 805. When the target pattern is imaged, the setting value of the tilt angle is the same in step 804 and step 808. However, when the calibrator is not reimaged, the setting value of the tilt angle is determined in step 804 and step 808. May be different. However, there is no problem if the tilt angle at the time of imaging the target pattern can be estimated within the allowable error range required for the subsequent target pattern analysis.

  Next, the imaging position is moved onto the semiconductor wafer (step 806), and the target pattern is imaged in step 808. Before the target pattern is imaged, a plurality of target patterns are imaged as required by the loop 811 while changing the imaging conditions as necessary in step 807.

4.2 Imaging sequence-Variation 2
(B) Sequence for imaging with calibrator after target pattern imaging (FIG. 13B)
A sequence may be considered in which an arbitrary tilt angle is set and the target pattern is imaged under an arbitrary imaging condition, and then the tilt angle setting is not changed and the calibrator is imaged to estimate the tilt angle. This has the advantage that by observing the calibrator after imaging the target pattern, the imaging conditions at the time of observing the target pattern can be referred to when determining the imaging conditions at the time of calibrator imaging. In other words, since the tilt angle may change due to changes in the imaging conditions, imaging the calibrator under imaging conditions similar to the imaging conditions during target pattern observation makes it possible to more accurately determine the tilt angle during imaging of the target pattern. Can be estimated.

  This sequence will be described with reference to FIG. First, the semiconductor wafer is attached to the SEM (step 813), then the imaging position is set on the semiconductor wafer on which the target pattern is formed (step 814), and the tilt angle is changed to an arbitrary set value (step 815). However, there is no problem if the order of steps 814 and 815 is reversed. After setting the tilt angle, the target pattern is imaged in step 816. Before the target pattern is imaged, a plurality of target patterns are imaged under a plurality of imaging conditions as required by the loop 821, while changing the imaging conditions as necessary in step 815.

  Next, the imaging position is moved onto the sample on which the calibrator is formed (step 817), and the calibrator is imaged in step 819. Prior to imaging the calibrator, the imaging condition is changed as necessary in step 818, and images of the calibrator under a plurality of imaging conditions required for tilt angle estimation at the time of imaging the target pattern in step 816 are looped through loop 822. Take images as needed.

  In FIGS. 13A and 13B, the imaging condition change (steps 803, 807, 814, 817) and the imaging position change (steps 801, 806, 814, 817) are shown separately. The imaging condition change includes stage movement or image shift, which indicates stage movement or image shift between calibrators or target patterns, while the imaging position change is stage movement between calibrators and target patterns. Is shown. When the estimation of the tilt angle at the time of setting an arbitrary tilt angle shown in FIG. 11 is estimated from the combination of the tilt image and the top-down image at the time of setting the tilt angle of the calibrator, the tilt angle setting value 0 The imaging of the target pattern (steps 808 and 816) in ° (top down) is skipped.

4.3 Imaging sequence-Variation 3
(C) Sequence for imaging the calibrator offline (FIG. 14)
The calibrator is imaged with a combination of the tilt angle and the imaging conditions that would be required when estimating the tilt angle before the semiconductor wafer is inserted or the target pattern is imaged, and the tilt angle estimated from the captured image group or the captured image group is captured. A sequence that is stored in the database along with the conditions can be considered. It is possible to estimate the tilt angle in the target pattern imaged under any subsequent imaging condition from the database.

  This sequence will be described with reference to FIG. A plurality of calibrators are imaged with a combination of the tilt angle to be imaged and the imaging conditions (the tilt angle and the imaging conditions are changed in steps 901 and 902, respectively, and the imaging is performed in step 903). The calibrator image is stored in a database 905. If necessary, the image processing unit 902 calculates an arbitrary tilt angle setting value or an estimated tilt angle value under imaging conditions using an image group stored in the database, and stores the estimated value in the database 905 as a library. To do. By using the database, it is possible to estimate from the database a tilt angle estimation in a target pattern imaged under an arbitrary imaging condition thereafter. In addition, even when the calibrator image captured under the same conditions as the tilt angle setting value or the imaging condition at the time of capturing the target pattern is not stored in the database, the tilt when the image is captured under the other conditions in the database From the estimated angle value, the estimated tilt angle at the time of imaging the target pattern can be calculated by means such as interpolation.

  At the time of calibrator imaging in the three imaging sequences shown in FIGS. 13 (a), 13 (b), and FIG. 14, a plurality of other calibrators that are the same or continuously arranged are observed, the observation direction is estimated in each image, By using the average value or median (median value) of the observation direction, it is possible to reduce variations in the estimated value of the observation direction due to individual differences of calibrators and image noise.

4.4 Estimated value correction due to changes in imaging conditions
Even when the setting value of the tilt angle is not changed between the target pattern imaging and the calibrator imaging, the actual tilt between the calibrator imaging and the target pattern imaging is caused by the change of other imaging conditions. The corner may be different. Therefore, it is possible to estimate the tilt angle at the time of image capturing of the target pattern by estimating the amount of change of the tilt angle accompanying the change in the image capturing condition and adding the amount of change to the estimated tilt value at the time of image capturing by the calibrator.

  Next, a method for calculating the amount of change in tilt angle associated with the change in the imaging condition will be described.

The amount of change in tilt angle due to stage movement, image shift, or magnification change in the imaging condition change is distributed in the semiconductor wafer surface of the tilt angle (for example, estimated from the tilt angle in the calibrator attached at position 704 in FIG. 7). Alternatively, it can be estimated from the distribution of the tilt angle in the field of view (for example, estimated from the tilt angle in the 706 calibrator in FIG. 7). In particular, regarding the image shift, the change in the incident direction of the electron beam due to the shift of the electron beam can be estimated from the geometric calculation of the optical system. In addition, the amount of change in the tilt angle due to the focus change during the change in the imaging condition can be estimated by experimentally or analytically determining the relationship between the control current value of the objective lens and the tilt angle. In other words, since the focus is changed by changing the control current of the objective lens, if the relationship between the control current value of the objective lens and the tilt angle is obtained, the value of the control current of the objective lens at the time of observing the target pattern is obtained. The tilt angle can be estimated.

5). How to use the estimation results (tilt angle calibration, 3D reconstruction, GUI display, etc.)
How to use the estimated tilt angle in observation or measurement of an observation image will be described.

5.1 Tilt angle calibration
The tilt angle is calibrated using the estimated tilt angle value by the calibrator, and the set value of the tilt angle can be matched with the actual tilt (corresponding to step 808 in FIG. 13).

  FIG. 15 shows a sequence for correcting the tilt angle. First, after setting an arbitrary tilt angle to be observed, a calibrator image is acquired (step 1001, corresponding to step 804 in FIG. 13A), and an estimated value of the tilt angle is calculated from the calibrator image (step 1002 (corresponding to the determination of the estimation parameter T shown in FIG. 10 described above)). A difference Δθ between the estimated value of the tilt angle and the set value is calculated, and when the absolute value of Δθ is larger than a preset threshold value (condition 1004), a beam tilt method is used so as to reduce the absolute value of Δθ. If so, the deflection of the electron beam is changed, and if the stage tilt method, the tilt of the stage is changed (step 1005). Thereafter, the loop 1006 observes the calibrator again, estimates the tilt angle, and determines whether the recalculated Δθ satisfies the condition 1004. This process is repeated until the condition 1004 is satisfied, and the estimated tilt angle is converged to the set value.

5.2 Length measurement or shape index value calculation
In a semiconductor pattern dimensional measurement represented by CD measurement, or in an observation image of a target pattern from a top-down direction or an inclination direction as disclosed in Japanese Patent Laid-Open No. 2004-022617, the semiconductor pattern is based on the brightness profile. An index value having a high correlation with the three-dimensional shape of the semiconductor pattern can be obtained from the dimension measurement value of an arbitrary shape portion. However, an error in setting the tilt angle during observation results in an error in the index value. According to the calibration method, (1) the dimension value is corrected by the tilt angle estimated by the calibrator, or (2) the observation image of the target pattern imaged from the observation direction equal to the set value by the tilt angle calibration described above. High measurement accuracy can be realized by using the measurement.

As a processing example of the above (1), a method of correcting the index value when estimating the side wall inclination angle ω of the cross section 1101 of the measurement object having a step shown in FIG. 19 will be described. Value dimensions measured in the sidewall portion based on the brightness profile of the top-down image of the observed (the observed length) and W _1. When the height H of the step is known, the side wall inclination angle ω is given by tan ⁻¹ (H / W ₁ ). However, when an angle error of Δθ is included in the observation direction, the observation length error ΔW resulting from the angle error is added to the side wall observation length, and W ₂ = W ₁ + ΔW. According to the calibration method, the angle error Δθ can be estimated, and the observation length error ΔW can be obtained by Htan (Δθ). Therefore, from W ₁ = W ₂ −ΔW, the side wall observation length W ₁ without an angle error in the observation direction can be obtained, and the accurate side wall inclination angle ω can be estimated.

  In the above processing example, the dimension value in the top-down image is measured, but the same applies to the dimension measurement in the tilt image, and the dimension of an arbitrary shape portion of the semiconductor pattern in the tilt image from an arbitrary observation direction. It is conceivable to correct the measurement value.

5.3 3D reconstruction
In three-dimensional profile reconstruction by stereo measurement using observation images of target patterns from multiple directions, (1) use the tilt angle estimated by the calibrator as an input value for the observation direction, or (2) calibrate the tilt angle described above. By using the observation image of the target pattern imaged from the observation direction equal to the set value and the set value of the tilt angle as input values, high reconstruction accuracy can be realized.

FIG. 16 shows a method for calculating a step (height) between two points P ₁ -P ₂ in a cross section 1101 of a measurement target having a step as an example of an algorithm for reconstructing a three-dimensional profile. For example, an absolute coordinate system 201 in the SEM is used as a reference for measuring a step, and the height of the absolute coordinate system 201 in the z-axis direction is measured as a step. In FIG. 16A, L ₁ and L ₂ represent distances on the image between the two points P _{1 and} P ₂ when the beam tilt angles are θ ₁ and θ ₂ , respectively, when observed from above. At this time, the step H between the two points P ₁ -P ₂ is calculated by using the above-mentioned L ₁ , L ₂ , θ ₁ , θ ₂ by geometric calculation, H = {(L ₁ −L ₂ ) · cos θ ₁ · cos θ ₂ } / Sin (θ ₁ −θ ₂ ).

Similarly, M ₁ and M ₂ in FIG. 16B represent distances on the image between two points P _{1 and} P ₂ when the stage tilt angles are θ ₁ and θ ₂ , respectively, when observed from above. At this time, the level difference H between the two points P ₁ -P ₂ is calculated by using the above M ₁ , M ₂ , θ ₁ , θ ₂ by geometric calculation, H = (M ₁ · cos θ ₂ −M ₂ · cos θ ₁ ) / It is given by sin (θ ₁ −θ ₂ ). Since the input error of tilt angles θ ₁ and θ ₂ affects the reconstruction of the three-dimensional profile in both beam tilt observation and stage tilt observation, reconstruction accuracy can be improved by accurately estimating and inputting the observation direction in the observed image. It is possible to improve.

Further, in FIG. 16 (a), (b) , there is shown a measurement example of the step H between the points P ₁ -P _2, inside and outside between the points P ₁ -P _2, more finely, subject Irregularities on the pattern surface can be measured. FIG. 16C illustrates a method of measuring the points P _{1 to} P ₇ on the target pattern surface using a beam tilt image. Step H _5-6 between the points P ₅ -P ₆ as an example, the beam tilt angle theta _1, the distance on the image between the two points P ₅ -P ₆ when observed from above as theta ₂ L _{1, 5-6,} with L _2,5-6, as in the case of FIG. _{16 (a), H 5-6 =} {(L 1,5-6 -L 2,5-6) · cosθ 1 · cosθ ₂ } / sin (θ ₁ −θ ₂ ). From the obtained shape measurement results, measure the subtle pattern shapes such as the target pattern height, line width, and sidewall inclination angle, as well as corner roundness, and detect and control semiconductor pattern quality inspection and semiconductor process variations. Is possible.

FIG. 16D shows points P _{1 to} P ₇ (indicated by white circles in FIG. 16D) on the target pattern surface 1101 obtained from design data and the heights of the points P _{1 to} P ₇ . Are measured using tilt images, respectively. Measurement points P ′ _{1 to} P ′ ₇ (indicated by black circles in FIG. 16D) 1102 is a multipoint approximate shape of the surface of the target pattern in which the black circles are connected by straight lines. Is displayed in an overlapping manner. In this example, it can be seen that the shape of the completed target pattern (the shape of the measurement data) has a lower step than the design value, and the corners near the point P ₁ are rounded. Thus, the target pattern state can be evaluated by displaying the shape of the design data and the shape of the measurement result side by side or overlapping each other, or by displaying the shape such as the height and rounded corners numerically. Further, by performing such measurement at multiple points on the surface of the target pattern two-dimensionally in the x and y axis directions of the absolute coordinate system 201, for example, a three-dimensional shape measurement such as a measurement shape 1205 in FIG. You can also get results.

5.4 GUI display
The estimated value of the tilt angle obtained by the calibrator can be used to display the estimated value of the actual tilt angle on the GUI, or the beam tilt image can be converted into a stage tilt image and displayed.

FIG. 17A shows an example of a GUI (Graphic User Interface) that displays an observation image and a tilt angle estimation result. Image displays 1201 and 1202 are beam tilt images having setting values of 5 ° and 10 ° in the observation direction, respectively (hereinafter referred to as image 01 and image 02, respectively). Image displays 1203 and 1204 are obtained by estimating and displaying the stage tilt image at the set tilt angle or the estimated tilt angle by calibration from the beam tilt images 1201 and 1202. Without actually observing the stage tilt image, the stage tilt image can be estimated and displayed from the beam tilt image and the observation direction.
Since the beam tilt image obtained by the deflection of the electron beam is geometrically different from the image seen when the human observes from an oblique direction, it is difficult to interpret sensuously. Therefore, it is effective in visual analysis to convert and display a stage tilt image that is closer to human observation. When a stage tilt image is observed, it can be converted into a beam tilt image and displayed. Further, the image display 1201 to 1204 can display an image subjected to image distortion correction using the result of image distortion estimation shown in FIG. 7C as an example.

Reference numeral 1205 denotes a three-dimensional display of a three-dimensional profile estimated from an observation image, and the three-dimensional profile can be displayed from an arbitrary observation direction.
Reference numeral 1206 denotes a vector that three-dimensionally represents the set value of the observation direction in the images 01 and 02 and the estimated value of the observation direction using the calibrator. The vector components (set values and estimated values of tilt angles in the x and y directions) are displayed at 1207. Further, the display of the three-dimensional profile and the display of the observation direction can be displayed two-dimensionally as shown in FIGS. 17B and 1208 and 1209, respectively. In the figure, reference numerals 1208 and 1209 are projections of the estimated three-dimensional profile and observation direction on the xz plane as an example. (1) to (4) in 1206, 1207, and 1209 correspond to each other. Further, an error between the set value and the estimated value of the observation direction can be displayed, the observation image of the calibrator and the inclination direction of the calibrator and the wafer can be displayed.

  The beam tilt image (1201, 1202) or stage tilt image (1203, 1204) of the target pattern, the estimated three-dimensional profile (1205) of the target pattern, or the three-dimensional display (1206) or observation of the observation direction as described above. The two-dimensional display of the direction (1209 or 1102 in FIG. 16D), the numerical display of the observation direction (1207), or the design shape of the target pattern obtained from the design data (1101 in FIG. 16D) is , Any combination can be displayed on the same GUI.

  In the above description, the method of calibrator placement and the method of detecting the geometric shape deformation of the calibrator in the observed image, mainly in the case where tilt observation is performed by the beam tilt method and the case where a pyramid type sample is used as the calibrator. The observation direction estimation method, the imaging / observation direction estimation sequence, and the like have been described. However, the present invention is not limited to this, and can be similarly applied to other tilt observation methods and calibrators.

  In the present invention, accurate observation and measurement can be realized by using the tilt observation image of the target pattern and the accurate estimated value of the observation direction. Even in the case of an SEM device having a tilt angle hysteresis characteristic in which the actual tilt angle varies each time it is set even if the set value is the same, or with poor reproducibility, the tilt angle is estimated and calibrated. Accurate and reproducible observation and measurement are possible. Furthermore, even if the same sample is observed and measured with different SEM devices, the difference in tilt angle between the different SEM devices is estimated, and the same measurement result can be obtained by correcting the machine difference due to the difference in tilt angle. It becomes like this. The measurement includes, for example, dimension measurement such as line width and contact hole diameter, or the above-described three-dimensional profile measurement, etc., and a highly accurate measurement value is an effective clue for detecting a change in a semiconductor process. Furthermore, tilt observation is particularly effective when measuring a target pattern called a reverse taper with an inclination angle of 90 ° or more (when observed from directly above, there is an unobservable region). The observation / measurement method of the present invention is important for highly accurate analysis.

1 is a diagram showing one embodiment of a system for realizing the present invention. It is a figure which shows the flowchart of tilt angle estimation. It is a figure which shows the method of imaging the signal amount of the electron discharge | released from on a semiconductor wafer. It is a figure which shows the variation of the sample using the crystal plane which is one Example of a calibrator with a known shape. It is a figure which shows the modification of the pyramid sample which is one Example of a calibrator with a known shape. It is a figure which shows the observation method of a calibrator, the distribution of the brightness value in a calibrator image, and the detection method of a line segment. It is a figure which shows the variation of the arrangement method of a calibrator. It is a figure which shows distortion correction of an image. It is a figure which shows the positional relationship of the incident direction of a wafer surface, a calibrator, and an electron beam. It is a figure which shows the concept of the tilt angle estimation method. It is a figure which shows one Example of the image acquisition sequence at the time of tilt angle estimation. It is a figure which shows one Example of the procedure which estimates the inclination direction of a calibrator which combined image processing, SPM, and a focus position measurement. It is a figure which shows the sequence which images a target pattern and a calibrator. It is a figure which shows the sequence which preserve | saves the captured image of a calibrator in a database, and refers the information of the said database for estimation of the incident direction of an electron beam. It is a figure which shows the calibration method of a tilt angle. It is a figure which shows the principle of the height measurement by stereo vision. It is a figure which shows one Example of GUI which displays the estimation result of a tilt angle. (A) is a perspective view of a pole type sample, (b) is a perspective view of a hole type sample, (c1) is an image of a pole type sample observed from above, and (c2) is a hole type sample from above. The observed image, (d), is a cross-sectional view of the hole 1807 and the pyramidal calibrator 1808. It is a figure which shows the correction method of the dimension measurement value of a semiconductor pattern.

Explanation of symbols

1301 ... Semiconductor wafer 1302 ... Electro-optic system 1303 ... Electron gun 1304 ... Primary electron 1305 ... Condenser lens 1306 ... Deflector 1307 ... ExB deflector 1308 ... Objective lens 1309 ... Secondary electron detector 1310, 1311 ... Reflected electron detector 1312 ~ 1314 ... A / D converter
1315 ... Processing / control unit 1316 ... GUI screen 1317 ... Stage
1319 ... Calibration sample 1320 ... Stage controller 1321 ... Deflection control unit 13151 ... Image memory 13152 ... CPU 101 ... Concave pyramid type sample
102 ... Convex pyramid sample 103 ... Mountain sample
104 ... Line and space type sample 701 ... Stage 702 ... Semiconductor wafer
703 ... Holder 710, 711 ... Pyramid type sample
712 ... Mountain type or line and space type sample

Claims (20)

The sample is irradiated with a focused electron beam and scanned, and generated from the sample by the irradiation 2
A method of observing the sample using an SEM image obtained by detecting secondary electrons or reflected electrons,
An SEM image of the surface of the calibration substrate on which the pattern having the known shape is formed by irradiating and scanning the surface of the calibration substrate on which the pattern having the known shape is formed by irradiating and scanning an electron beam focused from an oblique direction. Acquired,
Using the information of the shape of the pattern whose shape is known and the information of the acquired SEM image, obtain the angle of the oblique direction irradiated with the electron beam,
Adjust so that the obtained angle in the oblique direction becomes a desired angle,
A sample substrate on which a pattern is formed is irradiated with the electron beam from the adjusted angle in the desired oblique direction to obtain an SEM image of the sample substrate,
A method for observing a sample, characterized in that a three-dimensional image of a pattern on the sample substrate or a cross-sectional shape of the pattern is obtained by processing an SEM image of the sample substrate using information on the desired angle. The sample is irradiated with a focused electron beam and scanned, and generated from the sample by the irradiation 2
A method of observing the sample using an SEM image obtained by detecting secondary electrons or reflected electrons,
An SEM image of the surface of the calibration substrate on which the pattern having a known shape is formed by irradiating and scanning an electron beam focused on the surface of the calibration substrate on which the pattern having a known shape is formed, from an oblique direction. Acquired,
Using the information of the shape of the pattern whose shape is known and the information of the acquired SEM image, obtain the angle of the oblique direction irradiated with the electron beam,
SEM image of the sample substrate is obtained by irradiating the electron beam from the obtained oblique direction angle to the sample substrate on which the pattern is formed,
A three-dimensional image of the pattern on the sample substrate or a cross-sectional shape of the pattern is obtained by processing the SEM image of the sample substrate using information on the angle obtained from the SEM image of the pattern whose shape is known. The sample observation method. 3. The sample observation method according to claim 1, wherein the electron beam focused on the surface of the calibration substrate is irradiated from an oblique direction by tilting the electron beam. 3. The sample observation method according to claim 1, wherein the irradiation of the electron beam focused on the surface of the calibration substrate from an oblique direction is performed by tilting the calibration substrate. Irradiating the electron beam focused on the surface of the calibration substrate from an oblique direction is performed by tilting the electron beam, and an SEM image of the acquired sample substrate is expressed as the beam tilt with the calibration substrate. 3. The sample observation method according to claim 1, wherein the sample is converted into an image obtained when tilted at the same angle and displayed on a screen. An electron beam focused on the surface of the sample substrate on which the pattern is formed is irradiated and scanned from an oblique direction to detect secondary electrons or reflected electrons generated from the sample substrate and acquire an SEM image of the pattern. A method for observing a sample, comprising: a step; and obtaining a three-dimensional image of the pattern or a cross-sectional shape of the pattern on the sample substrate by processing the SEM image of the pattern acquired in the step of acquiring the SEM image In the step of acquiring an SEM image of the pattern, the shape obtained by irradiating and scanning the focused electron beam from an oblique direction onto the surface of the substrate on which the pattern having a known shape is formed is known. Electrons that irradiate a pattern with a known shape using information obtained by processing an SEM image of the pattern The substrate on which the electron beam or the pattern having the known shape is formed is adjusted so that the angle in the oblique direction of the beam becomes a desired angle, and the angle in the oblique direction of the irradiated electron beam becomes the desired angle. In the step of obtaining an SEM image of the pattern by irradiating the surface of the sample substrate with the adjusted state and obtaining the cross-sectional shape of the pattern, the information of the desired angle of the pattern is used in the step of obtaining the sectional shape of the pattern. A method for observing a sample, comprising processing an SEM image. An electron beam focused on the surface of the sample substrate on which the pattern is formed is irradiated and scanned from an oblique direction to detect secondary electrons or reflected electrons generated from the sample substrate and acquire an SEM image of the pattern. A method for observing a sample, comprising: a step; and obtaining a three-dimensional image of the pattern or a cross-sectional shape of the pattern on the sample substrate by processing the SEM image of the pattern acquired in the step of acquiring the SEM image In the step of obtaining the cross-sectional shape of the pattern, the shape obtained by irradiating and scanning the focused electron beam from an oblique direction onto the surface of the substrate on which the pattern having a known shape is formed is known. Using the information on the angle of the oblique direction irradiated with the electron beam obtained by processing the SEM image of the pattern, The method of observing a sample and obtains the cross-sectional shape of the three-dimensional image or the pattern of the pattern of the sample substrate by treating the pattern SEM image of the obtained sample substrate in the step of obtaining the SEM image. 8. The electron beam focused on the surface of the substrate on which the pattern having a known shape is formed is obliquely applied by tilting the electron beam. Sample observation method. The electron beam focused on the surface of the substrate on which the pattern with the known shape is formed is irradiated from an oblique direction by tilting the substrate on which the pattern with the known shape is formed. The sample observation method according to claim 6 or 7. The electron beam focused on the surface of the substrate on which the pattern having a known shape is formed is irradiated by tilting the electron beam, and an SEM image of the acquired sample substrate is obtained by the shape. A method for observing a sample according to claim 6 or 7, wherein a substrate on which a known pattern is formed is converted into an image obtained when the substrate is tilted at the same angle as the beam tilt and displayed on a screen. . 8. The sample observation method according to claim 1, wherein the pattern having a known shape is a pattern having a crystal plane exposed by etching a material. The pattern having a known shape was exposed by etching silicon (Si) (111).
8. The sample observation method according to claim 1, wherein the sample is a pattern formed by exposing a surface or a surface in a direction equivalent to the (111) surface. The pattern having a known shape is a concave quadrangular pyramid-shaped pattern, and an inclination angle of each surface of the quadrangular pyramid is known. The observation method of the sample of description. The pattern having the known shape is a pattern formed by exposing the (111) plane exposed by etching the silicon (Si) or the equivalent direction to the (111) plane after measuring the shape of the pattern. The method for observing a sample according to any one of claims 1, 2, 6 and 7. 8. The sample according to claim 1, wherein the pattern having a known shape is a mountain-shaped pattern, and an inclination angle of each surface of the mountain-shaped is known. Observation method. The sample observation according to claim 1, wherein the pattern having a known shape is placed in the vicinity of the sample substrate on a table on which the sample substrate is placed. Method. An apparatus for observing a sample using an SEM image,
Table means on which a sample can be placed and set to a desired tilt angle;
SEM image acquisition means for acquiring a SEM image obtained by irradiating and scanning a focused electron beam onto a sample placed on the table means and detecting secondary electrons or reflected electrons generated from the sample by the irradiation. When,
Image display means for processing the image acquired by the SEM image acquisition means and displaying it on a screen;
Control means for controlling the SEM image acquisition means;
Calculating means for obtaining an incident angle of the focused electron beam incident on the substrate on which the pattern having the known shape is formed from the SEM image of the substrate on which the pattern having the known shape acquired by the SEM image acquiring means is formed; With
The image display means processes the SEM image of the sample acquired by the SEM image acquisition means based on the information on the incident angle of the focused electron beam obtained by the calculation means, thereby obtaining a three-dimensional shape or a cross section of the sample. An observation apparatus characterized in that a shape is obtained and information on the obtained three-dimensional shape or cross-sectional shape of the sample is displayed on a screen. The observation apparatus according to claim 17, wherein the control means controls the SEM image acquisition means to set a tilt angle of the table means to a desired angle. 18. The observation apparatus according to claim 17, wherein the control unit controls the SEM image acquisition unit to set a tilt angle of the focused electron beam that irradiates the sample to a desired angle. The image display means sets the SEM image of the sample acquired by setting the tilt angle of the electron beam focused by the SEM image acquisition means to a desired angle, and the substrate on which the pattern having a known shape is formed. 18. The sample observation apparatus according to claim 17, wherein the sample observation apparatus converts the image into an image obtained when the electron beam is tilted at the same angle as that when the electron beam is tilted and displays the image on a screen.

Images