JP2007178294A - Edge detection method and edge detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve precise length measurement of a pattern by determination of edge position with high accuracy and satisfactory reproducibility excluding effect of noise and so forth of an edge part of a line profile, in an edge detection method and an edge detection device detecting the edge position of a pattern to be subjected to length measurement on an image. <P>SOLUTION: The edge detection method includes: a step for acquiring a plurality of line profiles at different places across a pattern to be subjected to length measurement; a step for generating a model curve from a curve near the edge of the acquired line profiles; a step for superposing the generated model curve and the plurality of profiles, respectively, to calculate fitting points on the basis of difference in brightness, respectively; and a step for calculating as the edge position the points in which the position of the point where the model curve crosses a prescribed a threshold is added to the calculated fitting point, respectively. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an edge position detection method and an edge detection apparatus for detecting an edge position of a pattern to be measured on an image.

  Conventionally, the length of a pattern on a mask or LSI acquired with a scanning electron microscope or the like is determined by determining the positions of the left and right ends of the pattern on the pattern image and measuring the distance between the two. The width dimension is automatically measured.

  At this time, the positions of the left end and the right end of the pattern are the positions of the center points of the line profile (the line profile when the horizontal axis is distance and the vertical axis is brightness for scanning across the pattern) (or the predetermined threshold value 02). . And 0.8 and the midpoint position when crossing 0.8) was determined, and the distance between the positions (left end, right end) was automatically measured as a dimension.

  Also, for the end of the line profile (the end of the pattern), first-order and second-order curves are determined in advance, and the point where the curve matches the line profile is determined as the edge position (left end, right end) The distance between the obtained positions (left end, right end) was automatically measured with the pattern dimension.

  However, the above-described conventional method has a problem that accuracy and further reproducibility cannot be obtained when noise is applied to the line profile at the edge portion of the pattern.

  In the latter latter method described above, the edge portion of the pattern is determined to be the point that most closely matches a predetermined first-order or second-order curve. There is a problem that the curve is not necessarily optimal and sufficient accuracy and reproducibility cannot be obtained.

  In order to solve these problems, the present invention generates a model curve from one or a plurality of line profiles acquired from the measurement target region, and the difference between the generated model curve and the curve near the edge of the line profile. The fitting point having the smallest absolute value of the value is obtained, and the shift amount at the predetermined threshold value is added to this to determine the edge position.

  According to the present invention, a model curve is generated from one or a plurality of acquired line profiles, and a fitting point having a minimum absolute value of a difference between the generated model curve and a curve near the edge of the line profile is obtained. By determining the edge position by adding the shift amount at the predetermined threshold value, the edge position of the edge portion of the line profile is determined as much as possible to determine the edge position with high accuracy and good reproducibility. It becomes possible to realize the length.

  According to the present invention, a model curve is generated from one or a plurality of acquired line profiles, and a fitting point having a minimum absolute value of a difference between the generated model curve and a curve near the edge of the line profile is obtained. Accurate measurement of the pattern is performed by determining the edge position by adding the shift amount at the predetermined threshold and determining the edge position with high accuracy and good reproducibility, eliminating the influence of noise etc. of the edge part of the line profile as much as possible. It was realized.

FIG. 1 shows a system configuration diagram of the present invention.
In FIG. 1, an electron optical system 1 is a lens barrel such as a scanning electron microscope, and includes an electron gun that generates an electron beam, a focusing lens that focuses the generated electron beam, and a focused electron beam. An objective lens that is narrowed down on the sample 3, a two-stage deflection system for scanning the surface of the electron beam narrowed down on the sample 3 (scanning in the X and Y directions), and an electron beam that is narrowed down further It consists of a detector that detects secondary electrons, light, and reflected backscattered electrons that are emitted when the surface of the specimen 3 is scanned with a beam. The image of the surface of the specimen 3 (secondary electron images, backscattered electrons) Image) and the like.

  The sample chamber 2 is a container that is evacuated by a vacuum evacuation system (not shown) and stores the sample 3 in a vacuum.

  The sample 3 is a sample to be measured, and is formed with a pattern to be measured, such as a mask or LSI.

  The PC 11 is a personal computer and executes various processes according to a program. Here, the image acquisition means 12, the shift amount calculation means 13, the edge position calculation means 14, the image file 15, the output file 16, and the display device 17 are displayed. , And input / output means 18 and the like.

  The image acquisition means 12 issues an instruction to a control system (not shown) and operates the electron optical system 1 to image the surface of the sample 3 (for example, an image of a mask pattern portion (secondary electron image, reflected electron image)). Is known.

  The shift amount calculation means 13 calculates the position (shift amount, see FIG. 4A) of the point where the model curve generated from the line profile intersects the predetermined threshold (see FIGS. 2 to 5). Will be described later).

The edge position calculation means 14 calculates the edge position of the pattern on the image (
(It will be described later with reference to FIGS. 2 to 5).

The image file 15 holds an image acquired from the sample 3.
The output file 16 stores the results of calculation and measurement of edge positions and the like.

The display device 17 displays an image or the like.
The input / output device 18 is various input devices (for example, a mouse and a keyboard) and various output devices.

Next, the operation of the configuration of FIG. 1 will be described in detail according to the order of the flowchart of FIG.
FIG. 2 shows a flowchart for explaining the operation of the present invention.

FIG. 2A shows a flowchart for determining a model curve.
In (a-1) of FIG. 2, S1 uses a cubic curve obtained from the first profile as a model curve. This is obtained by obtaining a plurality of profiles crossing the pattern on the image of FIG. 3B (to be described later) at different positions and approximating the curve near the edge of the first profile as it is with a cubic curve to determine a model curve. Means that.

  By the above S1, a curve near the first edge of the line profile crossing the pattern is approximated by a cubic curve, and a model curve is generated, thereby automatically generating a model curve suitable for the line profile. It becomes possible. Then, the generated model curve and the first, second,..., 50th profiles are overlapped with each other, and the fitting point is calculated for each position where the sum of the absolute values of the differences becomes the smallest ((c) in FIG. 3). (See (d)).

  In FIG. 2A-2, S11 determines the first cubic model curve from the vicinity of the edge of the first profile. This generates a model curve representing (approximate) a curve near the edge of the first profile with a cubic curve, similar to S1 described above.

  S12 compares with the first model curve near the edge of the second profile. In a well matched place, the second cubic curve approximation and the first model curve are averaged. As a result, a model curve of a cubic curve averaged between the edges of the first and second profiles is generated.

  S13 repeats this up to the 50th as a new model curve. The cubic curve finally obtained is used as a model curve.

  Through S11 to S13, a curve near the first edge of the line profile crossing the pattern is approximated by a cubic curve to generate a model curve, and then the model curve and the second line profile are averaged. Processing to create a cubic curve is repeated up to the 50th line profile, and the finally obtained cubic curve is calculated as a model curve to approximate the first to 50th line profiles to a cubic curve. Thus, it is possible to automatically generate a model curve with higher accuracy and better reproducibility by averaging.

FIG. 2B illustrates a fitting point calculation flowchart.
In FIG. 2 (b), S21 calculates the sum of absolute values of the difference between the brightness of the model curve and the profile at each position while shifting near the edge of the model curve and fits the place where the sum is minimized. Position (fitting point). This is because, as shown in FIGS. 3C and 3D described later, the brightness of the model curve and the line profile is shifted while shifting the model curve generated in FIG. 2A near the edge of the line profile. The sum of absolute values of the differences is obtained, and a fitting point at which the sum is minimum (for example, the position of the model curve where the minimum value is obtained) is calculated.

  By the above S21, it is possible to automatically calculate the fitting points that most closely match the model curve calculated in FIG. 2A and the edges of the first, second,..., 50th line profiles. Become.

FIG. 2C shows a flowchart of edge detection.
In FIG. 2C, in S31, the difference between the maximum value and the minimum value of the model curve is set to 1 (normalized).

  In S32, the position of the threshold value designated on the model curve is obtained. As shown in (e) of FIG. 4 to be described later, this obtains a position (horizontal axis is also a position) that intersects a threshold value (for example, 0.8, the optimum value is obtained by experiment) specified on the model curve.

  In S33, the distance from the start point (left end) of the model curve to the threshold value is set as a shift value. As shown in FIG. 4 (e), the distance from the start point (left end, fitting point) of the model curve to the position intersecting with the threshold value calculated in S32 is obtained as a shift value.

S34 is added to the fitting points found in each profile.
S35 is each edge position. In S34 and S35, the shift values shown in FIG. 4 (e) obtained in S33 are added to the fitting points calculated in S21 for the first to 50th line profiles, and the edge positions are determined.

  Through S31 to S35 described above, a position that intersects a predetermined threshold (for example, 0.8) on the model curve is obtained, and the distance between this position and the start point of the model curve (corresponding to the fitting point) is calculated as a shift value. It is possible to calculate the edge position of each line profile by adding the calculated shift value to the fitting points fitted to the model curve in S21 for the first to 50th line profiles. As a result, a model curve can be dynamically and optimally created from one line profile or a plurality (or all) of line profiles, and the threshold value obtained by experiment is the position of the intersection with the model curve ( The optimum edge position can be determined by calculating the edge position by obtaining the shift value and adding it to the fitting point, and the edge position of the pattern with high accuracy and good reproducibility can be automatically calculated on the image. .

3 and 4 are explanatory diagrams of the present invention.
FIG. 3A shows an example of a model curve. The illustrated model curve shows an example of a third-order model curve automatically generated in (a-1) or (a-2) of FIG. 2 described above. (A-1) in FIG. 2 is a model curve generated by approximating a curve near the edge of one (for example, the first) line profile in the 1st to 50th line profiles with a cubic curve. Further, in (a-2) of FIG. 2, the model curve of the cubic curve created at the first of the first to 50th line profiles is averaged with the second, third,..., 50th line profiles. It is a model curve of a later cubic curve.

  As described above, by generating a model curve from the vicinity of the edge of the line profile of the pattern to be measured, it is possible to dynamically and automatically create a model curve optimal for the line profile on the image.

  FIG. 3B shows how a plurality of line profiles are acquired from different locations across the pattern on the mask. Here, 50 line profiles from the first to the 50th are acquired.

  FIG. 3C shows a state in which the sum of absolute values of luminance differences is obtained while shifting the model curve to the first line profile near the edge.

  FIG. 3D shows a state where the fitting position (fitting point) of the first line profile is calculated. This is calculated by using the fitting point as the position where the sum total calculated in FIG. Note that the calculation is simplified, and in FIG. 3C, the position at which the difference in luminance (or the difference in absolute value of luminance) is minimized while shifting the model curve to the first line profile near the edge is the fitting point. You may make it ask.

  FIG. 4E shows how the shift value of the model curve is calculated. The model curve is dynamically created automatically from the line profile in FIG. 3A, and the intersection with the threshold of the model curve (for example, 0.8, which is an optimum value calculated by experiment) The distance from the start point (minimum value, position corresponding to the fitting point) of the model curve is calculated as a shift value.

  FIG. 4F shows how the edge position of the first profile is determined. The fitting point of the first profile is calculated in (d) of FIG. 3 described above, and the shift value of the model curve calculated in (e) of FIG. 4 is added to the calculated fitting point of the first line profile. Add the edge position to calculate the edge position. Similarly, the second to 50th edge positions are calculated. Then, the average value of the calculated edge positions is calculated and determined as the edge position of the pattern. Compared to the case where the model curve is fixed, the measurement reproduction accuracy is 0 in the experiment in which the edge position is calculated by automatically generating the model curve of the cubic curve dynamically from the line profile in FIGS. 2 to 4 of the present invention. It was improved by about 1 nm (20%) (see FIG. 5).

  FIG. 5 shows an example result. The state of change of 3σ (deviation) on the vertical axis when the threshold value on the horizontal axis is changed is shown. The threshold value on the horizontal axis represents the value of the threshold value that intersects the model curve of FIG. 4E described above, and the deviation 3σ is the known deviation (3σ) of the edge position obtained for the first to 50th profiles. It is.

  5A shows an example in the case of the conventional method (model curve is fixed), and FIG. 5B is the method described in FIGS. 2 to 4 of the present invention (model curve is dynamically calculated). An example of the case is shown. Each of the drawings is a curve obtained by experimenting under the same conditions. The method of the present invention shown in FIG. 5B has a good measurement reproducibility (3σ) and is about 0.1 nm (20%). I was able to improve.

  According to the present invention, a model curve is generated from one or a plurality of acquired line profiles, and a fitting point having a minimum absolute value of a difference between the generated model curve and a curve near the edge of the line profile is obtained. Accurate measurement of the pattern is performed by determining the edge position by adding the shift amount at the predetermined threshold and determining the edge position with high accuracy and good reproducibility, eliminating the influence of noise etc. of the edge part of the line profile as much as possible. The present invention relates to an edge detection method and an edge detection apparatus to be realized.

It is a system configuration diagram of the present invention. It is an operation | movement explanatory flowchart of this invention. It is explanatory drawing (the 1) of this invention. It is explanatory drawing (the 2) of this invention. It is an example of a result.

Explanation of symbols

1: Electro-optical system 2: Sample chamber 3: Sample 11: PC (personal computer)
12: Image acquisition means 13: Shift amount calculation means 14: Edge position calculation means 15: Image file 16: Output file 17: Display device 18: Input / output device

Claims (5)

In an edge position detection method for detecting an edge position of a pattern to be measured on an image,
Obtaining a plurality of line profiles at different locations across the pattern to be measured;
Generating a model curve from a curve near the edge of the acquired line profile;
Calculating the footing points based on the difference in luminance by superimposing the generated model curve and the plurality of profiles, respectively;
An edge position detection method comprising: calculating a point obtained by adding a position of a point where the model curve crosses a predetermined threshold to the calculated footing point, and an edge position.   2. The edge position detection method according to claim 1, wherein the footing point is a point at which the luminance difference is minimum, the absolute value of the luminance difference is minimum, or the sum of the absolute values of the luminance differences is minimum. .   As the model curve, a curve in the vicinity of an edge in any one line profile that crosses the pattern is set as the model curve, or a curve in the vicinity of the edge in the first line profile is set as the first model curve. The edge position according to claim 1 or 2, wherein a model curve is obtained by repeating the averaging process with the first line profile to form a new model curve for a predetermined number of line profiles. Detection method.   The edge position detection method according to claim 1, wherein the model curve is approximated by a cubic curve. In an edge position detection device for detecting an edge position of a pattern on an image,
Means for acquiring a plurality of line profiles at different locations across the pattern to be measured;
Means for generating a model curve from a curve near an edge of the acquired line profile;
Means for calculating a footing point based on a difference in luminance by superimposing the generated model curve and the plurality of profiles, respectively;
An edge position detection apparatus comprising: a point obtained by adding a position of a point where the model curve crosses a predetermined threshold to the calculated footing point, and an edge position.

Images