JP4696097B2 - Sample image forming method and charged particle beam apparatus - Google Patents

Description

  The present invention relates to a method for forming a sample image and a charged particle beam apparatus, and more particularly, to form a high-resolution image that is not affected by image drift at a high magnification, or a sample image suitable for obtaining an accurate dimension value of a measurement object. The present invention relates to a method and a charged particle beam apparatus.

  In a charged particle beam apparatus typified by a scanning electron microscope, desired information (for example, a sample image) is obtained from a sample by scanning a finely focused charged particle beam on the sample. In such a charged particle beam apparatus, higher resolution is progressing year by year, and the observation magnification required with higher resolution is higher. In addition, the beam scanning method for obtaining the sample image includes a method of accumulating a plurality of images obtained by high-speed scanning to obtain a final target image, and one low-speed scanning (one frame image capturing time: 40). There is a method of acquiring a target image in about 80 seconds from a second). As the observation magnification increases, the influence of visual field drift on the acquired image becomes more serious. For example, in a method of acquiring a target image by integrating (frame integrating) image signals obtained by high-speed scanning for each pixel, if there is a drift based on charge-up of a sample during image integration, the field of view is shifted Since the pixels are integrated, the target image after the integration is blurred in the drift direction. In order to reduce the influence of drift, the number of integration frames may be reduced to shorten the integration time. However, in this case, a sufficient S / N ratio cannot be obtained.

  In the method of acquiring an image by low-speed scanning, if there is a drift during image capture, the field of view flows in the drift direction, so that the image is deformed.

  In Japanese Patent Laid-Open No. 62-43050 (Patent Document 1), a pattern for drift detection is stored, and an image of this pattern is periodically acquired to detect a deviation from the stored pattern to perform beam irradiation. A technique for correcting the position is disclosed.

  Japanese Patent Application Laid-Open No. 5-290787 (Patent Document 2) obtains two images based on electron beam scanning on a specific observation area, and pattern matching for specifying the amount of displacement and the direction of the displacement. And a technique for integrating the images by moving the pixels by the specified deviation.

JP 62-43050 A JP-A-5-290787

  In the technique disclosed in Japanese Patent Laid-Open No. 62-43050, when the observation magnification is several hundred thousand times, the control accuracy of the beam irradiation position becomes insufficient. For example, when an image of 1280 × 960 pixels is acquired at an observation magnification of 200,000 times, the size of one pixel on the observation visual field (on the sample) is about 0.5 nm. As miniaturization of measurement objects progresses, measurement and evaluation at higher magnification is required, but in such a situation, when applied to an apparatus that forms a final image by accumulating multiple images, An image shift (drift) of several nanometers or less works as “blurring” of the image during frame integration.

  The technique disclosed in Japanese Patent Application Laid-Open No. 62-43050 corrects drift by controlling the scanning position of an electron beam (for example, image shift deflection) and suppresses image shift. The practical limit is several nanometers to several tens of nanometers, and it is almost impossible to perform position correction (drift correction) on an image with a magnification of several hundred thousand times or more at the pixel level. Moreover, since it takes time until the drift is stabilized, there is a problem that the throughput is lowered.

  On the other hand, the technique disclosed in Japanese Patent Application Laid-Open No. 5-290787 can make a certain evaluation regarding the position correction at the pixel level between images, but has the following problems.

  In general, image data that has not undergone image integration has a poor S / N ratio, so that it is difficult to detect a shift between images, and it is difficult to correct with high accuracy. In addition, increasing the probe current (electron beam current) can increase secondary electron emission and improve the S / N ratio, but in the case of easily charged samples, the field of view of the electron beam is moved by charging. As a result, the difference between the images acquired at different timings becomes more conspicuous, and as a result, it is difficult to correct with high accuracy. In addition, when a sample that is vulnerable to electron beam damage is irradiated with a large beam current, there are problems such as sample destruction and evaporation.

  An object of the present invention is to provide a sample image forming method and a charged particle beam apparatus suitable for realizing the suppression of visual field shift with high accuracy while suppressing the influence of charging by charged particle beam irradiation.

  In order to achieve the above object, in the present invention, in a sample image forming method in which a charged particle beam is scanned on a sample and an image is formed based on a secondary signal emitted from the sample, a plurality of images obtained by a plurality of scans are obtained. A sample image forming method characterized in that a plurality of composite images are formed by synthesizing the images, and the images are combined by correcting misalignment between the composite images, and a further composite image is formed. A charged particle beam apparatus for realizing the method is provided.

  As described above, by forming a composite image and correcting the misalignment on the composite image, it is possible to detect misalignment between images having a sufficient S / N ratio without increasing the beam current. It is possible to suppress “blurring” of an image at the time of frame integration by correcting a high positional deviation. Other objects and other specific configurations of the present invention will be described in the section of the best mode for carrying out the invention.

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

  FIG. 1 is a schematic configuration diagram of a scanning electron microscope which is an example of the present invention. A voltage is applied between the cathode 1 and the first anode 2 by a high voltage control power source 20 controlled by a computer 40, and the primary electron beam 4 is extracted from the cathode 1 with a predetermined emission current. An acceleration voltage is applied between the cathode 1 and the second anode 3 by a high-voltage control power source 20 controlled by a computer 40, and the primary electron beam 4 emitted from the cathode 1 is accelerated and proceeds to the subsequent lens system. .

  The primary electron beam 4 is converged by the converging lens 5 controlled by the lens control power source 21, and after the unnecessary region of the primary electron beam is removed by the diaphragm plate 8, the converging lens 6 controlled by the lens control power source 22, And the objective lens 7 controlled by the objective lens control power source 23 is converged as a minute spot on the sample 10. The objective lens 7 can take various forms such as an in-lens system, an out-lens system, and a snorkel system (semi-in-lens system). A retarding method is also possible in which a negative voltage is applied to the sample to decelerate the primary electron beam. Furthermore, each lens may be composed of an electrostatic lens composed of a plurality of electrodes.

  The primary electron beam 4 is scanned two-dimensionally (XY direction) on the sample 10 by the scanning coil 9. The scanning coil 9 has a secondary signal 12 such as a secondary electron generated from the sample 10 by irradiation of a primary electron beam supplied with a current by a scanning coil control power supply 24, travels to the upper part of the objective lens 7, and then secondary The signal is separated from the primary electrons by the orthogonal electromagnetic field generator 11 for signal separation and detected by the secondary signal detector 13. The signal detected by the secondary signal detector 13 is amplified by the signal amplifier 14 and then transferred to the image memory 25 and displayed on the image display device 26 as a sample image. The secondary signal detector may detect secondary electrons or reflected electrons, or may detect light or X-rays.

  An address signal corresponding to the memory position of the image memory 25 is generated in the computer 40, converted into an analog signal, and then supplied to the scanning coil 9 via the scanning coil control power source 24. For example, when the image memory 25 has 512 × 512 pixels, the X direction address signal is a digital signal that repeats 0 to 512, and the Y direction address signal is when the X direction address signal reaches 0 to 512. It is a digital signal of 0 to 512 that is incremented by 1. This is converted into an analog signal.

  Since the address of the image memory 25 corresponds to the address of the deflection signal for scanning the primary electron beam, a two-dimensional image of the deflection area of the primary electron beam by the scanning coil 9 is recorded in the image memory 25. The signals in the image memory 25 can be sequentially read out in time series by a read address generation circuit (not shown) synchronized with a read clock. The signal read corresponding to the address is converted into an analog signal and becomes a luminance modulation signal of the image display device 26.

  The image memory 25 has a function of storing (combining) images (image data) in an overlapping manner for improving the S / N ratio. For example, a single completed image is formed by storing images obtained by eight two-dimensional scans in an overlapping manner. That is, a final image is formed by combining images formed in one or more XY scanning units. The number of images (frame integration number) for forming one completed image can be arbitrarily set, and an appropriate value is set in consideration of conditions such as secondary electron generation efficiency. Further, an image desired to be finally acquired can be formed by further overlapping a plurality of images formed by integrating a plurality of sheets. When the desired number of images is stored, or after that, blanking of the primary electron beam may be executed to interrupt information input to the image memory.

  In addition, when the number of integrated frames is set to 8, when the 9th image is input, a sequence may be provided in which the 1st image is deleted and 8 images remain as a result. In addition, when the ninth image is input, it is also possible to perform weighted averaging such that the accumulated image stored in the image memory is multiplied by 7/8 and the ninth image is added thereto. .

  A two-stage deflection coil 51 (image shift deflector) is disposed at the same position as the scanning coil 9, and the position (observation field of view) of the primary electron beam 4 on the sample 10 can be controlled two-dimensionally. The deflection coil 51 is controlled by a deflection coil control power supply 31.

  The stage 15 can move the sample 10 in at least two directions (X direction and Y direction) in a plane perpendicular to the primary electron beam.

  From the input device 42, it is possible to specify an image capturing condition (scanning speed, total number of images), a visual field correction method, and the like, and output and storage of an image.

  The apparatus according to the present invention has a function of forming a line profile based on detected secondary electrons or reflected electrons. The line profile is formed based on the amount of detected electrons when the primary electron beam is scanned one-dimensionally or two-dimensionally or the luminance information of the sample image. The obtained line profile is, for example, on a semiconductor wafer. It is used for measuring the dimension of the formed pattern. The apparatus according to the present embodiment can further include an interface 41 for transferring image data to an external device or the like, and a recording device 27 for storing the image data in a predetermined storage medium.

  In the description of FIG. 1, the control device is described as being integrated with or equivalent to the scanning electron microscope. However, the control device is of course not limited thereto, and is described below with a control processor provided separately from the scanning electron microscope body. Such processing may be performed. In this case, a detection signal detected by the secondary signal detector 13 is transmitted to the control processor, or a signal is transmitted from the control processor to a lens or a deflector of the scanning electron microscope, and via the transmission medium. An input / output terminal for inputting / outputting a transmitted signal is required.

  Alternatively, a program for performing the processing described below may be registered in a storage medium, and the program may be executed by a control processor that has an image memory and supplies necessary signals to the scanning electron microscope. That is, the embodiment of the present invention described below is also established as an invention of a program that can be used in a charged particle beam apparatus such as a scanning electron microscope equipped with an image processor.

Example 1
In the embodiment of the method for improving the S / N ratio by integrating TV scanning images, the processing flow of FIG. 2 will be described in detail below. FIG. 5 is a diagram schematically showing the processing of FIG.

First step (S2001):
From the input device 42, the frame integration number N0 of each captured image and the number N1 of captured images are designated. At this time, the total number of frames integrated in the final image is N0 × N1. Usually, when N0 is about 2 to 8, and N1 is about 10 to 50, a necessary S / N can be secured according to the purpose. Note that when each image is acquired by a slow scan that is slightly slower than the TV scan, it is possible to set N0 to one. In the interlaced TV scan, one frame is obtained by two scans. In such a case, N0 can be set to 2. In this condition setting, it is desirable to form the plurality of sample images in a state where optical conditions (electron beam convergence condition and scanning condition) are fixed in order to facilitate later detection of misalignment.

Second step (S2002):
When the start of image capture is instructed from the input device 42, N1 images (F1, F2,..., FN1) of the frame integration number N0 are continuously acquired in the same visual field.

Third step (S2003):
F1 is set in the memory area of the target image F0.

Fourth step (S2004):
A sharpened image F0a is created from the target image F0. As the sharpening process, a technique such as using an image filter that emphasizes an edge in an image can be used.

Fifth step (S2005):
A sharpened image F2a is created from F2.

Sixth step (S2006):
A positional shift between the sharpened images F2a and F0a of F2 is detected. For detection of misregistration, arithmetic processing such as image correlation can be applied. However, the present invention is not limited to this, and any image arithmetic method capable of detecting misregistration is applicable.

Seventh step (S2007):
The pixels of the original image F2 are shifted by the visual field shift amount detected in the sixth step, added to the F0 image, and returned as the target image F0 again.

Eighth step (S2008):
F2 is replaced with F3, and the fourth to sixth steps are repeated to perform addition processing for correcting the positional deviation for all the N1 images.

  In this embodiment, the finally obtained image is an image obtained by integrating N0 × N1 frames. However, since the image is blurred only by the integration of N0 frames due to the influence of drift, it is directly N0 × Compared with the case of integration in the N1 frame, image blur due to drift is reduced to 1 / N1. By providing such a sequence, it is possible to eliminate the positional shift in the two-dimensional image plane direction between images acquired at different timings, which occurs based on charge-up on the sample, etc., or to suppress image blurring, or Can be eliminated.

  An example of the results obtained in this example is shown in FIG. FIG. 3A shows an image (1280 × 960 pixels, magnification of 200,000 times) acquired by normal frame integration. Drift during image integration accumulates and appears as “blur” of the image. ing. FIG. 3B is an image obtained by acquiring 10 images having the frame integration number of 1/10 of FIG. 3A and integrating the 10 images while correcting the positional deviation. In FIG. 3B, the total number of frames accumulated in the image is the same, but only the drift accumulated in each accumulated image is “blurred” in the final image. The “blurring” of the image due to drift is also reduced to 1/10 of FIG.

  Since the amount of deviation changes according to the type of sample and optical conditions, it is desirable to set N0 and N1 according to the S / N ratio. The number of scans (number of images) necessary to ensure the required S / N ratio is determined based on the required image quality and the secondary electron generation efficiency of the sample. N0 and N1 may be determined. Also, by inputting at least one of the parameters indicating the sample conditions (ease of charge-up, etc.) and the total integration number, the frame integration number (N0), or the number of captured images (N1), It is also possible to create a sequence that determines the other two parameters. According to such a configuration, it is possible to easily set the apparatus conditions simply by inputting specifications necessary for observation.

  In this embodiment, the misalignment between the images integrated in the frame is corrected. However, the present invention is not limited to this, and the misalignment correction is executed in units of arbitrary plural frames or for each arbitrary number of captured images. Anyway. At this time, if the image to be compared for positional deviation detection does not have a certain S / N ratio above a certain level, it is considered that the deviation detection accuracy is lowered, so that a predetermined S / N ratio is secured. It is desirable to set the required number of images as the frame integration number (NO), and then set the number of captured images (N1) for obtaining the S / N ratio required for the final sample image.

  In the present invention, the positional deviation may be corrected and stored in the image memory 25, or a frame memory corresponding to the total number of frames multiplied by the number of captured images is prepared, and when displaying a sample image, it is transferred to an external image storage element. At the time of transfer, the positional deviation between the sample images may be corrected before the transfer to the external storage element or in the external storage element.

  If at least an image memory for storing composite images, an image memory for storing images before composite processing, and an image memory for images to be captured are prepared, they are captured one after another by scanning with an electron beam. Images can be synthesized sequentially.

  In this embodiment, in order to further facilitate the setting of N0 and N1, a reference image is stored for each combination of N0 and N1 for a specific sample, and this image can be read when N0 and N1 are set. If comprised in this way, it will become possible for the operator to perform appropriate setting of N0 and N1 while referring to the reference image.

  When the drift is fast, the number of deviation correction is increased by reducing the number of N0, and when the drift is not so fast, it is desirable to increase the number of N0 in order to improve the image quality of the image to be compared. . For example, as a means for appropriately setting the number of N0 and N1, a means for adjusting N0 and N1 stepwise may be provided. When the total number of accumulated sheets is set to 50, the combinations of N0 and N1 can be 1 × 50, 2 × 25, 5 × 10, 10 × 5, 25 × 2, 50 × 1, but these By providing means for adjusting and means for displaying the actually accumulated image, the operator can set appropriate N0 and N1 from the synthesized image without being particularly aware of the technique related to the present invention. It becomes.

  By providing the adjusting means as described above, it is possible to easily select an appropriate combination when the combination of N0 and N1 changes, for example, when the image quality changes when the combination of N0 and N1 is changed. It becomes possible.

  Further, a means for adjusting the degree of deviation correction is provided, and when “the degree of deviation correction is large” is set, N1 is set to be large, and when “the degree of deviation correction is small” is set, N0 is made small. The same effect can be achieved even when configured to be set as described above.

(Example 2)
The processing flow of FIG. 4 will be described in detail below.

First step (S4001):
A frame integration number N0 and an image number N1 per image are set.

Second step (S4002):
Two images with the frame integration number N0 are continuously acquired.

Third step (S4003):
A sharpened image is created from the two acquired images, and a positional deviation between the images is obtained.

  If the deviation exceeds a predetermined allowable value, each image before the position correction integration includes noticeable “blur” due to drift. It is possible to inform the operator that the size is too large by the display function.

Fourth step (S4004):
The difference between the two images is corrected and added, and this image is registered as F0.

Fifth step (S4005):
The field of view is moved in a direction to cancel the positional shift obtained in the process of S4003. As the moving means in this case, both the case of using an electric visual field moving means (image shift deflector) and the case of using a stage are possible depending on the amount of movement. Usually, when the movement amount is small, image shift is used together. If the amount of movement is large, a stage or image shift is used if necessary. By canceling the visual field shift by image shift or stage, the shift between images can be compressed even if there is a relatively large drift, so there is an effective visual field (region where the visual field overlaps between images) after correcting the positional shift by image processing. The problem of narrowing is improved.

Sixth step (S4006):
Get the next image.

Seventh step (S4007):
A sharpened image of the acquired image and a sharpened image of F0 are created, and a positional deviation between the images is obtained.

Eighth step (S4008):
F0 and the positional deviation of the image acquired in S4006 are corrected and added, and this is newly set as F0.

Ninth step (S4009):
The field of view is moved in a direction to cancel the positional shift obtained in the process of S4008.

Tenth step (S4010):
The processes S4006 to S4009 are repeated to acquire and integrate N1 images.

  According to the configuration as described above, a large drift component is corrected by the stage and beam deflection, and further, a minute drift at the pixel level is corrected at the time of image integration. A resolution image can be acquired.

  On the other hand, in order to minimize the influence of drift, it is necessary to correct the misalignment and shorten the acquisition time of each image to be integrated, but the limit is the image required for detecting misalignment. Determined by the S / N ratio. Therefore, when the drift amount exceeds a predetermined value, each image for correcting the positional deviation itself becomes blurred due to drift. When the amount of drift leading to such a result is detected, a means for displaying that it is difficult to acquire a high-resolution image or interrupting the measurement may be provided. This makes it possible to eliminate the adverse effect of operating the apparatus wastefully in a state where it is clearly difficult to acquire a sample image.

(Example 3)
The processing flow of FIG. 6 will be described in detail below.

First step (S6001):
A first image F1 for drift detection is acquired.

Second step (S6001):
The target image F0 is acquired under a predetermined slow scanning condition. By performing the slow scan in this way, more secondary electrons can be generated compared to the case of performing the fast scan, so that a high contrast image can be obtained.

Third step (S6001):
A second image F2 for drift detection is acquired.

Fourth step (S6001):
A shift amount ΔF (ΔFx, ΔFy) between the images F1 and F2 is detected.

Fifth step (S6001):
The amount of deformation in the horizontal and vertical directions of the target image is obtained from the image shift amount ΔF.

Sixth step (S6001):
The target image F0 is deformed to construct a new image F0 ′.

  Here, the process of the fifth step will be described in detail below with reference to FIG.

Assuming that the shift amount between the images F1 and F2 for drift detection captured in the image memory is ΔF (ΔFx, ΔFy) and the difference between the capture times of the images F1 and F2 is ΔT, the drift velocity (Vx , Vy)
Vx = ΔFx / ΔT, Vy = ΔFy / ΔT
Calculated by On the other hand, when the capture time of the target image F0 is T0, at the start and end of scanning of the image F0,
ΔF0x = Vx × T0 in the X direction
ΔF0y = Vy × T0 in the Y direction
This means that the field of view is only shifted. Therefore, as shown in FIG. 7, it is predicted that the target image F0 in the image memory is obtained when the drift does not occur by deforming the target image F0 in the image memory by F0x (Y direction) and ΔF0y (X direction). The image F0 ′ is restored.

  In this embodiment, the images F1 and F2 for drift detection are captured before and after the acquisition of the target image F0. However, of course, the present invention is not limited to this, and before F0 is captured or F0 is captured. Thereafter, F1 and F2 can be taken in continuously. In such a case, the deformation at the time when F0 is acquired is estimated from the displacements of F1 and F2, and the image F0 ′ expected thereon is restored.

  The images F1, F2, and F0 are stored in the image memory, respectively, and F1 and F2 are read based on the necessity determination of image restoration, and restoration processing is performed.

  According to the above configuration, it is possible to accurately grasp the shape of the observation target that is deformed by drift.

  In particular, when performing slow scanning, since the sample is continuously irradiated with the electron beam for a long time, and the sample image deformation that takes charge of the sample becomes remarkable, the application of the technique of this embodiment is extremely effective. In this embodiment, an example has been described in which two images for drift correction and one image to be corrected are described. However, the present invention is not limited to this, and an arbitrary number of images can be set.

  In the apparatus of this embodiment, an option for selecting whether or not image restoration is necessary on a graphical user interface (GUI) as shown in FIG. 16 so that the operator can judge whether or not image restoration is necessary. Is provided. FIG. 16 shows an example in which selection is performed by a pointing device or the like on the image display device, but the present invention is not limited to this, and it may be set by other known input setting means.

  While observation using an electron beam apparatus such as an electron microscope has a need to form a sample image with high accuracy, there is a need to suppress irradiation of an electron beam as much as possible to reduce sample damage. In the case of the apparatus according to the present invention, if the deformation of the sample due to drift can be suppressed, the sample image can be formed with high accuracy. Further scanning is required. That is, since the electron beam scanning time increases, the risk of sample damage due to electron beam irradiation increases.

  By providing the above options, it is possible to switch between the scanning mode for acquiring F1 and F2 and the mode for not performing scanning, and the operator restores in view of the status of the observation target or conditions for forming an appropriate sample image. The sample image can be formed under the conditions that the operator truly desires.

  In addition, by registering how much the deformation is corrected in a graph, it can be used for setting the scanning speed and determining the necessity of drift correction. In addition, the configuration is such that the correction amount and deformation amount with respect to the number of times of electron beam scanning and the irradiation time of the electron beam are stored and displayed, so that it is possible to grasp how much the electron beam is scanned to deform the sample image. It becomes possible.

Example 4
An embodiment in which the drift correction technique is applied to line profile integration will be described with reference to FIG. For measurement of pattern dimensions on a wafer, a signal distribution (using a line profile) is usually obtained when a pattern to be measured is line scanned with an electron beam. When the sample is an insulator, high-speed scanning is performed in order to prevent disturbance of the signal due to charging. Therefore, the S / N of the profile is poor with a signal obtained by one line scanning, and measurement with high reproducibility is difficult. . Therefore, normally, the signal distribution obtained by a plurality of scans is integrated to form a profile to be measured. At this time, if drift occurs in the direction of line scanning, the integrated line profile becomes dull, so that the measurement accuracy decreases.

  Therefore, each profile obtained by a plurality of line scans is stored, profile position correction is performed so that the correlation between the profiles becomes the highest, and the profiles are integrated. In this case, each pre-integration profile can be a signal acquired by one scan, but if the scanning speed is high, the signal acquired by one scan is too bad for the S / N. Thus, the line scans are simply integrated a minimum number of times within a possible range of correlation, and this can be used as each pre-integration profile. This method improves the dullness of the profile even if there is a drift, and can generate a profile with a high S / N, so that measurement with high reproducibility becomes possible.

  Further, the necessity of position correction may be selected on the setting screen described with reference to FIG. Furthermore, in semiconductor inspection equipment, the reliability of the measurement can be improved later by issuing an alarm when the position correction amount exceeds a certain threshold value and when it becomes clearly large, or by flagging the measurement. It may be possible to confirm. In this case, the correction amount, integration profile, pre-integration profile is stored, or if a line profile is formed based on a two-dimensional scan image, the integrated image or pre-integration image is stored, and the image is displayed later. You may make it display on an apparatus.

  According to the above configuration, when a measurement target is erroneously measured with a pattern in which the same element is adjacent, such as a repeated pattern, the confirmation can be easily performed.

  In particular, when a line profile is formed by performing a one-dimensional scan, the measurement speed can be increased as compared with the case of performing a two-dimensional scan. On the other hand, the measurement accuracy cannot be confirmed by referring to the sample image. In this embodiment, it is possible to form a line profile with high accuracy even when one-dimensional scanning capable of realizing high-speed measurement is performed. For example, pattern length measurement is performed based on the line profile. Even with such an apparatus, it is possible to perform highly accurate length measurement based on a line profile formed with high accuracy.

  When measuring the pattern width of a line pattern having roughness, the length measurement range is expanded in a direction perpendicular to the length measurement direction, and length measurement based on the line profile is performed at a plurality of different locations therebetween. On that basis, it is conceivable to average a plurality of length measurement values, or to measure a variance value of roughness based on a plurality of length measurement values obtained within an expanded length measurement range. In this case as well, the present embodiment can be applied.

  For example, if line profile integration with position correction is performed in units of a plurality of measurement points, and average values and dispersion values are measured after that, these results can be obtained with high accuracy.

  Rather than performing position correction in units of measurement points, if the position profile of another part is corrected with respect to a certain reference line profile, the line profile may be shifted at each measurement point due to charge-up etc. Even in this case, the integrated line profile can be properly formed and accurate length measurement can be performed. According to such a configuration, it is possible to easily realize the reliability evaluation of the electrical characteristics of the semiconductor element pattern and the like regardless of the presence or absence of roughness.

  Also, when measuring at multiple locations, if the measured value at one location is extremely different compared to other locations, part of the line pattern may be extremely thin or the measurement may fail. There is also a possibility. In this case, an error message may be issued or measurement results such as sample images and line profiles and measurement conditions at that time may be registered together, and the data may be read and confirmed later.

(Example 5)
FIG. 9 is a diagram for explaining an example in which the same pattern is measured a plurality of times and a correct dimension is predicted from a change in the pattern dimension with time. Some measurement objects using an electron beam are damaged by electron beam irradiation and contract or evaporate. In this case, as the beam irradiation amount increases, the pattern size becomes smaller, so the measurement itself becomes an error factor.

  The same pattern is measured multiple times in order to predict the error produced by the measurement itself and evaluate the correct dimensions. Since the beam irradiation amount increases in proportion to the number of measurements, the deformation of the pattern also increases accordingly. Therefore, by obtaining the relationship between the number of measurements (proportional to the beam irradiation amount) and the dimension measurement value from a plurality of measurement results, it becomes possible to predict the pattern dimension before the beam irradiation or before the contraction at the start of the beam irradiation. . The apparatus according to the embodiment of the present invention incorporates a sequence for automatically performing the above dimension prediction.

  In order to determine later whether or not the dimension prediction has been accurately performed, a table in which the measured values with respect to the number of measurements are graphed may be stored so that it can be output to a display device or an external output device later. For example, in the case where drift occurs simultaneously with the contraction of the observation target, there is a possibility that the size prediction is not properly performed due to the influence of the drift. Therefore, the operator can properly predict the size by referring to the previous graph. It can be confirmed whether it was a thing. If the sample image obtained at that time is stored in association with the stored graph, the appropriateness of the size prediction can be confirmed while referring to the sample image.

  When the graph records an abnormal transition, the graph may be selectively stored or a predetermined flag may be set. For example, if an abnormal displacement is observed in the graph showing the transition of the dimensional displacement, something may have occurred in the electron beam device at that time, and the dimensional prediction may not be performed properly. is there. If the graph and sample image at that time can be selectively confirmed, the operator can efficiently confirm the appropriateness of the dimension prediction without performing unnecessary confirmation.

  In the description of the present embodiment, the horizontal axis of the graph is “measurement count”, but the present invention is not limited to this. For example, parameters such as “scan count” and “time” may be used. The vertical axis is not limited to the “measured value”, and may indicate a ratio to a normal value (design value).

  When this embodiment is used in an apparatus for measuring a semiconductor pattern, etc., a dummy pattern having the same conditions as the measurement target pattern is created in the vicinity of the measurement target pattern, thereby contributing to the operation of the semiconductor element. Accurate measurement can be performed without shrinking the pattern.

(Example 6)
FIG. 10 shows an embodiment in which an image is acquired with a number of pixels larger than the number of pixels of the target image, and the positional deviation between the acquired images is corrected. In the present embodiment, for example, a case where the number of pixels of the target image is composed of 512 × 512 pixels is shown. In this example, the acquired image is acquired with 1024 × 1024 pixels. When the target image is integrated after correcting the positional deviation, an area that cannot be used as the target image is generated due to the deviation between the images. In this embodiment, an image is acquired in advance in an area wider than the number of pixels of the target image, and after integration, a 512 × 512 pixel area in the center is cut out to obtain a final target image.

  By acquiring a slightly larger image in this way, the end of the image is not cut off and cannot be seen even if drift occurs.

(Example 7)
FIG. 11 shows an embodiment in which abnormal images are removed and images are integrated. When images are abnormally shifted due to sudden disturbance while acquiring multiple images, or when an image is abnormally blurred, or an image after a certain image shows abnormal contrast due to charge during beam irradiation These images can be excluded from the original image to be integrated by detecting an abnormality by image processing. Abnormality detection can be performed by prescribing a visual field deviation amount that is regarded as abnormal in advance. With regard to blurring, abnormal images can be removed by performing calculation such as image differentiation and prescribing a threshold value for determining abnormalities in advance. Abnormal contrast can be removed by determining the histogram or determining that the correlation value with another image after visual field correction is abnormally decreased. By removing the abnormal information in this way, it is possible to stably acquire a high resolution image even if there is an unexpected cause.

  The image determined to be abnormal is removed, but in order to investigate the cause of the abnormality later, the image determined to be abnormal is stored in another image memory, and when the abnormality is recognized or abnormal The optical conditions (acceleration voltage of the electron source, emission current, etc.) before and after the determination are stored. According to such a configuration, it is easy to confirm for what reason an abnormal image has occurred. For example, if the timing at which the overcurrent flows to the cathode of the electron source coincides with the timing at which the abnormal image occurs, the cause is in the electron source and can be used as an index for replacing the electron source.

  Changes in current and voltage applied to optical elements such as extraction electrodes, acceleration electrodes, and scanning coils in an electron microscope are displayed in a time chart, and the timing at which an abnormal image is generated is superimposed on the time chart. By doing so, the operator can identify the cause visually.

  The abnormal frame removal technique described in the present embodiment is also applicable to the line profile integration described in the fourth embodiment.

  In this embodiment, an example in which abnormal images are mainly automatically removed has been described. However, the present invention is not limited to this example. For example, an image before integration is displayed on an image display device, and an image that an operator recognizes as abnormal is selectively displayed. A function that can be removed may be provided. At this time, if the configuration is such that a plurality of pre-integration images displayed side by side on the image display device can be selected by a pointing device or the like, the operator can select an image to be visually removed from the plurality of pre-integration images. Can do. A plurality of accumulated images may be displayed in order to identify an abnormal image with an image having a certain S / N as well as an image before accumulation.

(Example 8)
FIG. 12 illustrates an example of image integration for correction of visual field deviation of images acquired with a plurality of image signals. For example, when attempting to acquire an integrated image based on reflected electron signals, since the amount of reflected electron signals is generally small, it is necessary to increase the number of frames of each original image. This is because if an image with a small signal amount is acquired with a small number of frames and used as an original image, the S / N of the original image is extremely reduced, and a visual field shift between images cannot be detected. On the other hand, if the number of frames of the original image is increased, the time for capturing the original image becomes longer, so that the original image itself is blurred due to drift. In the present embodiment, a secondary electron image having a good S / N acquired simultaneously with the reflected electron signal is used to detect a field shift between images in each original image, and a plurality of the field shift amounts obtained by the reflected electron signal are detected. It is applied to the visual field shift between images.

  Since the secondary electron image and the backscattered electron image are acquired at the same time, and the field of view of the corresponding secondary electron image and the backscattered electron image completely coincide, It is possible to correct the field of view accurately with respect to the original image. By using a secondary electron signal image having a high S / N as a detected image of the amount of visual field deviation, the number of frames constituting the original image can be minimized, so that the original image itself does not blur due to drift. Examples of signals with poor S / N include the reflected electron signal shown in the present embodiment, for example, an X-ray signal and a sample absorption current, and the present embodiment is applied to various signals. be able to. In particular, when the element distribution (X-ray image) of the thin film sample is acquired with high resolution, the secondary electron signal in this embodiment can be used as a transmission electron signal. In general, by thinning a sample to several tens of nm, X-ray scattering within the sample does not occur, and an element distribution image with high resolution can be obtained.

  A configuration as shown in FIG. 17 is conceivable as a detection system for simultaneously detecting secondary electrons and reflected electrons. According to this configuration, two types of electrons (reflection) are emitted from the sample 1705 using the backscattered electron detector 1703 and the secondary electron detector 1704 respectively disposed above and below the objective lens 1702 that converges the primary electron beam 1701. The electrons 1706 and the secondary electrons 1707) can be detected simultaneously.

  Further, by using a detection system as shown in FIG. 18, secondary electrons and reflected electrons can be detected together. In the case of the configuration in FIG. 18, secondary electrons and reflected electrons 1803 emitted from the sample 1801 are accelerated by the retarding voltage 1802 applied to the sample 1801, and a secondary electron conversion electrode 1805 disposed on the objective lens 1804. Collide with. Secondary electrons and reflected electrons 1803 accelerated during the collision generate secondary electrons 1806 which are attracted to and detected by the secondary electron detector 1807.

  The energy filter 1808 can be applied with an energy filter voltage 1809 that is equal to or slightly higher than the retarding voltage 1802 applied to the sample. By applying such a voltage, only reflected electrons are selectively selected. It passes through the energy filter 1808.

  In the configuration as described above, each time a predetermined number of two-dimensional images are acquired, the voltage of the energy filter voltage 1809 is turned on / off or switched from strong to weak so that a secondary electron image and a reflected electron image are obtained alternately. . Then, the misalignment was detected using the secondary electron image, and the retarding technique was adopted by correcting the misalignment of the reflected electron image using the detected misalignment information and storing it in the image memory. In a scanning electron microscope, a reflected electron image without image blur can be obtained. In this embodiment, the reflected electrons and the secondary electrons are clearly separated. However, the present invention is not limited to this, and an energy filter voltage 1809 lower than the retarding voltage 1802 is applied to the energy filter 1808 to thereby detect the secondary electron detector 1807. The amount of electrons detected in may be increased. Since most of the electrons emitted from the sample are 50 eV or less, the number of frames constituting the original image can be minimized by using at least 50 eV or less of electrons as a detection image of the visual field shift amount. The voltage applied to the energy filter 1808 can be changed according to the purpose of analysis.

  The backscattered electron detector and the secondary electron detector are not limited to those described in the present embodiment, and various forms can be adopted. The X-ray detector is not particularly shown, but the existing X-ray detectors in general can be applied.

Example 9
FIG. 13 shows an embodiment in which images are integrated by correcting the field deviation only in a specific direction on the sample surface with respect to a plurality of images acquired. In the case of an image having a pattern only in a specific direction in the image, it can be detected with high accuracy for detecting a positional deviation in a direction orthogonal to the pattern, but in a direction parallel to the pattern. In this case, the positional deviation detection accuracy is extremely lowered. For such an image, by integrating the image with the field of view only in the direction orthogonal to the pattern, errors during field alignment can be prevented. The direction of the pattern can be specified by analyzing the frequency component of the image or binarizing the image into a line drawing.

  Even in the case where the visual field deviation is corrected only in a specific direction as described above, the accuracy of the length measurement result can be maintained in the apparatus for measuring the line width of the pattern on the semiconductor wafer. The pattern on the semiconductor wafer is often formed in a straight line, and the line width is almost the same no matter where it is measured on the same line pattern, so even if there is no deviation in the direction perpendicular to the pattern Accurate measurement is possible.

  When the target image is a line pattern and the two images to be integrated have a visual field shift as shown in FIG. 20A, the relationship between the image shift amount and the degree of coincidence is as shown in FIG. In FIG. 20B, blurring at the time of integration is corrected by overlapping images under the condition that the degree of coincidence is maximized. However, in a line-shaped pattern, the condition that maximizes the degree of coincidence is not a single point but a line shape. Exists. Therefore, the condition for overlapping images (maximum degree of matching) cannot be uniquely determined. Therefore, if the direction in which the image is shifted is orthogonal to the pattern, the shift amount between the images can be minimized and the blurring of the pattern can be corrected, so that the effective field of view of the integrated image can be maximized.

(Example 10)
FIG. 14 shows a specific method of detecting the amount of misalignment and adding the corrected image. In the input image 1401 and the input image 1402, a region 1403 having an appropriate size, for example, at the center of the input image 1401 is taken, and this is used as a template to perform template matching on the input image 1402. As a result, if the region 1404 is matched, the region 1403 and the region 1404 are overlapped, and a rectangular region (AND region) 1405 where the input image 1401 and the input image 1402 overlap is considered, and the AND region of the input image 1401 and the input image 1402 is considered. The portions that do not overlap 1405 are cut off to obtain input images 1406 and 1407 after alignment. Addition processing is performed using the post-positioning input images 1406 and 1407 as inputs.

  In this example, two input images are shown, but it is easy to expand to three or more. As an example of template matching, if the size of the input image is 512 × 512 pixels, there is a method in which normalized correlation processing is performed between the two pixels using 256 × 256 pixels in the central portion of the template. In this case, a position having the highest calculated correlation value is set as a matched position.

r (x, y) is the correlation value at (x, y), M _ij is the density value at point (i, j) in the template, and P _ij is the density at the corresponding point (x + i, y + j) in the input image. N is the number of pixels in the template.

  FIG. 15 shows another embodiment of the correction image addition method. As in FIG. 14, in the input image 1401 and the input image 1402, an area 1403 of an appropriate size, for example, at the center of the input image 1401 is taken, and template matching is performed on the input image 1402 using this as a template. As a result, if the region 1401 is matched, the region 1403 and the region 1404 are overlapped, and a rectangular region (OR region) 1501 including both the input image 1401 and the input image 1402 is considered, and the input image 1401 and the input image 1402 are Portions that do not overlap with the OR region 1501 are added, and the number of pixels is filled with 0 or an average value of the respective input images to obtain post-positioning input images 1502 and 1503. Addition processing is performed using the input images 1502 and 1503 after alignment as inputs. In this example, two input images are shown, but it is easy to expand to three or more.

(Example 11)
An example will be described in which the drift correction technique of the present invention is applied to automatic operation with a scanning electron microscope for semiconductor inspection. In order to perform normal automatic operation, a recipe file in which information such as measurement positions and observation conditions is registered in advance is created, and measurement positioning, observation, and measurement are executed according to this file. This method registers the environment to be set before executing the recipe file. A recipe execution environment screen is shown in FIG.

  The main sequence at the time of executing the recipe is to perform alignment for detecting the position of the wafer on the stage. At this time, image recognition is performed based on the image registered when the recipe is created. Next, the stage is moved to the measurement position, and an image is acquired at a relatively low magnification. Measurement pattern positioning is performed with high accuracy by image recognition (referred to as addressing), electrically deflected, and zoomed to a measurement magnification to measure pattern dimensions. Automatic focusing is performed before pattern positioning or measurement.

  When performing a recipe test or when there are a plurality of measurement wafers, the drift amount in the time from the pattern positioning image acquisition to the measurement image acquisition is measured for each measurement point on the first sheet. In the case of alignment, the drift amount after a few seconds is measured and stored. When the recipe is executed, the drift amount of each measurement point or alignment point is added as an offset after positioning. In the case of addressing, it is deflected to the position where the offset is added and zoomed to the measurement magnification. In this way, drift after positioning can be reduced, and it is not necessary to detect the drift amount of the second and subsequent measurement wafers during actual measurement using a recipe. Measurement is possible. Whether or not to perform drift correction in alignment, addressing, and measurement is determined by whether the drift correction switch for alignment 1901, the drift correction switch for addressing 1902, and the drift correction switch for measurement 1903 are ON or OFF, respectively.

  By providing an environment setting screen as described in the present embodiment, it is possible to arbitrarily set a specific technique for drift correction that changes depending on the measurement conditions and the sample state.

  In recent semiconductor manufacturing / inspection, it is common to store a plurality of semiconductor wafers in a cassette and handle them in cassette units. In such an apparatus that continuously measures a plurality of measurement objects, whether to perform drift correction based on the correction amount registered at the time of recipe creation, or to perform correction based on the drift amount actually measured for each wafer, A means for selecting is provided. With this configuration, the operator can determine whether to give priority to the measurement accuracy when there is an individual difference between the semiconductor wafers in the cassette or to give priority to the throughput, and reflect the result in the measurement.

  Means for selecting whether to perform correction based on a registered value registered at the time of recipe setting or to register a value obtained by detecting a drift amount in the first cassette in the cassette and use it for correction when performing offset correction Is provided. By configuring in this way, the operator determines whether to give priority to measurement accuracy when there is a manufacturing error between the test pattern or design value and the actual wafer pattern, or to give priority to throughput, and reflect it in the measurement. Can be made.

  The above description has been given of an example in which the operator selects a specific correction method. However, the present invention is not limited to this, and for example, by inputting the magnitude or presence / absence of a manufacturing error, the above specific correction method can be selected. A sequence for automatic setting may be provided.

  Although the embodiments so far have been described for the scanning electron microscope, the present invention is not limited to this, and the present invention can be applied to other charged particle beam apparatuses in which the sample image is shifted due to some drift generation factor.

It is a schematic block diagram of the scanning electron microscope for demonstrating one Example of this invention. It is a figure which shows the processing flow which correct | amends the positional offset of the some acquired image, and reconstructs a target image. It is a figure which shows the image acquired by the simple accumulation | accumulation, and the image which correct | amended the position shift and acquired the accumulation | storage after acquiring several images. It is a figure which shows the processing flow which combined the drift correction | amendment by control of the irradiation position of a beam, or a sample position, and the process which correct | amends and accumulate | stores the positional offset of several acquired images. It is a conceptual diagram which performs an integrating | accumulating process, correct | amending the position shift between several images. It is a figure which shows the processing flow for repairing the deformation | transformation by the drift of the image taken in by slow scanning. It is a conceptual diagram of the process which repairs the deformation | transformation of the slow scanning image by drift. It is a conceptual diagram which correct | amends and accumulate | stores the positional offset between each profile with respect to the several profile obtained by line scanning. It is a figure which shows the example which calculated the length measurement value which is not influenced by damage by estimating beam damage from the length measurement value of the image acquired by two or more sheets. It is a figure which shows the example which extracted the several image in the area | region wider than the target visual field, and cut out the center part in the area | region of the target visual field after the image integration which corrected the visual field shift | offset | difference between images. It is a figure which shows the example which correct | amended and integrated | accumulated the visual field shift | offset | difference except the image in which abnormality was detected from the acquired some image. It is a figure which shows the example which corrected and integrated | accumulated the visual field shift | offset | difference of the several image acquired by detecting a several image signal simultaneously. It is a figure which shows the example which correct | amended and integrated | accumulated the position shift of only the specific one direction with respect to several images. It is a figure for demonstrating the detection method of a positional offset amount, and the addition method of a correction image. It is a figure for demonstrating the other correction image addition method. It is a figure which shows an example of the GUI screen displayed on an image display apparatus. It is a figure which shows an example of the electron detection system of an Example of this invention. It is a figure which shows another example of the electron detection system of an Example of this invention. It is a figure for demonstrating the other example of the GUI screen displayed on an image display apparatus. It is a figure for demonstrating the principle of Example 9. FIG.

Claims (6)

A length measuring method for scanning a charged particle beam on a sample, forming an image of the sample based on a secondary signal emitted from the sample, and measuring a dimension of a pattern formed on the sample using the image In
Obtaining a plurality of images for a predetermined region of the pattern;
Specify one direction for measuring the predetermined area of the pattern,
Wherein comparing the plurality of first image and the other image of the image in the direction of displacement to perform the length measurement selectively corrected, integrating the image in which the shift amount is selectively corrected,
A length measurement method comprising performing the length measurement using the integrated image. In the length measuring method of Claim 1,
The method for forming a sample image is characterized in that the length measurement direction is a direction orthogonal to a longitudinal direction of a line pattern formed on the sample. In a length measuring device that measures the dimensions of a pattern formed on a sample,
A scanning electron microscope for acquiring a plurality of images for a predetermined region of the pattern;
Image processing means for processing an image acquired by the scanning electron microscope,
The image processing means
Specify one direction for the length measurement from the image of the predetermined area,
Wherein comparing the plurality of first image and the other image in the image displacement amount in the direction performing the length measurement selectively corrected, integrating the image in which the shift amount is selectively corrected,
A length measuring apparatus that performs the length measurement using the integrated image. In the length measuring device according to claim 3,
The length measuring apparatus characterized in that the image processing means analyzes the frequency component of the image or binarizes the image when specifying the length measuring direction. In the length measuring device according to claim 3,
The image processing means includes
A length measuring apparatus that corrects a deviation amount with respect to a direction orthogonal to a longitudinal direction of the pattern. In the length measuring device according to claim 3,
A length measuring apparatus that corrects the shift amount by shifting an image so that the degree of coincidence between the first image and another image in the length measuring direction is maximized.

Images