JP5151277B2 - Sample observation apparatus and correction method thereof - Google Patents

Description

  The present invention relates to a sample observation apparatus and a correction method thereof.

  Along with miniaturization of electronic devices such as LSIs, scanning electron microscopes capable of observing the fine structure of elements are frequently used in the manufacturing process of electronic devices.

  For example, when performing failure analysis of an electronic device, after identifying the defective portion in one chip by an electrical test or OBIRCH (Optical Beam Induced Resistance Charge) method, observe the defective portion with a scanning electron microscope. The cause of the defect is elucidated.

  At this time, CAD (Computer Aided Design) data including the design layout of the chip is input to the scanning electron microscope, so that the design layout and the observation image by the electron microscope are superimposed on the scanning electron microscope side. Attempts have also been made to automatically detect defective locations and improve the throughput of failure analysis.

  The scanning electron microscope is also used for the purpose of measuring a device pattern such as a metal wiring in addition to the observation of the chip.

  By the way, in order to perform the above observation and length measurement with high accuracy, the observation image of the electron microscope must accurately reflect the element shape.

  However, in the scanning electron microscope, the observation image is distorted around the image when a wide range of observation is performed at a low magnification, so that the electron microscope image and the CAD design layout cannot be overlapped well. Such distortion of the observed image is due to the fact that the magnetic field of the deflection coil for deflecting the electron beam is not uniform in the peripheral part of the deflection range, and the deflection amount of the electron beam in that part deviates from the set value. It is thought to occur.

  Moreover, since the degree of distortion of the observation image varies depending on the magnification of the electron microscope and the CAD design layout varies in size, it is extremely difficult to superimpose the observation image and the design layout at present.

  In addition, when superimposing the design layout and the observation image as described above, the scanning electron microscope is provided with an image rotation mechanism for the purpose of eliminating the rotational deviation thereof, and the image rotation mechanism can It is known that the above distortion occurs even when a microscopic image is rotated.

  Such distortion of the observed image does not appear only in the scanning electron microscope, but also appears in a focused ion beam (FIB) apparatus having a function of deflecting the electron beam to obtain an image of the sample.

  Although it is conceivable to correct the distortion of the observation image generated by these devices by hardware adjustment, the adjustment method is complicated, and the distortion characteristics differ depending on the individual devices. There is no situation. Furthermore, when measuring a device pattern with a scanning electron microscope, measurement accuracy on the order of nanometers is required, but at high magnification, distortion of minute observation images can be corrected only by hardware adjustment. Can not. Therefore, an evaluation apparatus that can accurately measure the length from a low magnification to a high magnification has not been realized.

  An object of the present invention is to provide a sample observation apparatus and a correction method thereof that can easily reduce distortion of an observation image.

According to one aspect of the present invention, (S1) a step of applying a charged particle beam transmitted through an electromagnetic lens system to a standard sample on which a correction pattern is formed, and (S2) the charged particles applied to the standard sample A step of obtaining an observation image of the correction pattern by detecting a beam with a detector; and (S3) a plurality of references of the correction pattern due to distortion of the observation image caused by the electromagnetic lens system. a step where the points calculated whether the amount of deviation of the position in the observation image, (S4) so that the positional deviation is eliminated, possess a step of correcting the image data of the observation image, the step (S3) is For each of the plurality of reference points, the position of the reference point in the observation image is specified as an actual position by pattern recognition, and the actual position of the reference point when there is no distortion is used as the starting point. A correction method for a sample observation apparatus is provided, which is performed by calculating a displacement vector having an end point as an end point .

Further, according to another aspect of the present invention, a charged particle gun that generates and accelerates a charged particle beam, an electromagnetic lens system that focuses the charged particle beam on a standard sample on which a correction pattern is formed, and the standard sample A detector that converts the intensity of the charged particle beam applied to the detection signal into a detection signal, and an image analysis unit that creates image data from the detection signal, and the image analysis unit is caused by the electromagnetic lens system. The degree of displacement of the reference points of the correction pattern in the observation image due to the distortion of the observation image of the correction pattern generated by the The position of the reference point is specified as an actual position by pattern recognition, and a positional deviation vector having the ideal position of the reference point when there is no distortion as the start point and the actual position as the end point Out, the sample observation apparatus is provided, which comprises correcting the image data such that the positional deviation is eliminated.

  Next, the operation of the present invention will be described.

  According to the present invention, since the image data of the observation image of the correction pattern is corrected so as to eliminate the positional deviation of the correction pattern caused by the electromagnetic lens system, the image rotation mechanism of the sample observation apparatus is used. In addition, it is possible to easily prevent distortion of the observation image that occurs when observing at a low magnification.

  Further, it is preferable to use a grid pattern composed of a plurality of grid lines as the correction pattern, and use an intersection of the grid lines as the reference point. In this case, each observation image of the grid line is mapped to a point in the Hough space by Hough transformation, and the position of the intersection of the plurality of grid lines can be calculated from the position coordinates of the point. In the Hough transform, even if the grid line is slightly curved due to image distortion, it is easy to obtain the coordinates of the grid point to capture the straight line in the real space as a point in the Hough space. Can do.

  Further, the reference point position shift is an observation condition including at least one of an acceleration voltage of the charged particle beam, a deflection amount of the charged particle beam by the electromagnetic lens system, a magnification of the observation image of the correction pattern, and a rotation angle of the observation image. You may associate and store in a database.

  In this way, when observing the observation sample under a specific observation condition, the observation sample is referred to the positional deviation in the database corresponding to the observation condition, so that the observation image of the observation sample is free from distortion. The image data can be corrected.

  According to the present invention, the distortion of the observation image can be easily reduced by correcting the image data so that the positional deviation of the reference point of the correction pattern is eliminated.

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

  FIG. 1 is a configuration diagram of a sample evaluation apparatus according to an embodiment of the present invention. Hereinafter, a scanning transmission electron microscope (STEM) will be described as a sample evaluation apparatus. However, the present invention is not limited to this, and the present invention can also be applied to a scanning reflection electron microscope (SREM) and an FIB apparatus.

  This sample evaluation apparatus 10 includes an electron gun 12, an electromagnetic lens system 13, a scanning / deflection coil 14, a diaphragm 15, a sample mounting unit 16, and a STEM detector 17 in a casing 11 whose inside is decompressed. Become. Among these, the electromagnetic lens system 13 includes first and second convergence lenses 13a and 13b and an objective lens 13c.

  The electron beam (charged particle beam) EB generated and accelerated in the electron gun 12 is made into an electron beam probe having a desired size and a desired current amount by the first and second converging lenses 13a and 13b. Since the electron beam probe formed by the first and second converging lenses 13 a and 13 b has an undesirably widened portion, the widened portion is cut by the diaphragm 15.

  The scanning / deflection coil 14 refracts the electron beam EB so that the electron beam EB scans the surface of the standard sample S mounted on the sample mounting unit 16. Further, the objective lens 13c refracts the electron beam EB so as to be focused on the surface of the standard sample S.

  The intensity of the electron beam EB transmitted through the standard sample S is converted into a detection signal I by the STEM detector 17 and is captured by the image analysis unit 19.

  The image analysis unit 19 captures the detection signal I in synchronization with the scanning speed of the scanning / deflection coil 14, thereby creating image data of an observation image of the scanned standard sample S. Based on the image data, an observation image is displayed on a CRT (not shown).

  The image analysis unit 19 also has a storage unit 20 for storing a database and the like described later.

  The standard sample S is used to calibrate the sample evaluation apparatus 10 and can be produced by forming a pattern on a silicon substrate by photolithography.

  FIG. 2 (a) is an enlarged plan view of such a standard sample S, and FIG. 2 (b) is a cross-sectional view taken along the line II of FIG. 2 (a).

  In the standard sample S shown in FIGS. 2A and 2B, a silicon oxide film is embedded as the correction pattern 101 in the groove 100 a formed in the silicon substrate 100. In such a standard sample S, for example, after a silicon oxide film is embedded as the correction pattern 101 in the groove 100a of the silicon substrate 100, the back side of the silicon substrate 100 is polished to expose the correction pattern 101 on the polished surface. Can be created.

  The planar shape of the correction pattern 101 is preferably a lattice as shown in FIG.

  By the way, the magnification range of a normal scanning transmission electron microscope or FIB apparatus is 200 to 2,000 times, 4,000 to 20,000 times, 40,000 to 200,000 times, and 400,000 to 10,000. 1,000 times. When the image size is 1024 × 1024 pixels and the pixel size is 72 dpi, the length of one frame and the size (image resolution) of one pixel in each magnification range described above are shown in Tables 1 to 4 below.

ここで、実際の測定における空間分解能は、画像分解能よりも低く、５pixel程度の誤差がある。従って、標準試料Sに形成された補正用パターン１０１をCRT上で認識するには、観察像での補正用パターン１０１の幅が５pixel程度なくてはならない。この点に鑑み、表１〜表４では、空間分解能を画像分解能の５倍にしている。

Here, the spatial resolution in actual measurement is lower than the image resolution and has an error of about 5 pixels. Therefore, in order to recognize the correction pattern 101 formed on the standard sample S on the CRT, the width of the correction pattern 101 in the observation image must be about 5 pixels. In view of this point, in Tables 1 to 4, the spatial resolution is five times the image resolution.

  It is preferable to prepare four types of standard samples S individually corresponding to these magnification ranges so that the correction pattern 101 can be resolved in each magnification range of Tables 1 to 4.

  However, in the calibration of a high magnification region of 400,000 to 10,000,000 times, it is difficult to fabricate the standard sample S by processing the silicon substrate 100 as shown in FIGS. A crystal having a known lattice constant, for example, a thin film made of a crystal material such as gold or silicon is preferably used as the standard sample S. In this case, the crystal lattice itself of these thin films becomes a lattice-shaped correction pattern.

  FIG. 3 is a principle diagram of the sample observation method in the sample evaluation apparatus 10.

  The scanning / deflection coil 14 includes coils 14a to 14c and a core 14e. The scanning direction A of the electron beam EB can be freely controlled by controlling the amount of current independently by each of the coils 14a to 14c and arbitrarily combining the polarities of these coils.

  Using this feature, when the scanning direction A is rotated by an angle θ, a rotated image obtained by rotating the observation image of the standard sample S by the angle θ can be obtained. The angle θ is a rotation angle with respect to the X axis of an arbitrary XY orthogonal coordinate system set for the standard sample S. Thus, the scanning / deflection coil 14 also functions as an image rotation mechanism.

  However, it is known that when an observation image is rotated using an image rotation mechanism, the observation image after rotation is distorted. This distortion will be described with reference to FIGS.

  FIG. 4A is a diagram illustrating the scanning direction A when the image rotation mechanism is not used. FIG. 4B is an observation image of the standard sample S obtained by scanning the electron beam EB as shown in FIG.

  As shown in FIG. 4B, when the image rotation mechanism is not used, the observed image 30 of the correction pattern 101 is not distorted.

  On the other hand, FIG. 5A is a diagram showing the scanning direction A rotated by an angle θ by the image rotation mechanism. FIG. 5B is an observation image of the standard sample S obtained by scanning the electron beam EB as shown in FIG.

  As a result of the rotation as described above, the observed image 30 of the correction pattern 101 is distorted from the normal position (dotted line) that is assumed to be free of distortion.

  Such distortion also occurs when observing at low magnification without using an image rotation mechanism. 6 and 7 are diagrams for explaining that the observation image is distorted at a low magnification.

  FIG. 6A is a diagram showing the scanning direction A when observing at high magnification. FIG. 6B is an observation image of the standard sample S obtained by scanning the electron beam EB as shown in FIG.

  When observing at a high magnification, since the scanning range of the electron beam EB is narrow, the electron beam EB is scanned with a uniform magnetic field near the center of the scanning / deflection coil 14 (see FIG. 3). Therefore, the scanning direction A has a shape in which a straight line is folded, and the observed image 30 is not distorted as shown in FIG.

  On the other hand, FIG. 7A is a diagram showing the scanning direction A when observing at a low magnification. FIG. 7B is an observation image of the standard sample S obtained by scanning the electron beam EB as shown in FIG.

  When observing at a low magnification, the scanning range of the electron beam EB is wide. Therefore, the electron beam EB is exposed to a non-uniform magnetic field at the peripheral portion of the scanning / deflection coil 14, and the scanning direction A is curved as shown in FIG.

  As a result, as shown in FIG. 7B, the observation image 30 becomes more distorted than in the case of FIG.

  In the present embodiment, the distortion of the observation image as described above is corrected as follows.

  FIG. 8 is a flowchart for explaining the correction method of the sample observation apparatus according to the present embodiment.

  The first step S1 in FIG. 8 has sub-steps T1 to T6, and the operator performs each sub-step manually.

  Among them, in sub-step T1, the acceleration voltage supplied to the electron gun 12 is set, and the lens conditions of the first and second converging lenses 13a and 13b and the aperture diameter of the aperture 13 are adjusted, so that the electron beam EB is predetermined. To converge to the size of.

  Next, the process proceeds to step T2, the optical axis alignment between the electron beam EB and the electromagnetic lens system 13 is performed, then the process proceeds to step T3, and the above-described standard sample S (see FIG. 2) is placed on the sample mounting unit 16. Place.

  Thereafter, the process proceeds to step T4, the sample mounting portion is controlled so that the electron beam EB is perpendicularly incident on the surface of the standard sample S, and the inclination angle of the standard sample S is adjusted.

Subsequently, the process proceeds to step T5, and observation conditions such as a magnification, a rotation angle by an image rotation mechanism, and an image size are set. The set observation conditions are stored in the storage unit 20 by the image analysis unit 19 is provided as the observation condition data D _1.

  Then, the process proceeds to step T6, and the objective lens 13c is focused on the surface of the standard sample S.

  After step S1 is completed in this way, the process proceeds to step S2.

  In this step, an observation image of the standard sample S is obtained by scanning the electron beam EB on the standard sample S and capturing the electron beam EB transmitted through the standard sample S with the STEM detector 17 in synchronization with the scanning speed. This series of operations is automatically performed by the sample evaluation apparatus 10.

FIG. 9 is an enlarged view of an observation image of the standard sample S obtained in this way. In the figure, in the observation image of the lattice-like correction pattern 101 (see FIG. 2) of the standard sample S, lattice lines extending in the X direction are denoted by reference numerals x ₁ to x ₂₁ and extending in the Y direction. Are indicated by y ₁ to y ₂₁ .

Thus the observation image of the correction pattern 101 thus obtained is stored in the storage unit 20 by the image analysis unit 19 as image data D ₂ is provided. The image data D ₂ is obtained by associating the position (X, Y) of each pixel with the luminance data I _(X , _{Y) at the} position (X, Y).

  Next, the process proceeds to step S3 in FIG.

  This step has sub-steps P1 to P3, and the image analysis unit 19 automatically performs each of these sub-steps P1 to P3.

In a first sub-step P1, to Hough transform an observation image of the grating lines _{x 1 ~x 21, y 1 ~y} 21 shown in FIG.

  FIG. 10 is a diagram for explaining the Hough transform.

  The straight line L in the real space in which the XY orthogonal coordinate system is defined is uniquely determined by the distance r from the coordinate origin O and the angle θ formed by the unit normal vector n with the X axis. In this way, any straight line in real space has a one-to-one correspondence with a point in (r, θ) space excluding the origin, and a map from a straight line in real space to a point in (r, θ) space This is called Hough transformation. The (r, θ) space is called a Hough space.

  Here, the inner product of the vector P from the origin O in the real space to an arbitrary point (x, y) on the straight line L and the normal vector n described above (the length of the normal vector n is 1). To) is equal to the distance r between the origin O and the straight line L. Therefore, the following equation (1) is established.

式（１）において点(x、y)を固定して、式（１）を満たす点(r、θ)の軌跡をHough空間内に描くと、一つの正弦曲線が得られる。その正弦曲線は、実空間の点(x、y)を通る直線群に対応する。

When the point (x, y) is fixed in the equation (1) and the locus of the point (r, θ) satisfying the equation (1) is drawn in the Hough space, one sine curve is obtained. The sine curve corresponds to a group of straight lines passing through a point (x, y) in real space.

  From this, equation (1) defines a correspondence from a point (x, y) in real space to a sinusoid in the Hough space corresponding to a group of straight lines passing through the point (x, y). You can also see it.

In the present embodiment, the real space is divided into a plurality of cells, and attention is paid to one of the cells. Then, when there are grid lines x _{1 to} x ₂₁ and y _{1 to} y ₂₁ passing through the cell, a sine curve corresponding to a group of straight lines passing through the cell is calculated by the equation (1). Further, the Hough space is also divided into a plurality of cells, and “1” is voted for the cell through which the above sine curve passes. When such a procedure is performed for all cells in the real space, the number of votes for all cells in the Hough space is determined.

  FIG. 11 is a diagram showing the distribution of the number of votes in the Hough space. In the figure, the number of votes is represented by shading, and the white portion indicates that the number of votes is large. Hereinafter, the distribution of the number of votes is referred to as a Hough transform image.

  12 to 15 show other examples of the Hough transform image of the correction pattern.

  12 and 13 are diagrams when the image rotation mechanism is used.

  Among these, FIG. 12A is an observation image when there is no distortion, and FIG. 12B is a Hough transform image of the observation image.

  On the other hand, FIG. 13A is an observation image when there is distortion, and FIG. 13B is a Hough transform image of the observation image.

  14 and 15 are diagrams in a low magnification region.

  Among these, FIG. 14A is an observation image when there is no distortion, and FIG. 14B is a Hough transform image of the observation image.

  On the other hand, FIG. 15A is an observation image when there is distortion, and FIG. 15B is a Hough transform image of the observation image.

In FIG. 11 to FIG. 15, a threshold is set in advance for the number of votes of the Hough transform image, and when there are cells with the number of votes equal to or greater than the threshold, the position coordinates (r ₁ , θ ₁ ) to ( r _n, is recognized as grid lines x ₁ ~x ₂₁ corresponding to θ _n), y ₁ ~y ₂₁ is present in the observation image.

However, the distortion of the grid lines _{x 1 ~x 21, y 1 ~y} 21 in the observation image is intense, there is no cell number of votes is equal to or greater than the threshold value, the grid lines _{x 1 ~x 21, y 1 ~y} 21 May not be recognized.

Therefore, in the next step P2 shown in FIG. 8, it is determined whether to recognize grid lines _{x 1 ~x 21, y 1 ~y} 21.

  If it is determined in step P2 that the grid lines can be recognized (YES), the process proceeds to step P3.

In step P3, based on the Hough transform image of the to calculate the position coordinates of the grid point which is the intersection of each grid line _{x 1 ~x 21, y 1 ~y} 21 as follows.

First, it is assumed that the coordinates of cells whose number of votes is equal to or greater than the threshold are (r ₁ , θ ₁ ) to (r _n , θ _n ) as described above. The straight lines in the real space corresponding to the two coordinates (r _i , θ _i ) and (r _j , θ _j ) are the following equations (2) and (3) from the above equation (1). Can be written as

そして、これら式（２）、（３）を未知数（x、y）についての連立方式と捉え、（x、y）について解くと次の式（４）が得られる。

Then, when these equations (2) and (3) are regarded as simultaneous methods for the unknown (x, y) and solved for (x, y), the following equation (4) is obtained.

式（４）で得られた解（x、y）は、座標（r_i、θ_i）、（r_j、θ_j）に対応する実空間の二つの格子線の交点、すなわち格子点の座標である。

The solution (x, y) obtained by Equation (4) is the intersection of two grid lines in the real space corresponding to the coordinates (r _i , θ _i ), (r _j , θ _j ), that is, the coordinates of the grid points It is.

Such calculations _{(r 1, θ 1) ~} (r n, θ n) By performing all of the two sets of all of the grid lines _{x 1 ~x 21, y 1 ~y} 21 in the observation image two sets x _s, the position coordinates of the grid point P _st of _{y t (x st, y st} ) is calculated.

Thus, by pattern recognition using the Hough transform, the positions of all lattice points P _st in the observation image are specified as actual positions reflecting the distortion of the observation image caused by the electromagnetic lens system 13. Become.

  As described above, the Hough transform recognizes the grid line in the real space by the number of votes for the cell in the Hough space, and the grid line can be obtained even if the center of the line width of the grid line in the real space is not known. Can be recognized. Therefore, even if the line width of the grid line is slightly narrower than the spatial resolution of the sample observation apparatus 10 and the observation image of the grid line is somewhat blurred, the grid line can be recognized using the Hough transform.

  Therefore, in the four types of standard samples S produced individually corresponding to the magnification ranges in Tables 1 to 4, the line width of the correction pattern 101 does not have to be set as the upper limit of the spatial resolution of each magnification. The line width of the correction pattern 101 can be set to an intermediate value of the spatial resolution in each table.

  For example, in each range of 200 to 2,000 times, 4,000 to 20,000 times, and 40,000 to 200,000 times, the line width of the correction pattern 101 is 3.5 μm, 250 nm, and 20 nm, respectively. do it.

  In the high magnification region of 400,000 to 10,000,000 times, a thin film made of a crystal material used as the standard sample S may be made of a crystal material having a lattice constant of about 1 nm.

Thereafter, as shown in FIG. 16, the distortion caused by the electromagnetic lens system 13 are two grid lines x _s when it is not observed image, the grid points of y _t the ideal grid point Q _st. Then, such an ideal grid point Q _st a starting point, and calculates the positional deviation vector [Delta] r _st to the end point of the actual lattice point P _st described above.

Thereby, it is possible to grasp how much each lattice point _Pst is actually displaced due to the distortion of the observation image caused by the electromagnetic lens system 13.

Note that the calculation of the misregistration vector Δr _st may be performed for all grid points. However, when the calculation is difficult, a plurality of representatives are selected from all the grid points, and the misregistration is performed only for those representatives. The vector Δr _st may be calculated.

Further, FIG. 16 illustrates the case where the observation image is rotated by the image rotation mechanism. However, when the observation image is observed in the low magnification region as shown in FIG. 17, the positional deviation vector Δr _st is calculated in the same manner as described above. Is done.

By the way, since each lattice point _Pst is discretely distributed in the observation image, the positional deviation of an arbitrary point of the observation image cannot be known as it is.

Therefore, by using a generally known two-dimensional linear interpolation method, a positional deviation vector Δr (X, Y) at an arbitrary point (X, Y) of the observation image based on the positional deviation vector Δr _st. Is calculated.

Then, the positional deviation vector [Delta] r (X, Y) component Show (ΔX (X, Y), ΔY (X, Y)) and represents a storage unit for image analysis unit 19 is provided with the same as the position error data D ₃ 20.

Its displacement data D _3, in association with the observation condition data D ₁ obtained in substep T5, preferably stored in the storage unit 20 as the positional deviation database DB.

  FIG. 22 is a diagram schematically showing the structure of the database DB.

The database DB includes observation conditions A, B, C, which include the acceleration voltage of the electron beam EB, the deflection amount of the electron beam EB by the electromagnetic lens system 13, the magnification of the observation image of the correction pattern, and the rotation angle of the observation image by the rotation mechanism. and ..., made in correspondence with the displacement data D ₃ obtained in each of these viewing conditions.

  Thus, step S3 is completed.

  Next, the process proceeds to step S4.

In step S4, with reference to the displacement data D _3, the luminance data I _{(X, Y)} of the image data D ₂ obtained in Step S2, and for all the points of the observation image _(X, Y) I ( _X- ΔX _{(X, Y} ), _Y- ΔY _{(X, Y))} are corrected. This correction is automatically performed by the image analysis unit 19.

  By this correction, the luminance of the point (X, Y) in the observed image is actually recognized as the luminance at the point (X−ΔX (X, Y), Y−ΔY (X, Y)). Thus, an observation image of the standard sample S free from distortion due to the electromagnetic lens system 13 can be obtained.

  Next, the process proceeds to step S9.

In this step, the Hough transform is performed again on the undistorted observation image obtained in step S4, and the position coordinates (x _st , y _st ) of each lattice point P _st are calculated again as in step P3. It is determined whether or not the positional deviation between the lattice point Q _st and the lattice point P _st is eliminated. This determination may be made by the operator or the image analysis unit 19.

  When it is determined that the positional deviation has been eliminated (YES), the correction of the sample observation apparatus according to the present embodiment is finished.

On the other hand, if it is determined that the positional deviation has not been eliminated (NO), steps S2 to S4 are performed again. In this case, for example, the positional deviation is eliminated by changing the two-dimensional linear interpolation method for the positional deviation vector Δr _st performed in step S4.

If it is determined in step P2 that the lattice lines x _{1 to} x ₂₁ and y _{1 to} y ₂₁ cannot be recognized (NO), the process proceeds to step S5.

  In step S5, the sub-steps T3 to T6 described in step S1 are manually performed by an operator.

  However, in the first step T3, instead of the standard sample S, a rough correction sample 40 as shown in the plan view of FIG. 18A is mounted on the sample mounting portion 16 (see FIG. 1).

  The rough correction sample 40 is formed by arranging a plurality of island patterns 40a in a matrix at an equal pitch.

  FIG. 18B is a cross-sectional view taken along line II-II in FIG.

  The island pattern 40a is made of, for example, silicon, and a silicon oxide film 40b is embedded between the patterns 40a. In such a rough correction sample 40, for example, a groove is formed on the surface side of a silicon substrate, a silicon oxide film 40b is embedded in the groove, and polishing is performed from the back side of the silicon substrate, and the silicon oxide film 40b is formed on the polished surface. Can be formed.

  After performing the sub-steps T3 to T6 on the rough correction sample 40, the process proceeds to step S6.

  In step S6, the electron beam EB is scanned on the rough correction sample 40, and the electron beam EB transmitted through the rough correction sample 40 is captured by the STEM detector 17 in synchronization with the scanning speed. Each observation image is obtained. This series of operations is automatically performed by the sample evaluation apparatus 10.

  FIG. 19 is an enlarged view of the observation image 50 of the island pattern 40a obtained in this way.

  As shown in the figure, each observation image 50 is observed to be shifted from the ideal position image 51 without distortion due to distortion of the observation image caused by the electromagnetic lens system 13.

Also, the observation image 50 is stored in the storage unit 20 by the image analysis unit 19 as image data D ₄ is provided. The image data D ₄ is the position of each pixel (X, Y), formed by its position (X, Y) luminance data I _{(X, Y)} in the made to correspond to.

  Next, the process proceeds to step S7. In this step, the following operation is automatically performed by the image analysis unit 19.

  First, how much each observation image 50 is located from the ideal position image 51 is calculated.

  FIG. 20 is a plan view schematically showing the calculation method.

In calculation, the center Q _nm of n row m-th column of the ideal position image 51 and the start point, and calculates the positional deviation vector [Delta] r _nm to ending the center P _nm of n row m-th column of the observation image 50.

Then, using a generally known two-dimensional linear interpolation method, a positional deviation vector Δr (X, Y) at an arbitrary point (X, Y) of the observation image is calculated based on the positional deviation vector Δr _nm. calculate.

Thereafter, the component display of the positional deviation vector Δr (X, Y) is represented as (ΔX (X, Y), ΔY (X, Y)), and this is stored as positional deviation data D _{5 in the} storage unit 20 included in the image analysis unit 19. To store.

Further, the luminance data I _(X , _Y) of the image data D ₄ acquired in step S6 is obtained by using I _(X- ΔX _{(X, Y} ), _Y- Correct with Δ _{Y (X, Y))} .

  As a result, the luminance of the point (X, Y) in the observed image is actually recognized as the luminance at the point (X−ΔX (X, Y), Y−ΔY (X, Y)). An observation image of the island pattern 40a free from distortion caused by the electromagnetic lens system 13 can be obtained.

  Thus, step S7 is completed.

  Next, the process proceeds to step S8, where it is determined whether or not each island pattern 40a has been displaced by the above correction. This determination may be made by the operator, or may be automatically made by the image analysis unit 19.

In the determination, a two-dimensional Fourier transform image 52 of the observation image 50 is obtained as shown in FIG. A point in the Fourier space is specified by spatial frequencies k _x and k _y corresponding to the X axis and the Y axis in real space.

The size of the Fourier transform image 52 is determined by the arrangement pitch (interval) of the observation image 50. For example, when the observation image 50 is not distorted, it is assumed that the arrangement pitch of the observation image 50 is d ₀ and the diameter of the circumscribed circle C ₀ of the Fourier transform image 52 is k ₀ . On the other hand, when the observed image 50 is distorted and the pitch d becomes larger than d ₀ , the diameter k of the circumscribed circle C of the Fourier transform image 52 becomes smaller than k ₀ .

Accordingly, the diameter k of the circumscribed circle C is determined that there is a positional deviation of the observation image 50 is smaller than k _0, it can be determined that the position deviation when the diameter k is equal to k ₀ are eliminated.

  As a result of such determination, when it is determined that the positional deviation has been eliminated (YES), the coarse correction is finished, the process returns to the sub-step T3 of step S1, and a precise measurement using the standard sample S and the Hough transform is performed. Make corrections.

On the other hand, if it is determined that the positional deviation has not been eliminated (NO), the process returns to steps S6 and S7, and the positional deviation is eliminated by changing the two-dimensional linear interpolation method for the positional deviation vector Δr _nm . So that

  Thus, the main steps of the sample observation apparatus correction method according to the present embodiment are completed.

  Thereafter, the correction sample S is removed from the sample mounting portion 16 (see FIG. 1), and an observation sample such as a semiconductor chip is placed on the sample mounting portion 16.

Then, the corresponding to the viewing conditions of the observation specimen position displacement database DB with reference to the displacement data D ₃ in (see FIG. 22), so that the distortion disappears in the observation image of the sample for observation, the output from the detector 17 The image data of the observation sample to be corrected is corrected.

The correction method is the same as in step S4. That is, the luminance data I _{(X, Y)} in the image data of the observation sample produced by the image analysis unit 19 is expressed as I _(X− ΔX _{(X, Y} )) for all points _{(X, Y} ). , _Y- ΔY _{(X, Y))} .

According to the embodiment described above, so that the position deviation of the lattice point P _st of the correction pattern 101 calculated disappears in step (S3), so corrects the image data D ₂ in step (S4), the image rotation It is possible to prevent distortion of an observation image that occurs when using a mechanism or when observing at a low magnification.

In addition, in the Hough transform performed for recognizing the lattice point _Pst , even if the lattice line deviates from the straight line due to image distortion and becomes slightly curved, the straight line in the real space is converted to a point in the Hough space. Therefore, the coordinates of the lattice points can be easily obtained by solving the simultaneous method of Equation (4).

  Further, by storing the result of the positional deviation as a positional deviation database DB (see FIG. 22), when observing an observation sample such as a semiconductor chip, the distortion of the observation image can be reduced with reference to the database DB. It can be corrected. Thus, for example, by inputting CAD data including the design layout of the chip to the sample observation apparatus 10, the sample observation apparatus 10 side automatically identifies a defective portion in the chip while accurately overlaying the design layout and the observation image. Can be recognized.

  The features of the present invention are added below.

(Appendix 1) (S1) A step of applying a charged particle beam transmitted through an electromagnetic lens system to a standard sample on which a correction pattern is formed;
(S2) obtaining an observation image of the correction pattern by detecting the charged particle beam applied to the standard sample with a detector;
(S3) calculating how much the plurality of reference points of the correction pattern are displaced in the observation image due to distortion of the observation image caused by the electromagnetic lens system;
(S4) correcting the image data of the observation image so as to eliminate the displacement, and
A correction method for a sample observation apparatus, comprising:

  (Supplementary Note 2) In the step (S3), the position of the reference point in the observation image is specified as an actual position by pattern recognition, and a deviation between the ideal position of the reference point and the actual position when there is no distortion is determined. The correction method of the sample observation apparatus according to attachment 1, wherein the correction is performed by calculation.

  (Supplementary Note 3) The step (S3) is performed by calculating, as a displacement vector, a vector having the ideal position as a start point and the actual position as an end point for each of the plurality of reference points. 3. A method for correcting the sample observation apparatus according to 2.

(Additional remark 4) While using the grid | lattice-like pattern which consists of a some grid line as said correction pattern, and using the intersection of the said grid line as said reference | standard point,
The step (S3)
(P1) mapping each observation image of the plurality of grid lines to a point in the Hough space by Hough transformation;
(P2) The method for correcting a sample observation apparatus according to appendix 2, further comprising: calculating the position of the intersection of the plurality of grid lines as the actual position from the position coordinates of the point in the Hough space. .

(Supplementary Note 5) In the step (P1), when a Hough transform image of the lattice-shaped correction pattern cannot be obtained,
(S5) applying the charged particle beam transmitted through the electromagnetic lens system to a rough correction sample in which a plurality of island-shaped patterns are arranged in a matrix; and
(S6) obtaining an observation image of each of the plurality of island patterns by detecting the charged particle beam applied to the rough correction sample with the detector;
(S7) Obtain the positional deviation of the observation image of each of the plurality of island-shaped patterns generated due to the electromagnetic lens system, and image data of the observation image of the island-shaped pattern so as to eliminate the positional deviation The correction method of the sample observation apparatus according to appendix 4, further comprising a step of correcting.

  (Supplementary Note 6) (S8) The method further includes the step of determining whether or not the observation image of each of the plurality of island patterns has been misplaced by performing the step (S7). 5. A method for correcting a sample observation apparatus according to 5.

  (Additional remark 7) The said step (S8) calculates | requires the two-dimensional Fourier-transform image of the observation image of the said several island-shaped pattern after the correction | amendment, and the diameter of the circumcircle of this two-dimensional Fourier-transform image has no position shift. The correction method of the sample observation apparatus according to appendix 6, wherein the correction is performed by determining that there is a displacement of the island pattern when the diameter is smaller than the case diameter.

(Supplementary Note 8) In the step (S3), the positional deviation is determined based on the acceleration voltage of the charged particle beam, the deflection amount of the charged particle beam by the electromagnetic lens system, the magnification of the observation image of the correction pattern, and the observation. Store it in a database in association with viewing conditions that include at least one of the image rotation angles,
When observing an observation sample under a specific observation condition, refer to the positional deviation in the database corresponding to the observation condition, so that the observation image of the observation sample is free from distortion. The correction method of the sample observation apparatus according to appendix 1, wherein the image data is corrected.

(Appendix 9) a charged particle gun that generates and accelerates a charged particle beam;
An electromagnetic lens system for focusing the charged particle beam on a standard sample on which a correction pattern is formed;
A detector that converts the intensity of the charged particle beam applied to the standard sample into a detection signal;
An image analysis unit that creates the image data from the detection signal;
The image analysis unit calculates how much the plurality of reference points of the correction pattern are displaced in the observation image due to distortion of the observation image of the correction pattern caused by the electromagnetic lens system. And correcting the image data so that the positional deviation is eliminated.

(Supplementary Note 10) The calculation of the displacement in the image analysis unit is as follows.
The position of the reference point in the observation image is specified as an actual position by pattern recognition, and a deviation between the ideal position of the reference point and the actual position when there is no distortion is calculated. 9. The sample observation apparatus according to 9.

(Supplementary Note 11) The calculation of the positional deviation in the image analysis unit is as follows.
Each observation image of the plurality of grid lines of the correction pattern made of a grid pattern is mapped to a point in the Hough space by Hough transformation,
The sample observation apparatus according to appendix 10, wherein the observation is performed by calculating the position of the intersection of the plurality of grid lines as the actual position from the position coordinates of the point in the Hough space.

(Supplementary Note 12) The positional deviation of the reference point is determined by the acceleration voltage of the charged particle beam, the deflection amount of the charged particle beam by the electromagnetic lens system, the magnification of the observation image of the correction pattern, and the rotation of the observation image. A storage unit storing a database associated with an observation condition including at least one of the angles;
When observing an observation sample under a specific observation condition, the image analysis unit refers to the positional deviation in the database corresponding to the observation condition so that the observation image of the observation sample is free from distortion. The sample observation apparatus according to appendix 9, wherein image data of the observation sample is corrected.

FIG. 1 is a configuration diagram of a sample evaluation apparatus according to an embodiment of the present invention. FIG. 2A is an enlarged plan view of a standard sample, and FIG. 2B is a cross-sectional view taken along line I-I in FIG. FIG. 3 is a principle diagram of a sample observation method in the sample evaluation apparatus. FIG. 4A is a diagram showing a scanning direction when the image rotation mechanism is not used, and FIG. 4B is an observation of a standard sample obtained by scanning an electron beam as shown in FIG. It is a statue. FIG. 5A is a diagram showing a scanning direction rotated by an angle θ by an image rotation mechanism, and FIG. 5B is a diagram of a standard sample obtained by scanning an electron beam as shown in FIG. It is an observation image. FIG. 6A is a diagram showing the scanning direction when observing at a high magnification, and FIG. 6B is an observation image of a standard sample obtained by scanning an electron beam as shown in FIG. It is. FIG. 7A is a diagram showing a scanning direction when observing at a low magnification, and FIG. 7B is an observation image of a standard sample obtained by scanning an electron beam as shown in FIG. It is. FIG. 8 is a flowchart for explaining a correction method of the sample observation apparatus according to the embodiment of the present invention. FIG. 9 is an enlarged view of the observation image of the standard sample obtained in step S2 of the correction method of the sample observation apparatus according to the embodiment of the present invention. FIG. 10 is a diagram for explaining the Hough transform. FIG. 11 is a diagram showing the distribution of the number of votes in the Hough space. FIG. 12A is an observation image of a standard sample when there is no distortion when an image rotation mechanism is used, and FIG. 12B is a Hough transform image of the observation image. FIG. 13A is an observation image of a standard sample when there is distortion when an image rotation mechanism is used, and FIG. 13B is a Hough transform image of the observation image. FIG. 14A is an observation image of the standard sample when there is no distortion in the low magnification region, and FIG. 14B is a Hough transform image of the observation image. FIG. 15A is an observation image of a standard sample when there is distortion in a low magnification region, and FIG. 15B is a Hough transform image of the observation image. FIG. 16 is a diagram illustrating a calculation method of the positional deviation vector Δr _st when the image rotation mechanism is used. FIG. 17 is a diagram illustrating a method of calculating the positional deviation vector Δr _st when observing in a low magnification region. FIG. 18A is a plan view of the rough correction sample 40 used in the embodiment of the present invention, and FIG. 18B is a cross-sectional view taken along the line II-II in FIG. . FIG. 19 is an enlarged view of an observation image of the island pattern of the rough correction sample. FIG. 20 is a plan view schematically showing a method for calculating the positional deviation amount of the observation image of the island pattern. FIG. 21 is a plan view showing a two-dimensional Fourier transform image of the observation image of the island pattern. FIG. 22 is a diagram schematically showing the structure of a database that stores misregistration data.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Sample evaluation apparatus, 11 ... Housing | casing, 12 ... Electron gun, 13 ... Electromagnetic lens system, 13a, 13b ... 1st, 2nd convergence lens, 13c ... Objective lens, 14 ... Scanning / deflection lens, 15 ... Aperture, DESCRIPTION OF SYMBOLS 16 ... Sample mounting part, 17 ... STEM detector, 19 ... Image analysis part, 20 ... Memory | storage part, 30 ... Observation image, 40 ... Sample for rough correction, 40a ... Island-like pattern, 40b ... Silicon oxide film, 50 ... Observation Image: 51 ... Ideal position image, 52 ... Two-dimensional Fourier transform image of observed image, 100 ... Silicon substrate, 100a ... Groove, 101 ... Correction pattern, S ... Correction sample.

Claims (4)

(S1) applying a charged particle beam transmitted through the electromagnetic lens system to a standard sample on which a correction pattern is formed;
(S2) obtaining an observation image of the correction pattern by detecting the charged particle beam applied to the standard sample with a detector;
(S3) calculating how much the plurality of reference points of the correction pattern are displaced in the observation image due to distortion of the observation image caused by the electromagnetic lens system;
(S4) correcting the image data of the observation image so as to eliminate the displacement, and
I have a,
In the step (S3), for each of the plurality of reference points, the position of the reference point in the observation image is specified as an actual position by pattern recognition, and the ideal position of the reference point when there is no distortion is used as a starting point. A correction method for a sample observation apparatus, which is performed by calculating a displacement vector having the actual position as an end point . Using a lattice-shaped pattern composed of a plurality of lattice lines as the correction pattern, and using an intersection of the lattice lines as the reference point,
The step (S3)
(P1) mapping each observation image of the plurality of grid lines to a point in the Hough space by Hough transformation;
(P2) The correction of the sample observation apparatus according to claim 1, further comprising a step of calculating a position of an intersection of the plurality of grid lines as the actual position from a position coordinate of the point in the Hough space. Method. In the step (S3), the positional deviation vector is calculated using the acceleration voltage of the charged particle beam, the deflection amount of the charged particle beam by the electromagnetic lens system, the magnification of the observation image of the correction pattern , and the observation image. Store in the database in association with the observation conditions including at least one of the rotation angles,
When observing an observation sample under a specific observation condition, refer to the positional displacement vector in the database corresponding to the observation condition so that the observation image of the observation sample is free from distortion. The correction method of the sample observation apparatus according to claim 1, wherein the image data is corrected. A charged particle gun that generates and accelerates a charged particle beam;
An electromagnetic lens system for focusing the charged particle beam on a standard sample on which a correction pattern is formed;
A detector that converts the intensity of the charged particle beam applied to the standard sample into a detection signal;
An image analysis unit that creates image data from the detection signal;
The image analysis unit determines how much the plurality of reference points of the correction pattern are displaced in the observation image due to distortion of the observation image of the correction pattern caused by the electromagnetic lens system. For each of a plurality of reference points, the position of the reference point in the observation image is specified as an actual position by pattern recognition, and the ideal position of the reference point when there is no distortion is the start point and the actual position is the end point A sample observation apparatus, wherein the sample data is calculated as a deviation vector and the image data is corrected so that the positional deviation is eliminated.

Images