JP5880134B2 - Pattern measuring method and pattern measuring apparatus - Google Patents

Description

  The present invention relates to a pattern measurement method and a measurement apparatus for accurately measuring the size and shape of a semiconductor device and a photomask used when the semiconductor device is manufactured by a lithography technique. More specifically, to measure the size and shape of a pattern by irradiating a pattern on a substrate such as a photomask while scanning with an accelerated electron beam and measuring the amount of secondary electrons emitted from the surface. The present invention relates to a pattern measuring method and a measuring apparatus therefor.

Semiconductor integrated circuits are being miniaturized and highly integrated in order to improve performance and productivity, and also with regard to lithography technology for forming circuit patterns, technological development for forming finer patterns with high accuracy Is underway. Along with this, a technique for measuring the dimension and shape of a pattern is also required with higher accuracy.
The pattern size of a semiconductor device or a photomask used for manufacturing the semiconductor device is less than 100 nm, and it is necessary to stably form a pattern of less than 25 nm as a next-generation device. As a method for measuring the dimension and shape of such a fine pattern at a specific position, a scanning electron microscope (hereinafter also abbreviated as “CD-SEM”) specially designed for dimension measurement is used. .

  The CD-SEM irradiates a pattern formed on the substrate while scanning the accelerated primary electrons emitted from the electron beam source in a two-dimensional manner by a deflector, and emits secondary electrons and backscattering emitted from the surface. Detect electrons. The scanned area is divided into vertical and horizontal directions, the divided area is defined as a pixel, the secondary electrons detected in each area, or the amount of backscattered electrons as the luminance of the pixel, the secondary electron image or the backscattered electron image (hereinafter referred to as the “electron”) , Also collectively referred to as “SEM image”).

The contrast of the SEM image is formed by the material of the object to be measured and the unevenness of the surface. The amount of secondary electron emission per primary electron is called secondary electron emission efficiency, and the contrast by the material of the SEM image is obtained by the fact that the secondary electron emission efficiency differs depending on the material. The secondary electron emission efficiency varies depending on the acceleration voltage of the primary electrons.
On the other hand, the contrast formed by surface irregularities is explained from the relationship between the penetration depth of primary electrons and the escape depth of secondary electrons. The probability that secondary electrons generated by inelastic scattering of electrons entering the solid surface from the surface or backscattered electrons whose traveling direction has changed due to elastic scattering escapes from the surface is the depth from the surface at the time of generation. It is known that it decreases exponentially with time. In general, the depth of secondary electrons that can escape from the surface is generally less than 5 nm, and the brightness of the SEM image is increased under the condition that a large amount of secondary electrons are generated in a region close to the surface.

  Conditions for generating a large amount of secondary electrons in the region close to the surface are, for example, conditions in which the primary electrons are inclined from the direction perpendicular to the surface (so-called gradient effect), and the surface is convex and close to the secondary electron generation position. There are many conditions (so-called edge effect). For this reason, in the case where a pattern is formed on a flat substrate such as a photomask, the brightness of the SEM image is increased at the end of the pattern.

The dimension measurement by the CD-SEM uses the phenomenon that the brightness of the SEM image is increased at the pattern edge as described above, and the interval between areas that appear white on the SEM image (hereinafter referred to as “white band”). The pattern dimensions are measured based on this.
Under conditions where the acceleration voltage is low as used in CD-SEM, the white band width is typically 10 to 15 nm. Accordingly, in order to accurately measure the pattern dimension, it is important which part of the white band is set as the pattern end. For example, a method of setting a relative value with respect to the peak of the brightness value of the white band as a threshold value, or obtaining the position of the inflection point of the original profile from the peak position obtained by differentiating the brightness profile, and setting it as the pattern end position. Etc. are used.

JP 2005-156436 A JP 2010-205864 A Japanese Patent Application No. 2011-39436

Yipping Zhao et al., “Characterization of Amorphous and Crystalline Rough Surface: Principles and Applications” Academic Press (2001)

  However, as the pattern becomes finer, the side wall of the actual pattern edge is not an ideal vertical shape, but as shown in FIG. 1, the ideal pattern cross-sectional shape (FIG. The actual pattern cross-sectional shape is as shown in FIG. For this reason, measurements including rounding of the upper surface end, deviation of the side wall angle from the vertical, and tailing of the pattern bottom have been required. Therefore, an SEM image is predicted by simulation based on the material and cross-sectional shape of the object to be measured, and a library composed of a large number of SEM image signals taking into account the shape variation is constructed. A method has been devised to obtain the dimensions of the object to be measured by matching. Such a technique is disclosed in Patent Document 1.

Another major cause of measurement error is that the object to be measured is charged due to the incidence and emission of electrons when the pattern dimension is measured by the CD-SEM. The influence of the charging of the measurement object on the measurement accuracy includes an incident position shift due to primary electron deflection, a change in secondary electron detection rate due to secondary electron deflection, and the like.
The variation of the SEM image due to the charging of the measurement object is very complicated. However, in recent years, the processing power of computers has greatly improved, and simulations incorporating many factors such as individual inelastic scattering and electron diffusion have become possible in addition to simple elastic scattering. For example, according to Patent Document 2, it is possible to generate an SEM image in consideration of the influence of charging from design data of a semiconductor pattern.

  In order to appropriately perform simulation calculations for various cross-sectional shapes using the above technique, it is necessary to optimize physical properties such as the composition, density, work function, and the like of the pattern forming material. For this purpose, the actual pattern cross-sectional shape is precisely measured with a transmission electron microscope or the like and collated with the simulation calculation results. However, there is a problem that a residual due to a variation factor that cannot be expressed only by a cross-sectional shape is large.

  Therefore, the present invention has been made in view of the background art, and a pattern measurement method and a pattern measurement apparatus capable of simulating a pattern shape more accurately according to actual conditions and obtaining a highly accurate measurement result. It is an issue to provide.

The pattern measurement technique according to the present invention extracts the roughness of the pattern edge and its spatial frequency component from the contour line of the pattern obtained from the secondary electron image in order to accurately measure the pattern dimension using the charged particle beam. The correction is made by collating with a simulation data library taking these into consideration.
That is, the pattern measurement method according to one aspect of the present invention irradiates electrons on the surface of an object to be measured, detects the relationship between the position where the electrons are incident and the amount of electrons emitted from the surface, and information on the shape of the surface And measuring the dimension of the pattern formed on the surface of the object to be measured, wherein as the information on the shape of the surface, a parameter characterizing the roughness of the pattern edge is extracted, and the parameter is The method includes a procedure for determining a dimension of a pattern by collating with a simulation result taken into consideration.

In the pattern measurement method according to one aspect of the present invention, it is preferable that the object to be measured is a photomask, and the parameters characterizing the roughness of the pattern end include a root mean square roughness, a correlation length, and It is preferable to include a roughness index.
In addition, a pattern measurement apparatus according to one embodiment of the present invention includes a measurement mechanism that irradiates a focused charged particle beam while scanning the object to be measured and measures electrons emitted from the object to be measured; A processing device that constructs a two-dimensional image from the luminance information using the value correlated with the scanned position and the measured number of electrons as luminance, and calculates contour information of the pattern from the luminance information, And a pattern dimension determining device for extracting a pattern dimension by correcting the contour information of the pattern using a simulation result considering the parameter.

  According to the present invention, parameters related to roughness are extracted from pattern edge information, and matching with a simulation library that takes this into account is performed. Can be reduced. Therefore, the pattern dimension can be measured with high accuracy.

It is explanatory drawing of what modeled an ideal pattern ((a)) and an actual cross section ((b)). It is explanatory drawing showing the flowchart of the pattern dimension determination process performed as a measuring method of one Embodiment of this invention. It is explanatory drawing showing the relationship between a correlation distance and roughness, (a) and (b) are different correlation distances, (a) is an example with a longer correlation distance than (b). It is explanatory drawing which compared distribution of deviation | shift amount from the average value of an edge part position.

Hereinafter, an embodiment of a pattern measuring method and a pattern measuring apparatus according to the present invention will be described with reference to the drawings as appropriate.
A pattern measurement apparatus according to the present invention is a CD-SEM that is a measurement mechanism that measures the electrons emitted from the measurement object by irradiating the measurement object with a focused charged particle beam (not shown) while scanning the measurement object. A processing device that constructs a two-dimensional image from the luminance information with a value correlated with the position scanned on the object to be measured and the measured number of electrons as luminance, and calculates contour information of the pattern from the luminance information, A pattern dimension determining device that extracts a parameter characterizing the roughness of the pattern edge and corrects the contour information of the pattern using the parameter to calculate a pattern dimension;

Specifically, the pattern measurement apparatus executes a pattern dimension determination process shown in FIG. 2 as a measurement method according to an embodiment of the present invention.
When the pattern dimension determination process is executed, as shown in FIG. 2, first, the measurement position of the pattern formed on the measurement object is specified based on the pattern design data (step S1). Next, an optimum measurement condition is retrieved from the database from the pattern shape at the designated measurement position and the pattern constituent material designated separately. The CD-SEM irradiates the object to be measured while scanning a focused charged particle beam, and measures electrons emitted from the object to be measured.

  The acquired SEM images are transferred to the processing device that controls the CD-SEM, and the processing device acquires a plurality of SEM images including the designated position from the CD-SEM (step S2). This SEM image may be obtained by scanning the number of pixels in the XY directions orthogonal to each other once, or may be obtained by integrating the results of scanning a plurality of times. In general, in order to reduce the effect of charging, it is often preferable to reduce the primary electron current value and scan at high speed. To obtain a good image with less noise, the result of multiple scans It is preferable that it is obtained by integrating | accumulating.

  In the following description, for convenience, it is assumed that scanning of the electron beam in the X direction is repeated by the number of pixels in the Y direction. The selection order of the pixels in the Y direction when obtaining one image can be selected, and there is no limitation on the order of integration. The scanning direction in the X direction may be only one side, or data may be acquired by reciprocating. In the following description, X and Y can be appropriately read according to the scanning method.

  Ideally, the brightness profile of the SEM image should be the same even if the scanning order of the electron beams is changed in this way. However, it is necessary for the incidence of electrons into the object to be measured and the emission of secondary electrons and backscattered electrons. When charging occurs, the SEM image also changes depending on the scanning order of the electron beams. By performing a plurality of simulations in consideration of such changes and creating a library, matching between the actual measurement image and the simulation library image can be performed with high accuracy.

  Next, the pattern dimension determining apparatus acquires the pattern end position by extracting the contour line of the SEM image from the designated pattern shape (step S3). A known technique can be used for the contour line extraction algorithm. The brightness change curve in the direction perpendicular to the contour line is interpolated with the brightness of each pixel, and the pattern edge obtained from the threshold value or maximum slope point is obtained. This can be done by stitching together.

  Next, the pattern dimension determining apparatus obtains roughness information of the pattern edge from the contour line information (step S4). In the case of a linear pattern, there are LER (Line Edge Roughness) based on the roughness of one end and LWR (Line Width Roughness) based on the roughness of both ends. The pattern measurement method of the present invention can be applied to both. Hereinafter, an example in which the pattern dimension determining apparatus executes the LER processing of the line pattern in the Y direction will be described. However, X and Y may be interchanged, and an oblique pattern can be easily expanded. .

  Hereinafter, the region for measuring the line width is abbreviated as FOV (Field of View). In the FOV, the luminance change in the X direction is acquired by the number of pixels in the Y direction. Next, the edge position of the pattern is acquired from the luminance value in the X direction for each Y value. The method for obtaining the end position may be determined by a relative threshold value with respect to the minimum luminance and the maximum luminance of the white band, or may be obtained from the peak position of the differential waveform obtained by taking a difference with a predetermined width. . In this case, a resolution smaller than the pixel size can be obtained by approximating a discrete value with a smooth curve by an interpolation method. In addition, the influence of random noise can be reduced by accumulating a predetermined width in the Y direction and averaging, but the resolution of the spatial frequency is lowered, so it is necessary to select an appropriate value. .

  Next, the end position is obtained by connecting the end positions in the Y direction. This may be obtained from one image by two-dimensional scanning once, or a longer end shape can be obtained by connecting a plurality of images. As the observation magnification is increased, roughness information having a high spatial frequency can be obtained with high sensitivity. However, since the observation area is reduced, it is difficult to obtain roughness information having a low spatial frequency. In such a case, it is an effective means to increase the effective value of the number of pixels by connecting a plurality of images.

  In step S4, when obtaining information on roughness from the edge information, first, an average position of the edge in the X direction is obtained within a range selected in the Y direction. When it is considered that the pattern end is a straight line parallel to the Y-axis, the average value may be obtained simply by selecting the end position in the X direction by the number of pixels in the Y direction. Further, even when it is regarded as a straight line inclined with respect to the Y axis, the straight line can be uniquely determined by the least square method based on the end position information for the number of pixels in the Y direction.

Alternatively, the best matching position may be obtained by superimposing it with ideal pattern edge position information calculated from the design data.
Next, the shift amount of the end position from the average position is acquired for each Y value in the FOV. The edge of the pattern when the number of pixels in the Y direction of the FOV is n, where x0 is the average X position in each Y value, and xi is the position in the X direction of the i-th pixel in the Y direction. The root mean square roughness Rq can be defined by Equation 1.

  Subsequently, the spatial frequency component in the Y direction of the edge roughness is analyzed. When (xi−x0) is expressed by g (i), the autocorrelation function C (τ) of the end position can be defined by Equation 2. Here, τ is a shift amount when calculating the correlation function. <> In Equation 2 represents an ensemble average.

From Equation 2, when τ = 0, that is, C (0) is the root mean square roughness. When there is no periodicity in roughness, the autocorrelation function gradually decreases as the shift amount τ increases. A shift amount that becomes C (0) / e (e is the base of natural logarithm) is called a correlation distance, and is represented by a symbol ξ here.
The relationship between the correlation distance ξ and the roughness is shown in FIG. Here, in FIG. 3, the Y-direction pixel number (horizontal axis) indicates an arbitrary point of the acquired SEM image as 0 (zero) and indicates 200 pixels therefrom. In addition, the unit of the end position (vertical axis) in the figure depends on the shape of the target pattern, but a unit of length (μm, nm, etc.) can be used.

  The curves in the two figures (a) and (b) have the same mean square roughness and different correlation distances, and (a) has a longer correlation distance than (b). As can be seen from the figure, even if the root mean square roughness is the same, the roughness characteristics differ greatly if the correlation distance is different. If the correlation distances are different, even if the root mean square roughness is the same, the area of the actual cross section is different, so that the amount of secondary electrons emitted from the pattern sidewalls is different. It is also conceivable that the secondary electrons emitted from the side wall are likely to be incident again on the side wall at a high spatial frequency.

  Another parameter that characterizes the roughness is called a roughness index, which is represented by the symbol α here. This represents the change of the roughness with respect to the spatial frequency, and corresponds to the slope of the graph plotted against the frequency by Fourier transform of the autocorrelation function.

  Such roughness is known to have self-affinity, and there is a relationship of D = α + 1 between the roughness index and the fractal dimension D. When the contour line on a two-dimensional plane such as an SEM image has self-affinity, α takes a value from 0 to 1. When α = 0, it is one-dimensional and can be regarded as a straight line. When α is 1, the fractal dimension is two-dimensional and corresponds to, for example, a mathematically defined Peano curve. Since the outline of the pattern is generally linear, in general, the value of α is relatively close to 0 rather than 1. In this way, the characteristics of roughness can be expressed numerically by the above three parameters, that is, the root mean square roughness R, the correlation distance ξ, and the roughness index α.

Next, using these three parameters, a library of SEM image signals is constructed and a search is executed (step S5). That is, in the process of step S5, a virtual shape having typical roughness is constructed using the three parameters R, ξ, and α.
For this purpose, a convolution method, a method using a Langevin equation, a Monte Carlo method, and the like are known. For example, a method described in Non-Patent Document 1 can be used.

  Next, based on the cross-sectional shape and material, and the electron beam irradiation conditions, the position of the cross section to be measured, the secondary electron emission amount, and the emission angle distribution are obtained. As this method, a simulation technique using, for example, a Monte Carlo method, which is used during the implementation of the method shown in Patent Document 2 or the pattern measuring method described in Patent Document 3 (unpublished) by the present inventor, is used. Can do.

Subsequently, a three-dimensional shape is constructed from the cross-sectional shape and the virtual shape, and secondary electrons emitted from each grid obtained by dividing the surface of the three-dimensional shape from the electron emission amount and the emission angle distribution, Count what can be considered to reach the detector. Specifically, tracks of electrons emitted from the surface are traced according to the respective distributions.
In a simple manner, the SEM image signal can be obtained by convolution according to R representing the magnitude of roughness for the SEM image signal by simulation from the cross-sectional shape. However, since the root mean square roughness R does not include frequency components, even if R is the same, the function used for convolution is not uniquely determined.

  FIG. 4 shows the distribution of the X coordinate position of the end position of the line pattern obtained from the actual SEM image over 1024 pixels in the Y direction. The root mean square roughness R determined from this distribution was 4.9 nm. If fitting is performed so that the distribution of the X coordinate position follows the normal distribution and is closest to the actual measurement value, the normal distribution graph shown in FIG. 4 is obtained. It can be seen that the actual distribution is slightly different from the normal distribution.

  Assuming that the pattern end position is an ideal sine curve, the end position distribution when the mean square roughness is 4.9 nm is the distribution shown as the sin function in FIG. Further, assuming that the pattern end position is composed only of white noise that has no correlation with the Y value, the distribution is not a curve having a peak as shown in FIG. 4 but a uniform distribution. Thus, in order to match the actual distribution, it is necessary to appropriately combine various frequency components.

  In order to appropriately extract such a frequency component, it is effective to use a power spectral density function obtained by performing Fourier transform on the correlation function C (τ) shown in Equation 2. As described above, it is shown in Non-Patent Document 1 that the power spectral density function is expressed by Equation 3 below when the end position has self-affinity.

  In Equation 3, k is a frequency, α is a roughness index, w is a root mean square roughness, and ξ is a correlation distance. Using this equation, the parameters can be obtained by fitting with a power spectral density function obtained from an actual SEM image.

  Next, matching is performed with the SEM image luminance profile synthesized based on the edge position that is artificially generated by the above method from the combination of these parameters (step S6 in FIG. 2). In addition, by performing convolution with a distribution function that uses the beam diameter of the primary electron of the SEM as a parameter to the SEM image brightness profile before matching, it is possible to perform more accurate matching. It is.

In this manner, after extracting the optimum cross-sectional shape from the library, the pattern end position can be determined by performing measurement based on the specifications (step S6). Therefore, the distance between two points on the SEM image can be measured with high accuracy. The measurement result including the determined end position and the like is output to the output device (step S7).
The embodiment of the present invention will be further described with reference to examples.

A photomask blank in which a film containing molybdenum silicide as a main component was formed to a thickness of 60 nm on a synthetic quartz glass substrate was prepared.
A chemically amplified electron beam resist was applied and dried on the molybdenum silicide film, exposed using an electron beam exposure apparatus, and subsequently baked and developed to form a resist pattern. Using this resist pattern as an etching mask, molybdenum silicide was etched using a dry etching apparatus, and the remaining resist was removed to obtain a photomask.

In order to measure the width of the line-shaped pattern formed on this photomask, an SEM image was acquired using a CD-SEM. The image has lines in the Y direction, and scanning of the electron beam when acquiring the image is in the X direction. An SEM image composed of 1024 pixels in the X and Y directions was obtained by integrating 64 scan images.
A process of forming a luminance profile in the X direction from the acquired SEM image, and calculating the X coordinate of the pattern end position from the maximum inclination point of the white band by smoothing and difference processing is repeated 1024 times in the Y direction, and each Y position A numerical sequence of end positions at was obtained. This was a numerical sequence in which the deviation from the average position was obtained by linear approximation by the least square method. From this deviation, a value of 4.9 nm was obtained as the root mean square roughness.

Next, an autocorrelation function C (τ) was obtained from this numerical sequence using Equation 2. This was spline interpolated to obtain a correlation distance ξ that is 1 / e of the value of C (0), and a value of 11.9 nm was obtained. In addition, the power correlation density function was obtained by Fourier transforming the autocorrelation function. From this slope, a value of 0.4 was obtained as the roughness index α.
From the combination of these numerical values, the relationship between the luminance profile and the end position of the actual pattern was obtained from the library calculated in advance by the above method. The same procedure was performed for the other end, and the dimension value of the linear pattern was obtained from the distance between the two points.

  By using the pattern measurement method and the pattern measurement apparatus of the present invention, when measuring a pattern with a CD-SEM, it becomes possible to perform accurate dimension measurement taking into account that the luminance profile changes depending on the roughness of the pattern. Can be manufactured with high accuracy.

DESCRIPTION OF SYMBOLS 10 ... Board | substrate 20 ... Pattern 21 with an ideal cross-sectional shape ... An example of an actual pattern

Claims (4)

Formed on the surface of the object to be measured by irradiating the surface of the object to be measured, detecting the relationship between the electron incident position and the amount of electrons emitted from the surface, and extracting information on the surface shape. A method for measuring the dimensions of a pattern,
A pattern measurement method comprising a step of extracting a parameter characterizing the roughness of a pattern edge as information relating to the shape of the surface and determining a dimension of the pattern by collating with a simulation result considering the parameter .   The pattern measurement method according to claim 1, wherein the object to be measured is a photomask.   The pattern measuring method according to claim 1, wherein the parameters characterizing the roughness of the pattern edge include a root mean square roughness, a correlation length, and a roughness index. A measurement mechanism that measures the electrons emitted from the object to be measured by irradiating the object with a focused charged particle beam while scanning it, and a value that correlates with the number of electrons measured and the position scanned on the object. A processing device that constructs a two-dimensional image from the luminance information as luminance, and calculates contour information of the pattern from the luminance information, extracts a parameter characterizing the roughness of the pattern edge, and performs simulation in consideration of the parameter A pattern measurement apparatus comprising: a pattern dimension determining apparatus that corrects contour information of the pattern using a result and calculates a pattern dimension.

Images