JP3104804B2 - Pattern size measurement method using charged beam - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a semiconductor device such as a VLSI,
In a length measuring device or an inspection device that measures the line width of an insulator pattern by irradiating a charged beam and detecting reflected electrons or secondary electrons, or the reproducibility of measurement due to the influence of defocus, noise, etc. The present invention relates to a pattern dimension measuring method using a charged beam for removing a drop and performing accurate measurement, and particularly to a method of determining waveform analysis conditions.

[0002]

2. Description of the Related Art Along with miniaturization of a pattern size of an LSI or the like, a dimension measuring device using a charged beam has been used. In this type of device, the electron beam is scanned in a direction perpendicular to the measurement pattern, and a reflected electron or a secondary electron is detected in synchronization with the scanning signal, thereby forming a secondary electron signal waveform in a direction perpendicular to the measurement pattern. Have gained. A method and an apparatus for obtaining a pattern dimension from this signal waveform include an inspection apparatus (JP-A-55-72807) for measuring the distance between edges by setting an appropriate slice level in the secondary electron signal waveform, and A method of linearly approximating each edge portion and obtaining a pattern size from the distance between the intersection points (Japanese Patent Laid-Open No. 61-80)
011) etc. are used. However, when performing such a measurement, irradiation conditions such as beam acceleration voltage and beam current, conditions for removing noise in a signal waveform, and analysis conditions for measuring a signal waveform are not optimized. In this case, the charge reproducibility occurs, the S / N ratio is poor, the contrast cannot be obtained, the margin for defocusing is small, and the reproducibility of measurement decreases. In optimizing such irradiation conditions and signal waveform analysis conditions, in the manufacturing process development stage, there are many sample types before the selection of process technology, and it is necessary to frequently determine irradiation conditions and signal waveform analysis conditions. Simple determination of the conditions is required. Also, at the production line monitoring stage, although the frequency of condition determination is low, it is necessary to grasp the dimensional drift during line operation for a large number of samples, so that measurement accuracy and automation must be advanced,
Since the required measurement accuracy is high, the accuracy of the condition setting is important.

In order to reduce the effect of charge-up,
The method of optimizing irradiation conditions when capturing signal waveforms is
This is disclosed in JP-A-2-159508. In this method, the waveform is captured by optimizing the acceleration voltage and the beam current so that the measurement reproducibility does not decrease due to the distortion of the secondary electron signal waveform. However, even when trying to optimize the irradiation conditions and obtain a signal waveform with good reproducibility,
Changes in the signal waveform due to noise and defocus are inevitable. For example, as a method of removing noise, there is a smoothing of a waveform in which each signal amount is replaced with an average value (simple average or weighted average) of several points in the captured signal waveform.

[0004]

FIG. 5 is a diagram showing a change in the signal waveform when the number of points used for the smoothing is increased, and FIG. 5A shows the original waveform. When the points increase from (a) to (d) and the number of points used for smoothing is small,
Since the effect of noise removal is small as in (b), spike-like noise remains, and a different edge due to noise may be erroneously measured as a pattern edge. On the other hand, if the number of points used for smoothing is too large, the signal waveform becomes dull as shown in (d), the edges of the pattern disappear, and accurate measurement cannot be performed.

On the other hand, FIG. 6 shows a case where the focal length is shifted.
It is a figure which shows the change of a secondary electron signal waveform, (b) is a signal waveform at the time of good focus. In the signal waveform in the case (b) where the focus is good, the rising of the pattern edge portion is steep. (A) and (c) show the case where the focus is slightly shifted up and down, and the rising of the edge portion becomes gentle. (D) is a case where the position is further shifted. When the focus is shifted, the slope of the waveform at the pattern edge portion becomes gentle, and the maximum and minimum positions of the signal are shifted. Along with this,
Pattern dimensions change.

FIG. 7 is a graph showing the measured size dependence. When the slice level is set high, the size is small with respect to defocus, but when the slice level is set low, the size is small with respect to defocus. Since the size is large, the accuracy and reproducibility of the size measurement are reduced in any case.

[0007] Since optimization of such waveform analysis conditions (determination of conditions) has been performed by trial and error depending on the intuition of an experienced person, when trying to measure a pattern of a new material, the conditions must be randomized. In addition, the work involved in searching for good analysis conditions required a lot of labor.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to provide a method for measuring a pattern dimension using a charged beam, in which the S / N ratio is lowered and the measurement accuracy is reduced due to defocus. It is an object of the present invention to provide a technology capable of preventing a decrease in the temperature.

Another object of the present invention is to provide a technique capable of efficiently determining waveform analysis conditions when measuring a new sample having no measurement experience.

The above and other objects and novel features of the present invention will become apparent from the description of the present specification and the accompanying drawings.

[0011]

Means for Solving the Problems In order to achieve the above object, the means of (1) of the present invention relates to a method of measuring a pattern dimension using a charged beam, wherein a signal waveform of the same pattern is fetched a plurality of times. While sequentially changing the smoothing parameter for removing noise and the pattern detection parameter for detecting the pattern position, the left and right edge positions of the pattern of each signal waveform are detected, and the variation in the pattern position between each waveform is small, and the parameter The combination of the parameters is determined so that the change in the edge position with respect to the variation is minimized, and thereafter, the focal length is sequentially changed, and the signal waveform of the same pattern is repeatedly acquired and stored. By changing the waveform analysis parameters, a parameter that minimizes dimensional variation is determined from the result of dimensional measurement of the stored waveform. And wherein the door.

The means of (2) is a method for measuring a pattern dimension using a charged beam, wherein a signal waveform is taken in, a smoothing parameter for removing noise of the waveform is fixed, and a pattern detection parameter is sequentially changed while a target pattern is changed. The left and right edge positions are detected to determine the area of the pattern detection parameter in which the pattern can be detected. Next, while sequentially changing the smoothing parameter, the above processing is repeated, and the area of the pattern detection parameter in which the pattern can be detected becomes The maximum smoothing parameter is determined as the optimal smoothing parameter, and the center value of the pattern detection parameter area in which the pattern can be detected at that time is determined as the optimal pattern detection parameter. By repeatedly acquiring and storing the waveform, changing the waveform analysis parameters for dimension measurement, From the results, sizing the storage waveform, and determines a parameter dimensional variation is minimized.

[0013]

According to the above-mentioned means, the parameters of the waveform analysis are determined so that the influence of the noise and defocus of the captured secondary electron waveform is minimized. First, smoothing parameters for removing noise are determined. For this reason, a plurality (about 10) of secondary electron waveforms of the same pattern are captured and stored. The smoothing and edge detection are sequentially performed on the stored waveform while changing the smoothing parameter, and the range of the smoothing parameter and the pattern edge detection parameter that can accurately detect the position of the pattern edge is determined.

Next, in order to minimize variations in dimension measurement due to defocus, a plurality of secondary electron waveforms (about 10) obtained by changing the focal length are stored. While changing the slice level for calculating the pattern size, measurements are sequentially performed on the stored waveforms, the slice level that minimizes the variation in the measured size is determined, and the waveform analysis conditions are optimized. By doing so, even if there is a change in the S / N ratio of the secondary electron signal waveform or a change due to defocus, the analysis condition of the waveform is determined so as to minimize the variation in the measured dimensions. Measurement with good measurement reproducibility is possible even for a waveform with a poor ratio or a sample that is difficult to focus.

The determination of these waveform analysis conditions is performed by trial and error based on the intuition of an experienced person because the following processing can be automatically executed if the secondary electron signal waveform is stored in advance. It can be performed more accurately and efficiently than the conventional method.

[0016]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1A and FIG. 1B are flow charts showing a procedure of a method for determining a waveform analysis condition in a pattern dimension measuring method using a charged beam according to one embodiment of the present invention. The method of determining the waveform analysis conditions according to the present embodiment is roughly classified into two, namely, a smoothing parameter / pattern detection parameter determination routine and a length measurement parameter determination routine.

Step 101: The stage is moved to a target pattern position and irradiation conditions are set.

<Smoothing Parameter / Pattern Detection Parameter Determination Routine: A> Step 102: In a focused state, a plurality of secondary electron signal waveforms of the same pattern and the same place are continuously acquired and stored (signals are repeated). Acquisition addition or averaging. The averaging is performed before storage). Step 103: One of the stored waveforms is sequentially read. Step 104: A smoothing parameter (described later) for removing noise and a pattern detection parameter for pattern detection are sequentially set. In this parameter setting, a predetermined range may be sequentially changed in a matrix, or a range to be changed may be determined from a pattern detection result using a previous parameter. For example, the smoothing parameter may be fixed, the pattern detection parameter may be sequentially changed, and the next smoothing parameter may be changed when the pattern cannot be detected. As for the order of change, changing the smoothing parameter can perform waveform smoothing in common, thereby reducing the time. Step 105: Using the parameters set in step 104, for the waveform read out in step 103,
Performs waveform smoothing and pattern detection. The left and right edge positions and parameters of the detected target pattern are stored.

This process returns to step 104 and is repeated by changing the smoothing parameter and the pattern detection parameter. Then, returning to Step 103, Step 10
2. Read another waveform stored in step 2
Change the pattern detection parameter and repeat. At this time, the combination of the smoothing parameter and the pattern detection parameter is the same as that of the first waveform. Step 106: For each combination of the parameters set in step 104, the edge position obtained in step 105 is statistically processed for each waveform to obtain an average value and variation.

From the above results, a combination of parameters in which the variation of the pattern edge position in each waveform is minimum and the average value (pattern position) does not change with the parameter change is determined as the optimal parameter.

<Length Measurement Parameter Determination Routine: B> Step 107: Set the focal length. The range of change is
The focal length is changed before and after the most focused state so that the pattern is not blurred and the edges are not easily recognized. Step 108: Capture and store the secondary electron signal waveform. Returning to step 107, the focal length is changed to capture and store the waveform. This process is repeated. As the number of repetitions increases, the accuracy of parameter determination can be increased, but it takes time. As a rough guide, 10
It is about. Step 109: One of the stored waveforms in the step 110 is sequentially read. Step 110: Set parameters (slice level) for pattern dimension measurement. Step 111: the parameters of step 110, step 1
The pattern size is measured using the waveform 09. The parameters and measurement results (dimensions) are stored. In this process, the dimension measurement is repeated by sequentially changing the pattern dimension measurement parameters. Next, returning to step 109, the next waveform is read out, and the measurement of the pattern dimension measurement parameters is changed to repeat the dimension measurement. At this time, the same parameters as used for the first waveform are used. Step 112: For each parameter of the pattern dimension measurement set in step 110, the edge position obtained in step 111 is statistically processed for each waveform to obtain a variation. The parameter that minimizes this variation is determined as the optimal parameter.

Hereinafter, detailed examples of the above items will be described.

<Routine for Determining Smoothing Parameter and Pattern Detection Parameter> As a smoothing parameter for removing noise, for example, a moving average in which the average value (sometimes weighted) of several points in proximity is used to replace the value. There are points, weighting values, and the like. If the score is small, spike-like noise remains because the effect of noise removal is small, and a measurement error occurs during dimension measurement. Conversely, if the score is too large, the waveform becomes dull and the pattern edge portion becomes inaccurate (see FIG. 5).

On the other hand, as a method of detecting a pattern edge position, an appropriate slice level is set in the obtained signal waveform, an intersection thereof is obtained, and this intersection position is sequentially checked as a pattern edge position candidate ( (A binarization method), a method of setting a slice level with a differentiated waveform in order to detect a place where the slope of the signal waveform is large, and sequentially checking as a pattern edge position candidate (differential method). If the slice level is too high in the binarization method, only one pattern may be detected when there are a plurality of patterns. On the other hand, if the slice level is too low, a lot of noise will be picked up, making it impossible to detect an accurate pattern edge. In the case of a waveform that is not symmetrical in the left and right, if the slice level is inappropriate, an edge may be detected on only one side. In the case of the differentiation method, since the same occurs, it is necessary to select an appropriate slice level.

FIG. 2 is a diagram showing an example of determining a smoothing parameter and a pattern detection parameter according to the present embodiment, wherein 1 is a pattern detectable area, 2 is the left-right asymmetry of the waveform, and the target pattern is affected by other patterns. Is a region where noise cannot be removed, 3 is a region where noise cannot be removed, 4 is a region where spike noise is erroneously recognized as a pattern edge, and 5 is a region where the waveform is dull and there is no edge.

In this embodiment, as shown in FIG.
The smoothing parameter for removing the noise and the slice level for detecting the pattern edge position enable accurate detection of the pattern position within a certain area. When the signal waveform is not symmetrical in the left and right directions, the slice levels for pattern detection may be set to different values at the left and right pattern edges.
In this case, in order to determine the slice level for detecting the left pattern edge and the slice level for detecting the right pattern edge, areas where edge positions can be detected as shown in FIG. 2 are obtained.

Note that this area is not always clearly drawn. However, the optimal area can be found using the following guidelines. If the edge detection is not successful due to the influence of spike noise (smoothing is insufficient or the slice level for pattern detection is low), a random shift occurs. In the case of obtaining from a plurality of waveforms, the variation of the edge position obtained from each waveform becomes large. In this case, the smoothing parameter may be increased, or the slice level for pattern detection may be increased. On the other hand, if the smoothing parameter is too large, the influence of noise is eliminated and the variation is reduced, but the waveform is dull, and the edge position shifts with an increase in the smoothing parameter. In such a case, the smoothing parameter may be reduced. If the slice level for pattern detection is too high, either the left or right edge cannot be detected, or an edge of a completely different pattern will be detected.

Since the area in which such a pattern can be detected varies depending on the obtained signal waveform, it is necessary to determine the condition depending on the material and shape of the pattern to be measured. In this case, setting the parameters in the common portion of the detectable area when a plurality of the same patterns are measured is more effective in improving the reproducibility of the measurement, since the influence is also affected by the reproducibility of the obtained waveform. In this waveform acquisition, the waveform may not be exactly the same, but may be shifted slightly, or waveforms of different locations having the same pattern may be used. Although the parameters are set using a plurality of waveforms, a detectable area as shown in FIG. 2 may be obtained with one waveform, and the parameter may be set to a value at the center of the area. . If the approximate parameters are obtained in this way, and the parameters are changed in the vicinity of the approximate parameters to detect patterns for other waveforms, the time for determining the parameters can be reduced. The procedure for implementing the method will be described with reference to FIG. 3 (a flowchart according to another embodiment of the present invention).

Step 201: The stage is moved to the target pattern position and irradiation conditions are set.

<Smoothing Parameter / Pattern Detection Parameter Determination Routine: A> Step 202: While in focus, a secondary electron signal waveform is fetched. Step 203: While fixing the smoothing parameter for removing the noise and sequentially changing the pattern detection parameter,
The left and right edge positions of the target pattern are detected and stored. Step 204: Determine the area of the pattern detection parameter in which the pattern can be detected. [If the slice level of the pattern detection is not appropriate, either the left or right edge cannot be detected.
Since the spike noise or the edge of a completely different pattern is erroneously detected, the edge position is suddenly changed by increasing the slice level. Step 205: The processing of steps 203 and 204 is performed by changing the smoothing parameter. Step 206: From step 205, the smoothing parameter that maximizes the area of the pattern detection parameter in which the pattern can be detected is set as the optimal smoothing parameter, and the center value of the pattern detection parameter area in which the pattern can be detected at that time is detected as the optimal pattern. Determine as a parameter.

<Measurement Parameter Determination Routine: B> As a method of length measurement, a method of measuring by setting a slice level, a method of approximating a signal waveform of a pattern edge portion and a signal waveform of a background portion by different functions, respectively, and starting from an intersection thereof A function approximation method for obtaining dimensions is used. In the former, the slice level is used, and in the latter, the extent of the waveform approximated by what function is used as a measurement parameter.

FIG. 7 shows the focal length dependency of the measured value when the slice level is used. When the focus shifts, the shift amount (variation) of the measurement dimension changes depending on the value of the slice level. For example, in FIG. 7, when the slice level is 30 to 40%, even if the focal length varies about 5 μm, the dimensional variation becomes 0.01 μm or less. However, when the slice level is 70%, if the focal length varies about 5 μm, the dimensional variation will be 0.02 μm. Therefore, when the variation (standard deviation) of the measured value when the focal length is shifted is plotted, the variation becomes minimum at a certain slice level as shown in FIG. If this value is set as a slice level, good measurement reproducibility can be maintained even if there is a focus variation. Here, the plot is made with the variation of the measured value, but the dimensions in the focused state,
Similar results can be obtained by plotting the difference between the absolute values of the defocused dimensions measured using the same slice level.

At this time, if the slice level is changed, the absolute value of the measurement also changes. Cross section SEM (Scanning E
If there is a deviation in the absolute value due to comparison of an electron microscope (electron microscope) or the like, a constant value may be added as the offset value.

In this manner, the conditions for analyzing the waveform can be determined. Since this method is routinely performed, the condition determination conventionally performed by an inexperienced operator based on intuition can be automatically performed including acquisition of a waveform. In addition, since the conditions are determined including the reproducibility of the waveform and the influence of defocus, the reproducibility and accuracy of the measurement can be improved by using these conditions.

It should be noted that such optimization of the waveform analysis conditions is necessary when the materials and shapes of the patterns to be measured are different. For example, there is a difference between a case where a convex pattern such as a wiring pattern is measured using the same resist material and a case where a concave pattern such as a through hole is measured.
Also, even with the same wiring pattern, an isolated wiring portion,
In the case where wiring is densely formed, the shape of the waveform changes because the ease of secondary electrons differs. Therefore, the analysis conditions of the present invention are determined in advance for each of these patterns,
The parameters need to be determined. However,
Once the parameters are determined, the parameters are determined so that the measured values of the dimensions do not vary even if they are affected by fluctuations in the S / N ratio of the secondary electron signal waveform and defocus. Even if the beam current value or the focal point shifts, measurement with good reproducibility can be performed.

Although the determination of the smoothing / pattern detection parameters and the determination of the length measurement parameters are performed continuously, when determining the conditions of a plurality of patterns / samples, only the determination of the smoothing / pattern detection parameters is performed first. And the length measurement parameters may be performed later.

Although the present invention has been described in detail with reference to the embodiments, it is needless to say that the present invention is not limited to the above-described embodiments and can be variously modified without departing from the scope of the invention. Absent.

[0039]

As described above, according to the present invention, even if there is a change in the S / N ratio of the secondary electron signal waveform or a change due to defocus, the variation in the measured dimensions is minimized. Since the analysis conditions of the waveform are determined, measurement with good measurement reproducibility can be performed.

The determination of these waveform analysis conditions is performed by trial and error based on the intuition of an experienced person, since the subsequent processing can be executed automatically if a secondary electron waveform is stored in advance. It can be performed more accurately and efficiently than the conventional method.

[Brief description of the drawings]

FIG. 1A is a flowchart showing a procedure of a method of determining a waveform analysis condition in a pattern dimension measuring method using a charged beam according to one embodiment of the present invention;

FIG. 1B is a flowchart showing a procedure of a method of determining a waveform analysis condition in a pattern dimension measuring method using a charged beam according to one embodiment of the present invention;

FIG. 2 is a diagram showing an example of determining a smoothing parameter and a pattern detection parameter according to the embodiment;

FIG. 3 is a flowchart showing a procedure of a method of determining waveform analysis conditions in a pattern dimension measuring method using a charged beam according to another embodiment of the present invention;

FIG. 4 is a view showing a slice level (length measurement parameter) dependency of a length measurement value variation in the present embodiment;

FIG. 5 is a diagram showing a change in a signal waveform due to a change in a noise removal parameter for explaining that it is necessary to optimize a waveform analysis condition;

FIG. 6 is a view showing a change in a signal waveform due to defocus for explaining that it is necessary to optimize a waveform analysis condition;

FIG. 7 is a view showing a change in a measured value due to defocus for explaining that it is necessary to optimize waveform analysis conditions.

[Explanation of symbols]

A: smoothing / pattern detection parameter determination routine, B:
Length measurement parameter determination routine, 1 ... pattern detectable area, 2 ... left-right asymmetry of waveform, area where target pattern cannot be recognized due to the influence of other patterns, 3 ... area where noise cannot be removed, 4 ... spike-like noise as pattern edge Region where misrecognition occurs, 5 ... Region where the waveform is dull and the edge disappears.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G01B 15/00-15/08 H01L 21/64-21/66

Claims (2)

(57) [Claims] In a pattern dimension measuring method using a charged beam, a signal waveform of the same pattern is taken in a plurality of times, and a smoothing parameter for removing noise of the waveform and a pattern detection parameter for detecting a pattern position are sequentially changed. Detect the left and right edge positions of the pattern of each signal waveform,
The combination of the parameters is determined so that the variation in the pattern position between the waveforms is small and the change in the edge position with respect to the variation in the parameter is minimized. Acquiring and storing a signal waveform repeatedly, changing a waveform analysis parameter for dimension measurement, and determining a parameter for minimizing dimension variation from a result of dimension measurement of the stored waveform. Dimension measurement method. 2. A pattern dimension measuring method using a charged beam, wherein a signal waveform is taken in, a smoothing parameter for removing noise of the waveform is fixed, and the left and right edge positions of a target pattern are changed while sequentially changing the pattern detection parameter. Is detected to determine the area of the pattern detection parameter in which the pattern can be detected, and then, while sequentially changing the smoothing parameter, the above processing is repeated so that the area of the pattern detection parameter in which the pattern can be detected is maximized. The parameter is set as the optimum smoothing parameter, the center value of the pattern detection parameter area in which the pattern can be detected at that time is determined as the optimum pattern detection parameter, and thereafter, the signal waveform of the same pattern is repeatedly obtained by sequentially changing the focal length. By changing the waveform analysis parameters for dimension measurement,
A pattern dimension measuring method, wherein a parameter that minimizes dimension variation is determined from a result of dimension measurement of the stored waveform.