JP2012142299A - Scanning charged particle microscope device, and processing method of image taken by scanning charged particle microscope device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To increase the precision of image processing by taking a high-quality sample image with reduced noise components by use of a scanning type charged particle microscope.SOLUTION: The scanning charged particle microscope device is capable of taking a high-resolution image in various conditions by calculating a beam intensity waveform based on imaging conditions and sample information, and by carrying out image restoration by using a resolution deterioration factor as a subject of a deterioration model in addition to the beam intensity waveform. In the scanning charged particle microscope designed for semiconductor tests and measurements, the use of an image after image restoration to perform a pattern size measurement, defect detection, and defect classification or the like enables the improvement of the precision of measurement and the rise in the precision of detection and classification of a defect.

Description

  The present invention relates to a scanning charged particle microscope that scans a sample surface using charged particles such as a scanning ion microscope or a scanning electron microscope to acquire an image, and in particular, improves resolution by image processing on the acquired image. The present invention relates to a scanning charged particle microscope capable of improving S / N and an image restoration method of an image acquired by a scanning charged particle microscope.

  In order to clearly observe a minute object, a scanning charged particle microscope having a very high resolution as compared with an optical microscope is widely used. In a scanning charged particle microscope, a focused electron beam is irradiated onto a target sample, and charged particles that are emitted from the sample or transmitted through the sample (which may be different from the charged particles to be irradiated) are detected. Information on the target sample at the beam irradiation position is obtained. By scanning the sample with a charged particle beam, an enlarged image of the target sample can be acquired. A scanning ion microscope (hereinafter referred to as SIM) and a scanning electron microscope (hereinafter referred to as SEM) are well known as one of scanning charged particle microscopes. In particular, in the semiconductor manufacturing process, not only for image observation, but also for the purpose of obtaining feature quantities of target samples such as semiconductor inspection and pattern measurement, detection of defects generated on semiconductor wafers, defect cause investigation, patterns It is used for the measurement of dimensions and shapes. With the miniaturization of patterns, the need for inspection of minute defects and high-precision measurement of patterns has increased, and it has become important to acquire high-resolution images. However, any scanning charged particle microscope has limited resolution. Due to diffraction aberration caused by the wave nature of the particles, chromatic aberration due to lens characteristics, and spherical aberration, the charged particle beam (ion beam, electron beam, etc.) has a beam intensity waveform spread by these aberrations. And enter the sample surface.

  The charged particle beam incident on the sample is generally diffused in the sample and then emitted from the sample or transmitted through the sample. These phenomena cause resolution degradation. Further, in the scanning charged particle microscope, there is a problem that the more the charged particle beam is incident, the more the sample is damaged or the imaging time becomes longer. For this reason, a sufficient amount of charged particle beam cannot be incident, and the amount of signal to be detected is reduced, so that the S / N of the captured image is lower in the scanning charged particle microscope than in the optical microscope. On the other hand, image restoration is known as image processing that can improve image resolution and S / N. Images obtained with devices such as cameras, telescopes, and microscopes always have resolution degradation and noise superposition. Image restoration is processing for estimating a clear and high S / N image (hereinafter referred to as an ideal image) obtained by removing resolution degradation and noise from these images. In normal image restoration, the resolution of the ideal image is first degraded by convolution with a degradation function indicating the resolution degradation level, and then an image with noise superimposed is modeled as a captured image, and then this model is used to reverse the captured image. Estimate the ideal image. Regarding image restoration, many studies have been conducted on celestial images and optical images (see, for example, Non-Patent Documents 1 and 2). Also in the scanning optical microscope and the scanning charged particle microscope, Patent Document 1 and Non-Patent Document 3 propose a technique for improving resolution using image restoration. Furthermore, for example, Non-Patent Document 4 describes a calculation method for obtaining a beam intensity waveform on the sample surface, which is one of the factors that degrade the resolution of a captured image.

  Further, Patent Document 2 describes that information on the side surface of a sample is obtained by irradiating the sample with an electron beam from an oblique direction to perform image capturing.

Japanese Patent Laid-Open No. 3-44613 JP 2001-15055 A A.K.Katsaggelos: Optical Engineering, 28, 7, pp.735-748 (1989) M.R.Banham and A.K.Katsaggelos: IEEE Signal Processing Magazine, pp.24-41 (Mar.1997) Y.I.Gold and A.Goldenshtein: Proc.SPIE, 3332, pp.620-624 (1998) J. Orloff: Handbook of Charged Particle Optics, CRC Press (1997)

  However, in the methods proposed in Patent Document 1 and Non-Patent Document 3, a fixed beam intensity waveform is used for the deterioration function, which has the following problems.

  For example, the beam intensity waveform may vary greatly depending on parameters that determine imaging conditions such as acceleration voltage, probe current, and beam opening angle, and also varies depending on the beam tilt angle when imaging with the beam tilted. Furthermore, the beam intensity waveform is easily affected by the performance of a lens or the like that focuses the charged particle beam, and even if it is a different device or the same device, the performance of the lens may vary depending on the state of the device. . Unless a degradation function suitable for these imaging conditions is used, resolution degradation due to the beam intensity waveform cannot be sufficiently reduced.

  For example, when the target sample has a sufficient height compared to the depth of focus, the beam intensity waveform varies greatly depending on the height of the sample. The reason for this is that, since a single image is picked up with a fixed focal position, the focus is focused when the focal position is close to the sample surface, but conversely, the farther from the sample surface, the more defocusing occurs. In this case, unless the deterioration function is changed according to the height information of the sample, the resolution of the out-of-focus region cannot be sufficiently improved.

  In addition, for example, in order to further improve the resolution, there are cases where it is desirable to reduce resolution degradation factors other than the beam intensity waveform, such as the influence of the diffusion of the charged particle beam in the sample and the influence of the beam scan.

  In addition, for example, when measuring the dimensions and shape of semiconductor patterns, in order to achieve stable measurement accuracy, it is necessary to perform processing to prevent measurement fluctuations due to differences in measured values between devices and changes over time of devices. It is. In this case, the major cause of the difference or fluctuation in the measured value is considered to be the difference in the beam intensity waveform. Therefore, the idea of reducing the difference in the beam intensity waveform by performing image restoration using the beam intensity waveform itself as a degradation function. Is valid. However, when the S / N of the captured image is low and image restoration must be performed in a short time, it is difficult to sufficiently improve the resolution. Therefore, even if the beam intensity waveform itself is used as the degradation function, It is very difficult to sufficiently reduce the difference in beam intensity waveform.

  In order to solve the above problems, in the present invention, charged particle beam irradiation optical system means for irradiating and scanning a focused charged particle beam on a pattern-formed sample, and this charged particle beam irradiation optical system means Charged particle detection optical system means for detecting charged particles generated from a scanned sample irradiated with a charged particle beam, and an image for obtaining a charged particle image of the sample by processing a signal detected by the charged particle detection optical system means In a scanning charged particle microscope apparatus including an acquisition unit and an image processing unit that processes a charged particle image of a sample acquired by the image acquisition unit, the image processing unit acquires an image acquisition condition of an image acquired by the image acquisition unit Alternatively, a restored image of the image acquired by the image acquisition unit is obtained using a deterioration function calculated using imaging information including at least one of information on the sampled image.

  Further, in the present invention, charged particles generated from a sample are detected by irradiating and scanning a charged particle beam focused by a scanning charged particle microscope apparatus on the sample on which the pattern is formed. In a processing method of an image acquired by a scanning charged particle microscope apparatus that acquires a particle image and processes a charged particle image of the acquired sample, at least one of an image acquisition condition of the acquired image or information of the imaged sample is acquired. A restored image of the acquired charged particle image is obtained using a deterioration function calculated using imaging information that includes the image pickup information.

  According to the present invention, in a scanning charged particle microscope, image reconstruction processing is performed using a deterioration function adapted to imaging conditions, sample information, and application, thereby observing a fine structure and calculating a feature amount of a target sample with high accuracy. Is possible.

It is one Example of the sequence which images and images-restores. It is a basic composition of SEM which is one embodiment of the present invention. It is an example of the degradation function for reducing the difference in resolution in two degradation models. It is one Example of the image restoration based on an iterative method. It is an example of the beam intensity waveform on the sample surface, which is one of the resolution degradation factors of the captured image. It is one Example showing the relationship between an aberration and imaging conditions. It is one Example of the step which calculates a beam intensity waveform according to imaging conditions. It is one Example of the step which calculates a beam intensity waveform according to imaging conditions. It is an example of the beam intensity waveform at the time of irradiating a charged particle beam to the sample with height. It is an example of the diffusion area | region of a charged particle beam. It is an example of the degradation function related to the scanning direction of a charged particle beam. It is one Example of the step which calculates the degradation function for reducing the resolution degradation by the beam intensity waveform and the resolution degradation by the spreading | diffusion of the charged particle beam in a sample. It is an example of a GUI screen that prompts the user to correct the degradation function. It is one Example of the sequence which performs a defect detection or defect classification using the decompression | restoration image obtained by image restoration in SEM for semiconductor inspections, such as a SEM inspection apparatus and defect review SEM. It is an example of the sequence which images a defect image automatically and classifies in defect review SEM. This is an example of a sequence for performing pattern dimension measurement and shape measurement using a restored image in a semiconductor measurement SEM such as a length measurement SEM. This is an example of a sequence in which an evaluation point image is automatically captured and a dimension and shape are measured in a semiconductor measurement SEM. It is one Example of the sequence which adjusts resolution by performing a restoration image with respect to the captured image with low resolution among the two images imaged on different imaging conditions, and performs pattern dimension measurement and shape measurement after that. It is one Example of the sequence which measures the state of an apparatus. It is an example of a GUI screen that prompts the user to correct the degradation function.

Embodiments according to the present invention will be described with reference to the drawings.
[Description of the principle of image restoration]
The present invention restores an image using an appropriate deterioration function in order to improve the resolution and S / N of a captured image. The principle of image restoration according to the present invention will be described below.

  FIG. 1 is an example of a sequence for capturing an image and performing image restoration. First, based on preset imaging conditions, an image is captured in 101 steps to obtain 111 captured images g. The imaging conditions are the device model, device ID, device state, and acceleration voltage, probe current, beam opening angle, beam tilt angle, focus position, etc., representing parameters at the time of imaging. . In step 102, a degradation function A is generated based on the imaging conditions and sample information. Next, by performing image restoration in step 103, 112 restored images f are obtained. Further, a function that allows the user to correct the deterioration function as in steps 104 and 105 may be provided. In step 104, the user is asked whether or not the deterioration function needs to be corrected. If necessary, the deterioration function corrected in step 105 is used to perform image restoration in step 103 to correct the deterioration function. Repeat until no longer needed. In the case of continuously capturing images, the processing of steps 104 and 105 may be performed on the first captured image, but the processing of steps 104 and 105 is usually performed on the second and subsequent captured images. Is not done.

  FIG. 2 shows the basic configuration of an SEM that is an embodiment of the present invention. The SEM includes, for example, an imaging device 201, a control unit 221, a processing unit 222, a storage unit 223, an input / output unit 224, and the like. In obtaining a captured image, a primary electron beam 203 is generated from the electron gun 202, and the primary electron beam 203 is focused on the surface of the sample 206 by passing through the condenser lens 204 and further passing through the objective lens 205. Next, electrons such as secondary electrons and reflected electrons generated from the sample 206 are detected by the detector 208, and a digital image is generated from the detection signal by the image generator 209, thereby acquiring a captured image. The captured image is stored in the storage unit 223. By moving the stage 207, it is possible to capture an image at an arbitrary position of the sample. The detector 208 may include a plurality of detectors such as a secondary electron detector that detects a large amount of secondary electrons and a reflected electron detector that detects a large amount of reflected electrons. Further, there may be a height measurement sensor 214 for measuring the height of the sample. Information on the side surface of the sample can be obtained by irradiating the sample with an electron beam from an oblique direction and capturing an image like the primary electron beam 210. As a method of irradiating an electron beam from an oblique direction, for example, Patent Document 2 discloses that a chromatic aberration on a sample surface is substantially reduced by dispersing the electron beam and a deflection unit 211 for deflecting the electron beam away from the optical axis. A method of tilting the electron beam by using the correction unit 212 for correcting the above is described. In addition to such a method, it is possible to take an image by irradiating a sample with a charged particle beam from an oblique direction by a method using a plurality of columns, a method of tilting the column, or a method of tilting the stage 207. is there. However, normally, the beam intensity waveform spreads as the sample is irradiated obliquely, and the resolution of the captured image decreases.

  The control unit 221 controls the voltage applied to the periphery of the electron gun 202, the focal position adjustment of the condenser lens 204 and the objective lens 205, the movement of the stage 207, the operation timing of the image generator 209, and the like. The processing unit 222 generates a deterioration function in step 102, restores an image in step 103, determines whether or not the deterioration function needs to be corrected in step 104, corrects the deterioration function in step 105, and the like.

  The storage unit 223 stores captured images, restored images, imaging conditions, sample information, and the like. Input of imaging conditions, output of captured images or restored images, correction of deterioration functions, and the like are performed by the input / output unit 224. In the captured image g, resolution degradation due to some factor and S / N reduction due to noise superimposition always occur. In the captured image g (x, y), noise n (x, y) is superimposed on the convolution of the degradation function A (x ', y') representing the state of resolution degradation and the restored image f (x, y). As an image,

It can be expressed as. Here, the restored image f is an image without resolution degradation due to the degradation function A and no superimposition of noise n. The noise n is often assumed to be white Gaussian noise independent of the restored image f, but may be noise that is not independent of f or noise other than Gaussian distribution, for example, according to Poisson distribution.

  Also, noise does not have to be additive, for example

Noise that can be expressed as shown in FIG. The restored image f does not necessarily need to be an image excluding all resolution deterioration factors and all noises. For example, if there are two resolution degradation factors represented by A ₁ (x ', y') and A ₂ (x ', y'), the degradation function A (x ', y')

In addition, the restored image f can be set so that these two resolution degradations do not occur, and the degradation function A (x ′, y ′) is set to A (x ′, y ′) = A ₁ (x ′, y It is also possible to reduce only one of the resolution degradations by setting ') or A (x', y ') = A ₂ (x', y ').

In addition, for example, it is practically impossible to completely remove noise from an image, and there is a possibility that a fine structure is removed as noise is strongly removed. In order to prevent this, it is possible to leave some noise in the restored image f. Possible causes of resolution degradation include the spread of the beam intensity waveform on the surface of the sample, the diffusion of the beam within the sample, and the effects of beam scanning. By performing image restoration using a degradation function corresponding to each resolution degradation factor, resolution degradation due to each phenomenon can be reduced. Further, the degradation function does not need to directly correspond to these resolution degradation factors. For example, when the beam intensity waveforms on the sample surface under different imaging conditions are functions A _a (x ′, y ′) and A _b (x ′, y ′) as shown in FIG.

A function A (x ′, y ′) that satisfies the above may be used as a degradation function. By using this deterioration function, it is possible to reduce the difference in resolution between the two beam intensity waveforms. As a calculation method of the function A (x ′, y ′), for example, the function A _a (x ′, y ′) and A _b (x ′, y ′) are Fourier-transformed, and the function FA _a ( f _x ', f _y '), FA _b (f _x ', f _y ') division FA _a (f _x ', f _y ') / FA _b (f _x ', f _y ') There is a method of inverse Fourier transforming the result. In the image restoration, the restored image f is obtained by solving Equation (1) in reverse for the given captured image g. As a solution for image restoration, the inverse function of the degradation function A or a function A ⁺ corresponding to the inverse function is obtained, and a non-iterative method in which A ⁺ is directly applied to the captured image g, or iterative processing is performed. There is an iterative method for obtaining a restored image f with high resolution. Non-iterative methods generally have the advantage that a restored image f can be obtained at high speed, and methods such as the Wiener filter are well known. On the other hand, the iterative method can generally obtain a restored image f with high resolution and good image quality.

FIG. 4 is an example of image restoration based on an iterative method. The iterative method, by updating iteratively the image f _i Yuki performing resolution enhancement and noise reduction of an image f _i, is a method of obtaining a restored image f. First, create an image f ₀ is the initial value of the image f _i in step 401 by using the captured image g. It images f ₀ may be captured image g per se, pictures and subjected to preprocessing such as noise removal to the captured image g, may be restored image obtained by the other image reconstruction method. Next, in step 402, an image g ₀ that is a convolution result of the image f ₀ and the deterioration function A is calculated. Then, the captured image g in 403 images f _i updating step, the image f _0, and updates the image f ₀ using image g _0, obtain an image f _1.

Hereinafter, until the end condition is satisfied in step 404, it updates the image f _i by repeating steps 402-404, and outputs the image f _i as the restored image f When the end condition is satisfied. Termination condition, constant or after iterations performed, after a certain processing time, or when the update of the image f _i is sufficiently small, such as when certain conditions are met there is an image f _i is considered. The 403 image f _i update step have been proposed many methods. For example, in the Richardson-Lucy method, which is widely known as an iterative method,

The image f _i is updated according to the following. In this method, the image f _i converges to the maximum likelihood solution when the noise follows the Poisson distribution. Also, for example, constraint conditions

There is also a method of updating f _i (x, y) so as to make the evaluation function H (f _i (x, y)) smaller. Here, σ is a standard deviation of the noise n. Equation (6), the image f _i is at equal to the restored image f in equation (1), the difference image g _i and captured image g is based on the property that becomes noise n. This problem is solved using Lagrange's undetermined coefficient η

It is possible to replace the problem of minimizing the, it is possible to slide into updating the image f _i using an optimization method such as the steepest descent method or Newton's method. Non-Patent Document 1 describes the detailed technique.

  FIG. 5 is an example of a beam intensity waveform on the sample surface, which is one of the factors that degrade the resolution of the captured image. In FIG. 5A, the beams (1) and (2) in FIG. 5 (a) are beams 301 and 302 having a certain width, which are focused on the surface of the sample 303 and incident so that the beam width becomes narrow. Yes. The beam 301 of (1) is incident from a direction perpendicular to the x axis and the y axis. On the other hand, the beam 302 in FIGS. 5A and 5B is perpendicular to the y-axis, but is tilted at a beam tilt angle θ with respect to the x-axis. 5 (b) to 5 (e) show beam intensity waveforms on the sample surfaces (parallel to the xy plane) of the beams 301 and 302. If all the properties other than the incident direction of the beams 301 and 302 are the same, FIG. As shown in (c) and (e), the respective beam intensity waveforms in the y direction, that is, the waveform 312 in Fig. 5 (c) and the waveform 314 in Fig. 5 (e) can be regarded as the same. 5B is compared with the waveform 313 in FIG. 5D, the beam 302 that is incident from an oblique angle appears to have expanded, and the beam intensity waveform on the sample surface is expanded. Since the resolution is deteriorated as it is held, the resolution is usually degraded as the beam tilt angle θ is increased.

Even when other conditions are different, the beam intensity waveform is different. Examples of values representing the spread of the beam intensity waveform include diffraction aberrations d _d caused by the properties of charged particles as waves, chromatic aberrations d _c , which are lens characteristics, and spherical aberrations d _s . Each aberration d _d , d _c , d _s is expressed as follows.

Where λ is the de Broglie wavelength of the charged particle, α is the beam opening angle on the sample surface, Φ is the acceleration voltage of the charged particle beam, ΔΦ is the variation of the acceleration voltage Φ, C _c is the chromatic aberration coefficient, and C _s is the spherical aberration coefficient is there. FIG. 6A shows an example in which these aberrations change depending on imaging conditions. The beam opening angle α and the aberrations d _d , d _c when the values of the chromatic aberration coefficient C _c and the spherical aberration coefficient C _s are constant. The relationship between d _s and the acceleration voltage Φ and each aberration d _d , d _c and d _s are shown. In FIG. 6B, the reason why the diffraction aberration d _d is changed by the acceleration voltage Φ is that the de Broglie wavelength λ of the charged particles is proportional to the −1/2 power of the acceleration voltage Φ. The chromatic aberration coefficient C _c and the spherical aberration coefficient C _s are values representing the characteristics of the lens for focusing the charged particle beam, and usually change depending on the acceleration voltage Φ, the beam opening angle α, and the like. The beam intensity waveform can be approximately obtained using each aberration d _d , d _c , d _s , or can be accurately obtained by a calculation method described in Non-Patent Document 4, for example. Further, the beam intensity waveform may be estimated by actual measurement for each condition. Just as the aberration changes with the acceleration voltage Φ and the beam opening angle α, the beam intensity waveform also changes with these conditions. The beam intensity waveform also changes depending on the charged particle beam current (probe current), the focal position, and the like. Therefore, when the resolution degradation factor due to the beam intensity waveform is included in the degradation function, in order to perform image restoration appropriately, degradation suitable for the imaging conditions such as acceleration voltage Φ, beam opening angle α, probe current, focal position, etc. It is necessary to use a function.

FIG. 7 is an example of the step of calculating the beam intensity waveform according to the imaging conditions. When performing image restoration using the beam intensity waveform as a deterioration function, the step of calculating the beam intensity waveform of 501 can be used as the step of generating the deterioration function of 102 in FIG. Parameters necessary for calculating the beam intensity waveform, which change according to the imaging conditions, are stored in advance as 502 for a plurality of imaging conditions. The parameters 502 are, for example, the chromatic aberration coefficient C _c and the spherical aberration coefficient C _s . Next, a parameter corresponding to the imaging condition given by the switch 503 is selected. After selecting this parameter, the beam intensity waveform corresponding to the imaging condition is calculated by calculating the beam intensity waveform at 504. If the imaging condition parameter that completely matches the given imaging condition is not stored in 502, the switch 503 may select the parameter in the imaging condition closest to the given imaging condition, for example. Alternatively, estimation may be performed using parameters corresponding to a plurality of imaging conditions close to a given imaging condition.

  FIG. 8 is another example of the step of calculating the beam intensity waveform in accordance with the imaging conditions. Beam intensity waveforms for a plurality of imaging conditions are stored in advance as 602. Next, the beam intensity waveform corresponding to the imaging condition given by the switch 603 is selected. If an imaging condition parameter that completely matches the given imaging condition is not stored in 602, the switch 603 may select a beam intensity waveform under the imaging condition closest to the given imaging condition, for example. It is also possible to interpolate beam intensity waveforms under a plurality of imaging conditions close to a given imaging condition.

  In some cases, sample information is required to calculate the degradation function. FIG. 9 shows a state in which a charged particle beam is irradiated on a sample having a large difference in elevation. In FIG. 9A, the charged particle beam 802 is focused on the surface of the sample 803, and its beam intensity waveform is not so wide as 812. On the other hand, the charged particle beam 801 is out of focus on the surface of the sample 803, and its beam intensity waveform is wider than the beam intensity waveform 812 as indicated by 811 as shown in FIG. When the height difference of the target sample is large with respect to the focal depth of the charged particle beam as in the sample 803, the focus may be lost as in the charged particle beam 801. In this case, the beam diameter s of the charged particle beam varies depending on the beam irradiation position, for example, 821 shown in FIG. 9B. Due to this change, captured images having different resolutions depending on the position are acquired. For such an image, in order to appropriately reduce resolution degradation due to the beam intensity waveform, information on the shape of the sample, particularly the height and focal position of the sample, is used for each beam irradiation position (x, y). It is necessary to obtain a beam intensity waveform and use the beam intensity waveform as a deterioration function A. When using a deterioration function A that changes depending on the beam irradiation position (x, y), the deterioration function is expressed as A (x ′, y ′; x, y), and the captured image g (x , Y)

become that way. The height of the sample may be measured by, for example, the height measurement sensor 214 shown in FIG. 2, or the height of the sample measured by a scanning probe microscope or the like may be used. Further, when the sample is a semiconductor pattern, CAD (Computer Aided Design) data for pattern design can also be used.

  FIG. 9 shows an example in which the deterioration function changes depending on the height of the sample. As another example in which the deterioration function changes depending on the sample information, there is a case where diffusion of a charged particle beam in the sample is included in the deterioration function.

  FIG. 10 shows an example of the diffusion region of the charged particle beam in the sample. Diffusion occurs due to the interaction between the charged particles and the sample, and the interaction varies depending on the material of the sample. Therefore, the width of the diffusion region differs depending on the material of the sample. For example, even when the same charged particle beam 901 is irradiated, the diffusion region is narrow as 911 in the sample 921, whereas the diffusion region as 912 in the sample 922. Can be wide. The diffusion region for each sample can be calculated by a Monte Carlo simulator or the like that calculates the trajectory of charged particles in the sample. Also, the diffusion region differs depending on the shape of the sample near the beam irradiation position. For example, when there is a step like the sample 923, the diffusion region near the step is the diffusion region at the position where there is no step in the vicinity. It has a different shape.

  Therefore, in order to reduce the resolution degradation due to the diffusion of the charged particle beam as much as possible, it is necessary to calculate the degradation function based on the shape and material of the sample. The shape of the sample may be measured using, for example, the height measurement sensor 214 shown in FIG. 2, or may be measured using a scanning probe microscope or the like. Further, when the sample is a semiconductor pattern, it may be obtained from CAD data for pattern design. The material of the sample may be acquired by, for example, EDX (Energy Dispersion X-ray spectrum) for examining the composition of defects, or may be acquired from CAD data for pattern design when the sample is a semiconductor pattern.

  Another example of a resolution degradation factor is degradation related to beam scanning. In the scanning charged particle microscope, the charged particle beam emitted from the sample or transmitted through the sample is scanned while scanning the charged particle beam with respect to the sample 1903 in FIG. To detect. While the number of charged particles is detected for each pixel 1902 while changing the beam irradiation position, since scanning is performed during detection, a value smoothed in the x direction, which is the scan direction, is obtained as a result. Further, as the scan time per pixel is shorter than the response time of the detector, the smoothing is performed in the x direction. In order to reduce these resolution degradation factors, a degradation function having a width only in the x direction, such as 1911 and 1912, can be generated according to the beam scan speed, and image restoration can be performed using the degradation function.

  FIG. 12 is an example of steps for calculating a degradation function for reducing resolution degradation due to the beam intensity waveform and resolution degradation due to diffusion of the charged particle beam in the sample. First, parameters necessary for calculating a beam intensity waveform for a plurality of imaging conditions are stored in advance as 1001. Next, a parameter corresponding to the given imaging condition is selected by the switch 1002. Further, the distance between the focal position and the sample surface position is calculated in step 1003 from the imaging conditions and the sample shape. Next, the beam intensity waveform is calculated in step 1004 based on the distance and the parameter output from the switch 1002. Next, the diffusion region of the charged particle beam in the sample is calculated in step 1005 using the beam intensity waveform and imaging conditions, the shape of the sample, and the material of the sample. Finally, the beam intensity waveform and the diffusion region are calculated. The degradation function is obtained in step 1006. If it is not necessary to calculate the beam intensity waveform according to the shape of the sample, step 1003 may be omitted. Further, when the resolution reduction due to the beam intensity waveform is not reduced, Steps 1002 to 1004 may be omitted.

In order to finely set the deterioration function, it may be necessary to correct the deterioration function as in step 105 in FIG. 1 after the generation of the deterioration function. FIG. 13 is an example of a GUI screen that prompts the user to modify the degradation function. This GUI screen is a GUI screen for setting a beam intensity waveform and a degradation function for reducing resolution degradation due to diffusion of a charged particle beam in the sample. For a given imaging condition 1101, there are an area 1104 for displaying a degradation function, an area 1103 for displaying a default value of a parameter representing the degradation function, and an area 1107 for setting a parameter representing the degradation function. The parameter representing the degradation function may be, for example, values at several positions (x ′, y ′) of the degradation function A (x ′, y ′), or Fourier transform of the degradation function A (x ′, y ′). The values at several frequencies (f _x ', f _y ') of the function F _A (f _x ', f _y ') may be used. Alternatively, for example, it may be a parameter used for calculating a deterioration function such as a chromatic aberration coefficient C _c and a spherical aberration coefficient C _s . In the area 1104, only a deterioration function corresponding to the parameter set in the area 1107 may be displayed, or a deterioration function corresponding to the default value of the parameter displayed in the area 1103 may be displayed. An imaging condition may be set from the GUI screen of FIG.

  In addition, on the GUI screen in FIG. 13, a restored image obtained by performing image restoration using a degradation function corresponding to a captured image or a parameter set in the area 1107 as in the area 1106 is displayed. There is an area. Further, there may be an area for displaying values such as resolution, noise amount, pattern dimension, and the like that can be calculated from each image, such as an area 1108. In the areas 1106 and 1108, the result of image restoration using the degradation function corresponding to the default parameter as displayed in the area 1103 may be displayed. Further, there may be an area 1102 for setting sample information necessary for calculating the deterioration function and an area 1105 for setting processing parameters for image restoration. These display areas and setting areas may be displayed separately using a plurality of GUI screens.

  An example in which the principle of image restoration described above is applied to a defect review SEM will be described below.

  The restored image with high resolution or high S / N by image restoration can also be used for fine structure observation, high precision measurement, etc. FIG. 14 is an example of a sequence for performing defect detection or defect classification using a restored image obtained by image restoration in a semiconductor inspection SEM such as an SEM inspection apparatus or a defect review SEM. Steps 101 to 105 are the same as those described in FIG. In step 1301, the restored image 112 is subjected to defect detection or defect type classification. When defect detection is performed on a conventional captured image, even if a low-contrast defect or a minute defect such as the defect 1311 is included in the image, the captured image has low resolution or low S / N. If so, the defect may not be detected. On the other hand, by performing image restoration, it becomes possible to make a defect appear as in the case of the defect 1312. Therefore, it becomes easier to detect the defect by detecting the defect in step 1301. Even when defect classification is performed in step 1301, the classification accuracy can be improved for the same reason.

  As a specific example of step 1301, FIG. 15 shows an example of a sequence for automatically capturing and classifying defect images in the defect review SEM. When a defect image is automatically picked up by a defect review SEM, various optical types as described in, for example, JP-A-2-170279, JP-A-2000-105203, JP-A-2001-255278, etc. The defect detection and the defect position are measured using a defect inspection apparatus or an optical microscope included in a review SEM as described in JP-A-2006-261162, and the measured defect position information is used. Observe in detail with review SEM. The defect position has low position accuracy, and is often not accurate enough to directly capture a high-magnification image with a review SEM. In this case, first, using the position information of the defect detected by the optical defect inspection apparatus or the optical microscope of the review SEM, the table on which the sample is placed is driven in Step 1701 to move the defect into the observation field of the SEM.

  Next, an SEM low-magnification image is captured in step 1702, defect detection is performed using the low-magnification image captured in step 1704, and defect position coordinates sufficient to capture the high-magnification image are acquired. At this time, in order to improve the performance of defect detection, a restored image is generated by performing image restoration on the low-magnification image after low-magnification imaging as in step 1703, and defect detection is performed using the restored image. be able to.

  Also, when detecting a defect in a repetitive pattern portion such as a memory portion, a low-magnification image of a pattern that does not include a defect obtained by imaging is stored and used as a low-magnification reference image. By comparing the stored low-magnification reference image with the image containing the low-magnification defect obtained every time, and detecting the defect, the process of imaging the low-magnification reference image every time can be omitted. Although the detection efficiency can be improved, an image with a reduced noise component is used as a low-magnification reference image by performing the above-described image restoration on the low-magnification reference image stored. be able to. As a result, by obtaining a difference image between the low-magnification reference image and the image including the low-magnification defect restored as described above, the defect can be detected with higher accuracy.

  Next, when a defect is detected in the defect detection step 1704, the position coordinates on the SEM of the detected defect are registered, and a SEM high-magnification image is taken at the defect position coordinates registered in step 1705, In step 1707, the high-magnification image is used to classify into defect types. At this time, in order to perform defect classification in step 1707 with high performance, image restoration is performed on a high-magnification image obtained by high-magnification imaging as in step 1706, and the restored image obtained by the image restoration is used. Defect classification. In step 1708, it is determined whether or not all the target defects have been detected. If not, the processes from step 1701 to 1707 are repeated for each target defect. The image restoration process in either step 1703 or 1706 can be omitted. The restored image generated in step 1706 is stored in a database 1808. Similarly, the restored image generated in step 1703 can also be stored in the database 1808.

  According to this embodiment, it is possible to observe a finer defect using an image from which a noise component has been removed. In addition, since a clearer image of the defect can be obtained, the feature amount of the defect can be obtained more accurately, and the classification accuracy of the detected defect can be improved.

  Next, an embodiment in which the principle of image restoration described above is applied to a semiconductor measurement SEM such as a length measurement SEM will be described below.

  FIG. 16 shows an example of a sequence for measuring pattern dimensions and pattern shapes using a restored image in a semiconductor measurement SEM such as a length measurement SEM. The sequence of 101 to 105 is the same as that in FIG. In step 1201, the size or shape of the pattern included in the image is measured for the restored image 112. In a length measurement SEM or the like, this processing is performed at a point (evaluation point) where dimension measurement or shape measurement is desired. In a length measurement SEM, a captured image is usually generated by detecting secondary electrons. If an image is generated so that the lightness value increases as the number of secondary electron detections increases, an image in which a line-shaped region (white band) having a large lightness value at the pattern edge can be obtained. Pattern dimension measurement and shape measurement are performed using a white band. However, there is a problem that the lower the resolution, the lower the accuracy due to the wider white band. For this reason, it is possible to improve accuracy by performing pattern dimension measurement and shape measurement on the restored image, rather than performing similar measurement on the captured image.

  FIG. 17 shows an example of a sequence in which an evaluation point image is automatically captured in a semiconductor measurement SEM and the dimensions and shape are measured. The evaluation point image is often taken at a high magnification of 100,000 times or more, but the position accuracy is lower than the image area of the evaluation point image. In view of this, in the sequence shown in FIG. 17A, first, an image of an addressing point is captured in step 1801 in order to adjust the position accuracy, and in step 1802, the positional deviation amount is corrected using the captured image. Next, in step 1803, an image of the evaluation point is captured, and in step 1804, a restored image is generated by performing image restoration on the captured image. In step 1805, the restored image is used to generate a pattern included in the image. Dimension measurement and shape measurement. The processing in steps 1801 to 1805 is repeated for each target evaluation score. The restored image generated in step 1804 can be stored in a database as in 1807.

  On the other hand, in the sequence shown in FIG. 17 (a), the misalignment is corrected in step 1802 using the captured image obtained by the addressing point image capturing in step 1801 as it is, whereas in FIG. 17 (b). In the sequence shown, the image of the addressing point imaged in step 1801 is restored in step 1807 to generate a restored image, and in step 1808, the misregistration amount is corrected using the restored image. This sequence is particularly effective when the resolution or S / N of the captured image at the addressing point is low and it is difficult to correct the positional deviation amount.

  In addition, instead of the sequence of FIG. 17A, a restored image is generated for the captured image at the evaluation point in step 1812 as shown in FIG. 17C. In the dimension measurement and shape measurement in step 1811, the restoration is performed. It is also possible to use the captured image instead of the image. Further, as shown in FIG. 17D, after the captured images of the evaluation points are stored in the database 1821, a restored image can be generated for the captured images read from the database separately in step 1822.

  FIG. 18 is another example of a sequence for performing pattern dimension measurement and shape measurement on a restored image. Images are captured under different imaging conditions 1424 and 1427 to obtain the respective captured images 1421 and 1423, and the captured images 1421 are processed to match the resolutions of the captured images 1421 and 1423. When the resolution of the captured image 1421 is higher than the resolution of the captured image 1423, the resolution can be matched with the captured image 1423 by a process such as convolution of a low-pass filter with the captured image 1421. On the other hand, when the resolution of the captured image 1421 is low, it is possible to match the captured image 1423 by performing image restoration in step 1404 on the captured image 1421.

The captured images 1421 and 1423 may be captured by the same device or different devices. In sequence 1432, first, an image is captured in step 1408 based on a preset imaging condition 1427 to obtain a captured image g _{2 of} 1423. Next, in step 1409, pattern dimension measurement or shape measurement is performed. In step 1407, a degradation model corresponding to the resolution degradation factor of interest is calculated using the imaging conditions and sample information. Similarly, in sequence 1431, an image is captured in steps 1401 based on the imaging condition 1424 to obtain a captured image g _{1 of} 1421. In step 1402, a deterioration model is calculated from the imaging condition 1424 and the sample information. Thereafter, a degradation function A is generated in step 1403 from the degradation model obtained in steps 1402 and 1407. When the deterioration models obtained in Steps 1402 and 1407 are A _a and A _b , for example, the deterioration function A can be obtained using Expression (4). Next, the restored image f _{1 of} 1422 is obtained by performing image restoration in step 1404 using the degradation function A. If necessary, the deterioration function can be modified as in steps 1405 and 1406. In step 1406, correction is performed so that the resolution of the restored image f ₁ of 1422 matches the resolution of the captured image g ₂ of 1423. After the correction of the degradation function is completed, pattern dimension measurement or shape measurement is performed using the restored image f _{1 in} step 1410.

In the embodiment of FIG. 18, but the combined resolution of the restored image f ₁ on the resolution of the captured image g _2, performs image restoration even for captured image g _2, restore the resolution of the reconstructed image g ₂ obtained as a result of it may be matched to the resolution of the image g _1. In steps 1405 and 1406, the user may be asked whether or not the deterioration function needs to be corrected, or a correction method thereof, or may be automatically corrected based on a predesignated criterion. As a method of automatically correcting, for example, the following formula

If the sum of squares S of the difference between the restored image f ₁ and the captured image g ₂ is larger than a certain value, it is considered that the deterioration function needs to be corrected in step 1405, and in the subsequent step 1406, If the resolution of the restored image f ₁ is higher than the resolution of the captured image g ₂ , the width of the degradation function is narrowed. Conversely, if the resolution of the restored image f ₁ is high, the width of the degradation function is widened. be able to. In addition, the reference image for comparison with the restored image f ₁ and the deterioration model output in step 1407 do not need to use the actually obtained captured image or the deterioration model corresponding to the captured image. May use a simulation image. Further, instead of performing comparison with the reference image, for example, the resolution or the like may be calculated from the restored image f ₁ and compared with the reference resolution or the like.

  According to the present embodiment, an image of a pattern with reduced noise components can be obtained, and by using this image, more accurate pattern dimension measurement and pattern shape measurement can be performed.

  The resolution of the captured image may change as the state of the apparatus changes with time. Especially when it is necessary to obtain a stable resolution, such as for pattern dimension measurement and shape measurement, it is necessary to always grasp the state of the apparatus and apply feedback to the degradation function. FIG. 19 shows an example of a sequence for measuring the state of the apparatus. First, a sample sample such as 711 is prepared, and an image of the sample sample is captured in step 701 to obtain a captured image g 712. Next, in step 702, the state of the apparatus is measured. As the state of the apparatus, for example, the resolution of the captured image g, the pattern dimension, and the like are measured. The state of the apparatus can be regarded as one of the imaging conditions.

  FIG. 20 shows a GUI screen that prompts the user to correct the deterioration function A in the deterioration function correction that is the step 1406 in FIG. 18 includes an area 1601 for displaying the imaging condition 1424 and the sample information 1425, and an area 1602 for displaying the imaging condition 1427 and the sample information 1426. Similarly to FIG. 18, there are an area 1605 for displaying a deterioration function, an area 1603 for displaying a default value of a parameter representing the deterioration function, and an area 1604 for setting a parameter representing the deterioration function. Further, there may be an area 1606 for setting processing parameters for image restoration.

In addition, the GUI screen in FIG. 20 includes an area for displaying the captured image g _{1 of} 1421, the captured image g _{2 of} 1423, and the restored image f ₁ , as an area 1607. In addition to these images, may be displayed to allow easy be combined with captured image g ₂ restored image f _1, for example, the difference image or the like of the captured image g ₂ and the restored image f _1, such as 1611 . Further, similarly to the area 1108, there may be an area 1608 for displaying values such as resolution, noise amount, and pattern dimension that can be calculated from each image.

In the first and second embodiments, the case where the SEM is used as the scanning charged particle beam apparatus has been described. However, it is obvious that the same image processing method can be applied even when the SIM is used.
While the present invention has been described in detail based on the best mode for carrying out the invention, the present invention is not limited to the best mode for carrying out the invention, and departs from the gist thereof. It goes without saying that various changes can be made without departing from the scope.

DESCRIPTION OF SYMBOLS 111 ... Captured image, 112 ... Restored image, 201 ... SEM, 202 ... Electron gun, 203 ... Electron beam, 204 ... Condenser lens, 205 ... Objective lens, 206 ... Sample, 207 ... Stage, 208 ... Detector, 209 ... Image Generator 210 ... electron beam 211 ... deflection unit 212 ... correction unit 213 ... electrode 221 ... control unit 222 ... processing unit 223 ... storage unit 224 ... input / output unit 301 ... charged particle beam 302 ... charged particle beam, 411 ... captured image, 414 ... restored image

Claims (2)

A charged particle beam irradiation optical system means for irradiating and scanning a focused charged particle beam on a pattern-formed sample;
Charged particle detection optical system means for detecting the same kind or different kind of charged particles generated from the sample scanned by being irradiated with the charged particle beam by the charged particle beam irradiation optical system means;
Image acquisition means for processing a signal detected by the charged particle detection optical system means to obtain a charged particle image of the sample;
A scanning charged particle microscope apparatus comprising image processing means for processing a charged particle image of the sample obtained by the image obtaining means,
The image processing means is acquired by the image acquisition means using a degradation function calculated using image acquisition conditions including at least one of image acquisition conditions of the image acquired by the image acquisition means or information of the imaged sample. A scanning charged particle microscope apparatus characterized by obtaining a restored image of an image. A charged particle beam focused using a scanning charged particle microscope is irradiated onto a sample on which a pattern is formed and scanned to detect charged particles generated from the sample and acquire a charged particle image of the sample. And a method of processing the acquired charged particle image of the sample,
Scanning characterized in that a restored image of the acquired charged particle image is obtained using a deterioration function calculated using image acquisition conditions including at least one of image acquisition conditions of the acquired image or information of the imaged sample. For processing images acquired with a scanning charged particle microscope apparatus.

Images