JP2009043567A - Sample image rotating method, and charged particle beam device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a sample image rotating method capable of correcting visual field deviation based on accurate visual field deviation detection; and a charged particle beam device. <P>SOLUTION: In the charged particle beam device provided with a means for rotating an image of a sample mounted on a sample base as one embodiment for solving the above problem, the sample image rotating method selects part of an image before rotating the sample image as a template, searches the inside of the image after rotating the sample by using the selected template, and corrects visual field deviation based on the distance between a part from which the template has been selected and a part which the image after rotating the sample has matched. The charged particle beam device is also provided. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a sample image rotation method and a charged particle beam apparatus, and more particularly, to a method and apparatus for measuring or correcting a visual field shift that occurs when a sample image is rotated.

  In an electron microscope, a sample image rotation technique is known in which an image is observed by rotating a visual field. When rotating the sample image, for example, the sample stage is centered on the rotation axis in the direction of the electron beam optical axis (the ideal optical axis of the electron beam when the electron beam is not deflected, hereinafter sometimes referred to as the z axis). Rotate the sample image by changing the excitation condition of the electromagnetic objective lens, or in the case of a scanning electron microscope, rotate the scanning direction of the electron beam, so-called raster rotation. There is technology.

  In order to correct the visual field shift that occurs when the sample image is rotated by changing the excitation current of the objective lens, the amount of movement is obtained based on the correlation between the sample images before and after the rotation. Japanese Patent Application Laid-Open No. H10-228561 describes a technique for calculating a deflection correction coil for deflecting the movement of the sample stage or a movement correction amount of the sample stage.

JP 2001-118535 A

  In particular, when the image is rotated by the sample stage, the observation target may escape from the observation field due to the mechanical accuracy of the sample stage. In addition, a visual field shift that occurs when the excitation condition of the objective lens is changed is also conceivable. With the technique disclosed in Patent Document 1, it is possible to effectively correct such a visual field escape, but there are the following problems to be solved.

  When the sample image is rotated, for example, in the case of an apparatus that performs rectangular scanning in the xy direction (a direction perpendicular to the z-axis direction), in the pre-rotation image and the post-rotation image, the region deviated from the pre-rotation image Since there is a newly included region, the field of view (FOV) does not match. That is, even if the field of view is not deviated at all, the degree of coincidence between the two images of the sample image before and after the rotation differs due to the change in the irradiation region of the electron beam caused by the rotation. There is a problem that you can not. Further, when the field of view is shifted before and after the rotation, the difference in the degree of coincidence between the two images becomes more prominent, and there is a problem that it is difficult to correct the field of view based on the specification of an appropriate amount of field shift.

  Hereinafter, a sample image rotating method and a charged particle beam apparatus capable of correcting the field deviation based on the detection of the field deviation with high accuracy, which solve the above-described problems, will be described.

  As one aspect for achieving the above object, in a charged particle beam apparatus having means for rotating an image of a sample placed on a sample stage, a part of the image before rotating the sample image is selected as a template, and the selection is performed. Rotate the sample image to search the image after rotating the sample using the template, and correct the visual field deviation based on the distance between the location where the template was selected and the matched location of the image after rotating the sample. A method and a charged particle beam apparatus are proposed.

  According to the above configuration, even if the beam irradiation area changes due to sample image rotation, or the beam irradiation area shifts due to sample field rotation due to sample image rotation, Deviation correction can be performed. Other specific configurations and effects of the present invention will be described in detail in the section of the best mode for carrying out the invention.

  Hereinafter, a technique for effectively correcting a visual field shift during sample image rotation in a charged particle beam apparatus capable of sample image rotation will be described. A specific processing method used at that time will be described below.

(1) Means for calculating movement amount by phase-only correlation The operation of the operation electron microscope having the above-described configuration will be described using an example of image correlation shown in FIG. An image obtained by cutting out a part of the scan image 1 is recorded as a scan image with a number of pixels of M × N as a registered image as f1 (m, n). Next, an image captured after the recording mode is recorded as a reference image f2 (m, n) in the storage device with the number of pixels of M × N.

  However, both are natural images, and m = 0, 1, 2,... M−1, n = 0, 1, 2,.

  Discrete Fourier images F1 (m, n) and F2 (m, n) of f1 (m, n) and f2 (m, n) are defined by (1) and (2), respectively.

However, u = 0, 1, 2,... M-1, v = 0, 1, 2,.
A (u, v) and B (u, v) are amplitude spectra, and θ (u, v) and φ (u, v) are phase spectra.

  In phase correlation, when there is parallel movement of an image between two images, the position of the correlation peak is shifted by the amount of movement.

  A method for deriving the movement amount will be described below.

  First, assuming that the original image f2 (m, n) has moved by r ′ in the x direction, f4 (m, n) = f2 (m + r ′, n).

  Equation (2) is transformed into Equation (3).

  By setting the amplitude spectrum B (u, v) as a constant, a phase image independent of the contrast of the image is obtained. The phase image F′4 (u, v) of f4 is expressed by Equation (4).

  By multiplying the phase image F′1 (u, v) by the complex function of F′2 (u, v), the composite image H14 (u, v) is expressed by equation (5).

  The correlation strength image G14 (r, s) is expressed by Equation (6) by performing inverse Fourier transform on the composite image H14 (u, v).

  From equation (6), when there is a positional shift amount r ′ in the X direction between two images, the peak position of the correlation intensity image is shifted by −r ′. In addition, since the correlation calculation is performed using the phase component, the movement amount can be calculated even if there is a difference in brightness or contrast between the two images. When there is a positional deviation amount in the X direction between two images, a peak occurs at a position of ΔG (pixel) from the center of the correlation strength image. For example, if there is a shift of 2 pixels in the X direction between two images, the composite image becomes a two-cycle wave. When this is subjected to inverse Fourier transform, a correlation intensity image is obtained, and a peak is generated at a position shifted by 2 pixels from the center. This ΔG (pixel) corresponds to a movement amount on the light receiving surface of the detector, and ΔG is converted into a movement amount Δx on the sample surface. Assuming that the diameter L of the light receiving surface of the detector, the magnification M of the transmission electron microscope on the light receiving surface, and the number of pixels Lm of the light receiving surface of the detector are given by equation (7).

  Δx is the amount of movement on the sample surface between the two images.

(2) Means for calculating the degree of coincidence Next, the amount of movement between images, the magnification, and the accuracy of the rotation angle will be described. In the correlation calculation using only the phase component, since only the phase is used mathematically, the peak appearing in the correlation strength is the δ peak. For example, if the image is shifted by 1.5 pixels between two images, the composite image becomes a wave of 1.5 cycles. When this is subjected to inverse Fourier transform, a δ peak appears at a position shifted by 1.5 pixels from the center of the correlation intensity image, but since there is no 1.5 pixel, the value of the δ peak is assigned to the first pixel and the second pixel. Here, taking the center of gravity of the pixel having a high degree of coincidence and calculating the true δ peak position from the assigned value, the calculation result can be obtained with an accuracy of about 1/10 pixel. Further, since the correlation intensity image is δ peak, the similarity between the two images is evaluated based on the peak height of the correlation intensity image. When the image f1 (m, n) and the peak height Peak (pixel) are assumed, the degree of coincidence (%) is shown in Expression (8).

  For example, when the number of processed pixels is 128 pixels × 128 pixels and Peak is 16384 (pixels), the degree of coincidence = (16384) / (128 × 128) × 100 = 100 (%).

(3) Processing Using Pattern Matching An example of template image search by image template matching is shown below.

  The correlation calculation represented by Expression (9) is performed on all the pixels in the designated area of the source image, and the point at which the matching coefficient (r) is maximum (1.0) is detected as the movement amount. At this time, the degree of coincidence is defined as r multiplied by 100.

ｆ：ソース画像
ｇ：テンプレート画像
ｎ：テンプレート領域内有効画素数
（１＜ｎ＜＝６５５３６：２５６×２５６相当）

f: source image g: template image n: number of effective pixels in template area (1 <n <= 65536: 256 × 256 equivalent)

  If this method is used, since the calculation formula itself for calculating the correlation coefficient normalizes the data, the degree of coincidence increases with respect to variations in brightness and blurring.

  FIG. 16 shows the relationship between the conventional blur (defocuser) and the movement amount. FIG. 17 shows the relationship between the blur and the movement amount of the present invention. From these results, it can be seen that the method of the present invention has a high recognition degree against blur.

  These operations are performed on one area of the source image corresponding to the area of the template image. In the normalized correlation search according to the present invention, three stages of setup, training, and search are set. In the setup, the template image is cut out from the input image, and the training is performed by using the template image of the normalized correlation search. Register as Next, in the search, a template registered in the training is searched. In the movement amount calculation, the movement position is calculated as shown in FIG. 10, and the calculation is performed with sub-pixel accuracy as shown in FIG. In the automatic adjustment, since the image moves at the center of the image, searching with a spiral as shown in FIG. 11 improves the search efficiency.

(4) Processing Using Neural Network As shown in FIGS. 8 and 9, the compression / decompression neural network is performed under the condition of reducing the difference between the input image and the restored image. However, there are a plurality of pixels having the same gray value on the smoothed image. Therefore, this conversion method is realized by minimizing a cost function to which a constraint condition that smoothes the distortion of the grid is added in order to solve a unique solution by the re-steep descent method. Equation (10) shows the cost function.

In the above formula, I is a resampled image, R is a restored image, and (dx _{i, j,} dy _{i, j} ) is an estimated value of a movement vector in the grid (i, j).

(5) Method of automatically correcting focus The ΔX calculated by the equation (7) is put into the equation (11) to calculate the defocus amount Δf.

(6) Stage Correction Method of the Present Invention As shown in FIG. 14, ΔX and ΔY to be corrected can be corrected by equations (12) and (13).

  Hereinafter, a specific example of visual field shift correction using the above arithmetic expression will be described with reference to the drawings. First, an embodiment of the present invention will be described with reference to the flowchart of FIG. FIG. 1 is a schematic configuration diagram of a scanning electron microscope which is one of charged particle beam apparatuses. The following description relates to a scanning electron microscope, but is not limited thereto, and can be applied to other charged particle beam devices such as a focused ion beam device that narrows and scans an ion beam. is there.

  As shown in FIG. 1, in the scanning electron microscope described in FIG. 1, the electron beam emitted from the electron gun 1 is passed through the first irradiation lens coil 2, the second irradiation lens coil 3, and the objective lens 6. Irradiate the sample placed on the sample stage 7. At this time, the electron beam is scanned by deflection by the first deflection coil 4 and the second deflection coil 5. The sample stage 7 is supported by an unillustrated moving mechanism so as to be movable with respect to the electron beam. The moving mechanism is a driving mechanism that can move in the xy direction, z direction, which is the vertical direction of the electron beam optical axis, and the rotational direction having a parallel rotation axis in the z direction. Is provided. The moving mechanism is controlled by a processing device such as a microprocessor 18 or the like.

  FIG. 7 is a diagram illustrating an example of a sample stage. The stage including the rotation mechanism 34, the X movement mechanism 36, and the Y movement mechanism 37 is driven by a signal supplied from the stage controller 35.

  Further, an image shift deflector for deflecting the irradiation position of the electron beam is provided (not shown). The image shift is not performed by providing a dedicated deflector, but by superimposing a deflection signal on the first deflection coil 4 and the second deflection coil 5 and performing scanning and image shift with the two coils. Also good.

  Each lens for focusing the electron beam is connected to a respective excitation power source 8, 9, 12. The ROM 29 storing the scanned image observation lens data outputs the data to the DACs 13, 14, and 17 to convert the lens system data into analog signals. Analog signals are output from the DACs 13, 14, 17 to the excitation power supplies 8, 9, 12, and currents are output to the lens coils 2, 3, 6 of each lens system.

  Secondary electrons are emitted by the electron beam irradiated on the sample, and the secondary electrons are detected by the secondary electron detectors 32 and 33. Image data is stored in the memory 31 and an image is displayed on the monitor 22. Next, as shown in FIG. 5, the sample rotation start angle and end angle are input.

  Next, rotation is started using the sample stage of the present invention. An image is captured and recorded as template 1. Next, the search area is also captured and recorded. At this time, in the arithmetic unit 20 or the like, the visual field escape may be roughly corrected by correcting the X and Y of the stage using the above equation (6).

  The search area specifies a search range using the template after image rotation. The template is set smaller than the image before rotation. This is because the field of view common to the image before rotation and the image after rotation is limited. Therefore, if the portion that changes before and after the rotation is included in the field of view, the matching degree of template matching is lowered, and high accuracy is achieved. This is because it is difficult to measure the degree of coincidence.

  Further, the template image may be rotated in accordance with the actual rotation angle in order to improve the matching degree with the image after rotation.

  Next, rotation starts. Correlation is calculated using the above formulas (3) and (4), and the degree of coincidence and the movement amounts of X and Y are calculated. At this time, the calculation may be performed using equations (1) and (2). Next, when the degree of coincidence is 70% or less, the process ends. In the case of 70% or more, correction is performed by adding the movement amount to the stages X and Y. At this time, image shift may be used together. Next, when the angle is not set, correction is repeatedly performed. When the set angle is reached, the process ends. It is also possible to capture an image after reaching the set angle and display and record the degree of coincidence with the image as shown in FIG.

  An embodiment of the present invention will be described below with reference to the flowchart of FIG. As shown in FIG. 1, the lens data for scanning image observation is output from the ROM 29 storing the lens data to each of the DACs 13, 14, and 17 to convert the lens system data into analog signals. Analog signals are output from the DACs 13, 14, 17 to the excitation power supplies 8, 9, 12, and currents are output to the lens coils 2, 3, 6 of each lens system. Secondary electrons are emitted by the electron beam irradiated on the sample, and the secondary electrons are detected by the secondary electron detectors 32 and 33. Image data is stored in the memory 31 and an image is displayed on the monitor 22. Next, as shown in FIG. 5, the sample rotation start angle and end angle are input.

  Next, rotation is started using the sample stage of the present invention. An image is captured and recorded as template 1. Next, the search area is also captured and recorded. At this time, the visual field escape may be roughly corrected by correcting the X and Y of the stage using the equation (6). At this time, the calculation may be performed using equations (1) and (2). Next, rotation starts. Correlation is calculated using equations (3) and (4), and the degree of coincidence and the amount of movement of X and Y are calculated. Next, when the degree of coincidence is 70% or less, the rotation angle of the stage is returned and correction is repeatedly performed so that the degree of coincidence exceeds 70%. In the case of 70% or more, correction is performed by adding the movement amount to the stages X and Y. At this time, image shift may be used together. Next, when the angle is not set, correction is repeatedly performed. When the set angle is reached, the process ends.

  It is also possible to capture an image after reaching the set angle and display and record the degree of coincidence with the image as shown in FIG.

  An embodiment of the present invention will be described below with reference to the flowchart of FIG. As shown in FIG. 1, the lens system data is converted into analog signals by outputting to the DACs 13, 14, and 17 from the ROM 29 storing the scanned image observation lens data. Analog signals are output from the DACs 13, 14, 17 to the excitation power supplies 8, 9, 12, and currents are output to the lens coils 2, 3, 6 of each lens system. Secondary electrons are emitted by the electron beam irradiated on the sample, and the secondary electrons are detected by the secondary electron detectors 32 and 33. Image data is stored in the memory 31 and an image is displayed on the monitor 22. Next, as shown in FIG. 5, the sample rotation start angle and end angle are input.

  Next, rotation is started using the sample stage of the present invention. An image is captured and recorded as template 1. Next, the search area is also captured and recorded. Next, correlation is calculated using equations (3) and (4), and the degree of coincidence is calculated. At this time, the calculation may be performed using equations (1) and (2).

  If the degree of coincidence is 70% or less, the field of view cannot be corrected and the process ends. When the degree of coincidence is 70% or more, rotation is started. As shown in FIG. 15, in the correlation calculation according to the present invention, the accuracy of the calculation of the movement amount at the time of image rotation is reduced unless the angle is 10 degrees or less. For this reason, when the rotation angle is 10 degrees or more, the rotation angle of the stage is returned and the search area is taken in again to perform correlation calculation. When the rotation angle is 10 degrees or less, correlation calculation is performed using equations (3) and (4), and the degree of coincidence and the movement amounts of X and Y are calculated. At this time, the calculation may be performed using equations (1) and (2).

  Next, the field escape may be roughly corrected by correcting X and Y of the stage using Expression (6). Next, when the degree of coincidence is 70% or less, the process ends.

  In the case of 70% or more, correction is performed by adding the movement amount to the stages X and Y. At this time, image shift may be used together. Next, when the angle is not set, correction is repeatedly performed. When the set angle is reached, the process ends. It is also possible to capture an image after reaching the set angle and display and record the degree of coincidence with the image as shown in FIG.

The schematic block diagram of a scanning electron microscope. The flowchart explaining the image rotation process (the 1). Flowchart for explaining the image rotation process (part 2). Flowchart for explaining the image rotation process (No. 3). Input screen for setting the sample rotation angle. The figure explaining the example which superimposes and displays the matching degree of matching on a scanning electron microscope image. The schematic block diagram of a sample stage. The conceptual diagram of a neural network. The figure explaining an example of a template registration screen. The figure explaining an example of the movement amount measurement between images. The figure explaining the example which performs a visual field search in spiral shape. The conceptual diagram of a phase correlation process. The figure explaining an example of the coincidence calculation method between images. The figure explaining the conceptual diagram of a stage correction. The figure explaining the relationship between a rotation angle and a coincidence degree. The figure explaining the relationship between a movement amount and a coincidence degree. The figure explaining the relationship between a movement amount and a coincidence degree (conventional example).

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Electron gun 2 1st irradiation lens coil 3 2nd irradiation lens coil 4 1st deflection coil 5 2nd deflection coil 6 Objective lens coil 7 Sample stage 8-12 Excitation power supply 13-17 DAC
18 Microprocessor 19 Storage device 20 Arithmetic device 21 CRT controller 22 Monitor 23, 24 I / F
25 Rotary encoder for switching magnification 26 Rotary encoder for input 27 Keyboard 28 RAM
29 ROM
30 Mouse 31 Memory 32, 33 Secondary electron detector 34 Rotating mechanism 35 Stage controller 36 X moving mechanism 37 Y moving mechanism

Claims (9)

  In a sample image rotation method for rotating an image of a sample placed on a sample table, the sample is irradiated with a charged particle beam to form an image, a part of the image is selected to form a template, and the template is A sample image, wherein the sample image is searched for in the rotated sample image, and field-of-view correction is performed based on a distance between the location where the template is selected and the matched location after the sample rotation. Rotation method. In claim 1,
A sample image rotation method comprising rotating the sample to a desired rotation angle by repeating the image formation, template formation, in-sample image search using the template, and visual field shift correction based on the search a plurality of times. In claim 1,
The sample image rotation method, wherein the sample image is rotated by rotating the sample stage. In a charged particle beam apparatus including a charged particle source, an objective lens that focuses the charged particle source, and a sample stage on which a sample is placed,
The sample stage has a rotation mechanism for rotating the sample, and the rotation mechanism is controlled by a processing device that controls the rotation mechanism by setting a desired sample rotation angle, and the processing is performed. The apparatus selects a part of an image formed by irradiating the sample with a charged particle beam, forms a template, searches the rotated sample image using the template, and selects the template. A charged particle beam apparatus comprising: a processing device that performs visual field shift correction based on a distance between a location and a matched location after the sample rotation. In claim 4,
The processing apparatus repeats the image formation, template formation, in-sample image search using a template, and correction of visual field deviation based on the search a plurality of times, and rotates the sample to a desired rotation angle. Particle beam device.   In a scanning electron microscope, an electron gun, a converging lens that irradiates an electron beam and a sample, a mechanism that deflects the electron beam, a sample, an objective lens that focuses the sample, a mechanism that captures a scanned image as image data, Two sample scan images at different electron beam rotation angles are used as a search area image in which the first image is searched for the entire image, and at the time of another rotation, a template image obtained by cutting the center of the image is used. A scanning electron microscope characterized by means for processing a template image to obtain an image movement amount and a correlation value. In claim 6,
As a search area image for searching the sample stage and two sample scan images at different sample rotation angles for the first image as a whole, a template image obtained by cutting the center of the image at the time of another rotation is used. A scanning electron microscope characterized by means for processing an image and the template image to obtain an image movement amount and a correlation value.   In a scanning electron microscope, an electron gun, a converging lens for irradiating an electron beam and a sample, a mechanism for deflecting the electron beam, a sample, an objective lens for focusing the sample, a sample stage, and a scanned image as image data As a search area image for searching the entire image of two sample scan images at different electron beam rotation angles with the capturing mechanism, the area smaller than the area where the blur due to the rotation angle does not occur during the other rotation A scanning electron microscope comprising means for processing a search area image and the template image using a template image to obtain an image movement amount and a correlation value.   In a scanning electron microscope, an electron gun, a converging lens that irradiates an electron beam and a sample, a mechanism that deflects the electron beam, a sample, an objective lens that focuses the sample, a mechanism that captures a scanned image as image data, Two sample scan images at different electron beam rotation angles are used as a search area image in which the first image is searched for the entire image, and at the time of another rotation, a template image obtained by cutting the center of the image is used. A scanning electron microscope having means for processing a template image to obtain an image movement amount and a correlation value and means for storing the image and the correlation value.

Images