JPH0555801B2 - - Google Patents

Description

Detailed Description of the Invention <Industrial Application Field> The present invention is applicable to minute patterns formed on a substrate with uniform light reflectance within a measurement area, such as a thin film pattern formed on a silicon wafer or a semiconductor process. Chromium (Cr) on glass masks used in
The present invention relates to a line width measuring method for measuring the width or pitch of a pattern, etc.

<Prior Art> A conventional line width measurement method of this type will be explained with reference to FIG.

In the figure, 1 is a microscope, 2 is a sample such as a silicon wafer or a glass mask set in the microscope 1, 3 is a binocular eyepiece for visually confirming the sample 2, and 4 and 5 are light sources. The light source 4 is used when measuring a reflective sample such as a silicon wafer, and the light source 5 is used when measuring a transmissive sample such as a glass mask. The optical image of the fine pattern formed on the surface of sample 2 is captured by CCD.
The image is formed on a photoelectric conversion element such as 6 and converted into an image signal. The image signal output from the CCD 6 is given to the signal processing device main body 7, where it is appropriately processed, and then given to the CRT 8. Note that the reference numeral 9 is a keyboard for specifying a measurement profile to be described later and for moving a cursor, and the reference numeral 10 is a printer for printing measurement results.

Figure 6 shows the fine pattern (shaded area) on the sample surface projected on the CRT8.

Along with the fine pattern on the sample surface, CRT8
Two cursors L ₁ and L ₂ are displayed. The operator moves these two cursors L ₁ and L ₂ by operating the keyboard 9, and specifies the pattern to be measured by sandwiching the pattern between the cursors L ₁ and L ₂ .

After positioning the sample as described above, a measurement profile corresponding to the sample is designated by operating the measurement profile key provided on the keyboard 9. Here, the measurement profile is a waveform symbolically showing the light intensity of an optical image of a fine pattern. In the case of a fine pattern with uniform reflectance formed on a substrate with uniform reflectance such as a silicon wafer or a glass mask, FIG. 7A1,
There are four types of measurement profiles as shown in B1, C1 or D1. In addition, Fig. 7 A2 to D2
1 is a cross-sectional view of a sample from which each measurement profile is obtained.

7A1 and B1 are measurement profiles of a fine pattern made of a metal coating such as chromium formed on a transmission type substrate such as a glass mask, and C1 and D1 in the same figure are measurement profiles of a fine pattern formed on a reflective type substrate such as a silicon wafer. This is a measurement profile of a resist film pattern. Since the stepped portion of the metal coating, such as the Kumuro pattern formed on the glass mask, is steep, the measurement profile is represented by two values, H level and L level, as shown in FIG. 7 A1 and B1. On the other hand, since the stepped portion of the resist film pattern formed on the silicon wafer is inclined, and the inclined portion causes diffuse reflection, the level of the measurement profile is as shown in Fig. 7C.
There are three stages such as 1 and D1.

As mentioned above, there are four types of measurement profiles, but in order to specify which part of the fine pattern is to be measured, the measurement profile keys for measuring line width are shown in Figures 8 1 to 10. It is equipped with 10 types of keys as shown. In the figure, the distance between the target edges indicated by broken lines is the line width to be measured.

Further, the keyboard 9 is equipped with 12 types of measurement profile keys as shown in FIGS. 1 to 12 for measuring the pitch of the pattern. The circle mark shown in each measurement profile is the target edge to be measured for pitch.

In the above manner, the operator selects sample 2.
After alignment and measurement profile specification, the specified line width (or pitch) is measured as follows.

For example, if the sample is a silicon wafer and the
Assume that a light intensity distribution as shown in the figure is detected by the CCD 6. First, the operator uses cursor operations to roughly sandwich the pattern whose line width is to be measured between the two cursors L ₁ and L ₂ . Suppose then that the operator operates the measurement profile key to specify, for example, the measurement profile shown in FIG. 8 and 5. In this case, edge 1 and edge 3 shown in FIG. 10 are detected, and the dimension between these two edges is measured. Specifically, when the measurement profile shown in FIG.
In the signal processing device main body 7 shown in _the figure, the first edge, edge 1
is the left edge, the second edge is the cursor _L2 , and edge 3 is the right edge. When the positions of the left and right edges are detected, the dimension between the two edges is calculated based on the pixel pitch of the CCD 6.

<Problems to be Solved by the Invention> As described above, in the conventional line width measurement method, when detecting a target edge corresponding to a specified measurement profile, a cursor that is positioned by an operator is used as a reference. The target edge is detected using an extremely simple method, such as setting the numbered edge as the left edge or the right edge depending on the specified measurement profile. Therefore, due to an operator's operation error, for example, two patterns may be moved with the cursor as shown in Figure 10.
If you sandwich it between L ₁ and L ₂ ', even though you actually want to measure the dimension W between edges 1 and 3,
There is a problem that the dimension W' between 3' is measured.

Furthermore, according to conventional measurement methods, cursor operations by the operator are essential, so when trying to measure the line widths of a large number of patterns, cursor operations must be performed for each pattern, which takes a long time. There is also the problem that it is necessary.

The present invention has been made in view of the above circumstances, and an object of the present invention is to eliminate the need for cursor operations by an operator and to accurately detect a target edge according to a specified measurement profile.

<Means for solving the problems> In order to achieve such an object, the present invention has the following features:
It has the following structure.

Measurement profile information corresponding to multiple samples is stored in advance, and the optical image obtained by enlarging and projecting the fine pattern on the sample is converted into a multi-gradation image signal.
From this multi-tone image signal, two target edges that match the specified measurement profile information and are on the line to be measured are detected, and the dimension between the two edges is calculated to determine the width or width of the fine pattern. In a line width measurement method that measures pitch, the direction in which the edge changes and the size of the edge are stored as the profile information, and the edge is detected and the edge is determined based on the multi-tone image signal. The changing direction and edge size are detected, and when the edge changing direction and edge size match the information of the specified measurement profile, regarding the detected edge, that edge is set as a target edge. Detected.

<Function> As explained in Fig. 7, in the case of a fine pattern with uniform reflectance formed on a substrate with uniform reflectance such as a silicon wafer or a glass mask, the profile can be classified into four types. . There are four types of edges that appear in this profile: (1) large rising edges, (2) small rising edges, (3) large falling edges, (4) small falling edges. Moreover, the order in which these edges line up is: ・ (1)→(3)→(1)→(3)→… … … ・(1)→(3)→(2)→(4)→(1) →(3)→(2)→(4)→…
In either case, there is a regularity in the order in which they are lined up, and based on this regularity, the present invention is able to identify edges in an image signal obtained by converting an optical image of a fine pattern. To automatically determine which edge is a target edge.

Figure 7 A1, B by the operator.
Among the profiles 1, C1, and D1, which profile matches the fine pattern on the sample, and which edge in that profile is the target edge is specified. . Next, based on the rise or fall and magnitude relationship of a certain edge in the image signal with its neighboring edges, the certain edge is determined based on the regularity of the order in which the edges are lined up. Identify which of the above types (1) to (4) corresponds to. Therefore, it is examined whether the type of that certain edge matches the target edge in the profile specified by the operator, and if it does not match, the same process is performed for adjacent edges in sequence until they match. If they match, it is determined that the edge corresponds to the edge specified by the operator.

As described above, according to the present invention, in the process of detecting a target edge, the image signal is Since it is determined whether or not the edge matches the specified edge in the specified profile, the target edge is automatically detected, so cursor operation is not required and it is easy to use. , and the target edge corresponding to the measurement profile is quickly detected.

<Example> Hereinafter, an example of the present invention will be described in detail based on the drawings.

FIG. 1 is a schematic block diagram of an apparatus using a measuring method according to an embodiment of the present invention.

In this figure, the parts indicated by the same reference numerals as those in FIG. 5 are the same constituent parts, so the description thereof will be omitted here.

The optical image of the fine pattern on the sample surface magnified by the microscope 1 is projected onto the CCD 6. The CCD 6 outputs multi-gradation image signals in parallel corresponding to the light intensity of the optical image. This image signal is sent to the parallel-to-serial converter 11
The signal is converted into a serial image signal by the A/D converter 12, and then converted into a digital signal by the A/D converter 12, which is then applied to the RAM 13. Timing pulse generator 14 provides operating timing to parallel-to-serial converter 11, A/D converter 12, and address designator 15. The address designator 15 counts these timing pulses and gives the counted value to the RAM 13 as an address signal. RAM specified by this address signal
Each pixel signal of the CCD 6 is stored in the 13 storage areas. The CPU 16 performs signal processing described later on the image signal stored in the RAM 13, processing for detecting target edges corresponding to a measurement profile, calculation of dimensions between target edges, etc.
The processing program is executed according to the processing program stored in 7. In addition, code 18 is CRT8, keyboard 9,
and an input/output interface for connecting the printer 10 to the CPU 16. keyboard 9
is provided with the measurement profile keys and the like described in FIGS. 8 and 9.

Next, the measurement method according to this embodiment will be explained through an explanation of the operation of the above-mentioned apparatus.

For example, when measuring the line width of a resist pattern on a semiconductor wafer, an operator positions the pattern at a point (for example, the center of the field) within the microscope field of view or CRT display range. continue,
Depending on the measurement profile key, Fig. 7 A1, B
Among the profiles 1, C1, and D1, the C1 profile that matches the target edge to be measured is specified, and it is specified that the target edges in that profile are the edges at both the left and right ends.
After (or before) the above operations, when the measurement start key on the keyboard 9 is pressed, the CCD
6 is processed after appropriate preprocessing such as sensitivity correction between each pixel of the CCD 6 and averaging processing. The image signal for each pixel of the CCD 6 is subjected to, for example, differential processing to detect edges. And each of these detected edges is
The starting and ending positions of each edge, the size of the edge calculated from the difference therefrom, and the positive or negative sign depending on the rising or falling edge are stored in the RAM 13, respectively. An edge number (for example, 1 to N ₀ ) is assigned to each detected edge in order from, for example, the left end in the measurement area. Figure 2a shows an image signal on the line to be measured that has been preprocessed, and Figure 2b shows an image signal that has been subjected to arithmetic processing such as differential processing to detect edges and determine the edge size and positive/negative sign. are shown respectively. Note that such arithmetic processing does not necessarily need to be performed at the preprocessing stage, and may be performed during the process of target edge detection processing, which will be described later.

After the above processing is performed, processing for detecting the left and right target edges corresponding to the edges of the specified measurement profile and processing for calculating the dimensions between the target edges are performed. Hereinafter, this processing procedure will be explained according to the flowchart shown in FIG.

Step S1: Set edge selection number “N” to “1”
Set to . In this way, for example, the leftmost edge in the image signal within the measurement area (the edge with edge number "1" shown in FIG. 2) is selected, and whether or not this edge is the target is identified in the following steps.

Step S2: Set a flag F to "0" to determine whether to identify the left edge or the right edge of the specified measurement profile. When flag F is "0", processing is performed to search for the left edge, and when flag F is "1", processing is performed to search for the right edge.

Step S3: The size of edge “N” (edge “1” at the start of measurement), that is, edge “1”
The size of is read from the RAM 13, and this value is also set in the area "N ₁ " in the RAM 13. Hereinafter, the edge set in area "N ₁ " will be referred to as edge "N ₁ ".

Step S4: Edge “N+1” (Edge “2”)
The size is read and this value is set in the area "N ₂ " in the RAM 13. Below, the area "N ₂ "
The edge set to ``N 2'' is called edge ``N ₂ ''.

Step S5: Determine whether one of the edges "N ₁ " and "N ₂ " is rising and the other is falling. If there is dust on the sample surface, proceed to the step shown in Figure 3.
As noted in S5, there is profile a where two adjacent edges are both rising, and profile b where two adjacent edges are both falling.
Edges caused by such causes are excluded from the identification target. If such an abnormal edge is detected, the process advances to step S6.

Step S6: Next, to select an edge,
Set the edge selection number “N” to “N+1” (i.e., N
=2) and return to step S4. As a result, the edge "1" read in step S3
(set to ``N <sub>1</sub> '') and edge ``3'' newly read in the second step S4 (set to ``N <sub>2</sub> ''), the type of edge ``1'' is identified. As a result, edge "2" caused by adhesion of dust or the like is skipped, and the type of edge "1" is identified based on the combination with edge "3".

Step S7: If it is determined in step S5 that one of the edges "N ₁ " and "N ₂ " is a rising edge and the other is a falling edge, it is determined whether the edge "N ₁ " is a rising edge. do. Such a judgment is made by checking the sign of the edge size (differential value) set in "N ₁ ".

Step S8: Examples of cases in which edge "N ₁ " is determined to be a rising edge include profiles c to f added to step S7. However, since profiles e and f do not exist in the profiles shown in FIG. 7, it is necessary to eliminate edges that form such profiles e and f. Therefore, in step S8, by determining whether the sizes of edge "N ₁ " and edge "N ₂ " are approximately equal, abnormal edges are excluded from the identification target in the next step S9. ing. In addition, the above profiles e and f are
This is often caused by Fresnel diffraction at the pattern periphery on the substrate, and its occurrence frequency is quite high, so eliminating it from identification targets is
Extremely useful.

Step S9: If the sizes of edge ``N <sub>1</sub> '' and edge ``N ₂ '' are not equal, change the edge selection number ``N'' to ``N+'' to select the next edge.
1” and return to step S4.

Step S10: Edge “N ₁ ” and Edge “N ₂ ”
Profiles c and d are judged to be equal in magnitude, that is, (large rise) →
(large fall) or (small rise) → (small fall). Therefore, in order to determine whether the edge "N ₁ " has a large rise or a small rise, the value of the edge "N+2" is read and this value is set in the area "N ₃ " in the RAM 13. Hereinafter, the edge set in area "N ₃ " will be referred to as edge "N ₃ ".

Step S11: Edge “N ₂ ” and Edge “N ₃ ”
Determine whether the sizes are approximately equal.

Step S12: Edge “N ₂ ” and Edge “N ₃ ”
If the size is equal, step S11
Since the profile j is as added to
Edge "N ₂ " is a large falling edge, and edge "N ₃ " is a large rising edge.

Step S13: Therefore, it is determined that the edge "N ₁ " is a large rise.

Step S14: In step S11, it is determined that the sizes of edge "N ₂ " and edge "N ₃ " are not equal because edge "N ₁ " is small, as in profile k added to step S11. There are two cases: a rising edge, and a large rising edge like profile l. Etsuji “N ₁ ”
In order to determine which type of edge ``N <sub>2</sub> '' and edge ``N ₃ '' are, the sizes are compared.

Step S15: The edge "N ₂ " is smaller in the case of profile k, so in this case it is determined that the edge "N ₂ " is a small falling edge and the edge "N ₃ " is a large rising edge. do.

Step S16: Therefore, it is determined that the edge "N ₁ " has a small rise.

Step S17: In step S14, it is determined that the edge "N ₂ " is not smaller than the edge "N ₃ " (that is, the edge "N ₃ " is smaller) because it is the case of profile l. , in this case, the edge ``N ₂ '' has a large fall,
The edge "N ₃ " is judged to be a small rise.

Step S18: Therefore, it is determined that the edge "N ₁ " is a large rise.

Step S19: In step S7, it is determined that the edge "N ₁ " is not a rising edge, that is, the edge "N ₁ " is a falling edge in the case of profiles g, h, and i. Therefore,
In this step S19, it is determined whether the sizes of the edge "N ₁ " and the edge "N ₂ " are approximately equal.

Step S20: Both edges ``N <sub>1</sub> '' and ``N ₂ '' are equal only in the case of profile g, so in this case, edge ``N <sub>1</sub> '' has a large fall, and edge ``N ₂ '' has a large fall. It is judged that it is rising.

Step S21: In step S19, it is determined that the two edges "N ₁ " and "N ₂ " are not equal in the case of profiles h and i. Profile h is (small fall) → (large rise), profile i is (large fall)
→ (small rise). In this step S21, edge "N ₁ " and edge "N ₂ "
The type of edge is identified by comparing its size with that of the edge.

Step S22: If the edge "N ₁ " is smaller, it is profile h, so the edge "N ₁ "
is determined to be a small falling edge, and edge ``N ₂ '' is determined to be a large rising edge.

Step S23: Edge “N ₁ ” is edge “N ₂ ”
If the edge is not smaller than (that is, the edge "N ₂ " is smaller), then the profile is i, so it is determined that the edge "N ₁ " is a large falling edge and the edge "N ₂ " is a small rising edge. .

Step S24: Step S13, S16, S18, S20,
When the type of edge "N ₁ " is identified in S22 and S23, the process proceeds to step S24, where it is checked whether the flag F is "0" and the edge "N ₁ " is identified.
Determine whether the type of edge is to be compared with the left or right edge of the specified measurement profile.

Step S25: If flag F is "0",
It is checked whether the type of the left edge of the profile specified in advance by the operation of the measurement profile key by the operator matches the type of the edge "N ₁ ".

Step S26: If they do not match, set the edge selection number "N" to "N+1" and proceed to step S26.
Return to S3 and select the next edge (for example, edge "1"
If the left edge does not match the desired left edge, the type of edge "2") is identified.

Step S27: If the type of edge "N ₁ " matches the type of left edge of the measurement profile, the position of edge "N ₁ " is stored as a left edge.

Step S28: After completing step S27, flag F is set to "1".

Step S29: Next, set the edge selection number "N" to "N+1" and return to step S3.
Identify the type of the next edge (for example, edge "2").

Step S30: The type of the next edge (edge "2") is identified through each process from step S3 to step S23, and when step S24 is reached, the flag F is "1" at this time, so step
Proceed to S30. In this step S30, it is checked whether the edge "N ₁ " (edge "2") matches the type of right edge of the specified measurement profile.

Step S31: If the type of edge “N ₁ ” (edge “2”) does not match the type of the right edge,
Set the edge selection number "N" to "N+1", return to step S3, and select the next edge (for example,
Identify the type of edge "3").

Step S32: If the type of edge "N ₁ " matches the type of the right edge, the position of that edge is stored as the right edge.

Step S33: When two edges matching the types of left and right edges of the specified measurement profile are identified, it is determined whether the pattern having these edges is the closest pattern to the center of the visual field. Since the center of the visual field corresponds to the center position of the CCD 6, its position is known. Therefore, by comparing the distances from the positions of the identified edges to the center of the visual field, the pattern closest to the center of the visual field is determined.

Step S34: If the pattern is not the closest to the center of the field of view, change the edge selection number "N" to "N+
1'', return to step S2, identify the next edge, and repeat the above-described process until the pattern closest to the center of the visual field is found.

Step S35: When the pattern closest to the center of the visual field is found, the dimension between the left and right edges of that pattern is calculated from the positions of the left and right edges.

Step S36: Output the calculated line width to the printer and end the process.

Next, a process for measuring the line widths of all patterns within a measurement area corresponding to a designated measurement profile will be described based on FIG. 4.

Since steps S1 to S32 are the same as the flow shown in FIG. 3, they are omitted in FIG.

Step S37: When measuring all patterns, following step S32, it is checked whether all edges within the measurement area have been identified. This means that the edge selection number "N" is the edge number (1 to N ₀ ) assigned in the preprocessing described above.
This is done by checking whether it is equal to the maximum value of N 0 (ie, N ₀ ).

Step S38: If all edges have not been identified, change the edge selection number “N” to “N+
1” and follow the steps shown in Figure 3.
Return to S2 and perform the next edge identification process.

Step S39: When all edges have been identified, the memorized dimensions between the left and right edges are calculated.

Step S40: Output the calculated line width of each pattern to the printer, and end the process.

In the above-described embodiment, the case where the line width of a pattern is measured has been described, but the same process is performed when measuring the pitch of a pattern. However, in the case of pitch measurement, specify the desired profile from the measurement profiles shown in Figure 9,
In this case, there is only one type of target edge.

Further, in the above-described embodiment, the edge identification processing is performed sequentially from the edge of the visual field, but the identification processing may be performed sequentially from the center of the visual field. Furthermore, the processing procedures and processing contents shown in the embodiments can be modified in various ways. Such variations also include
The edge type of the image signal is identified based on the magnitude relationship between the rising or falling edge and the edge, which is a feature of the present invention, and when this edge type matches the edge type of the specified measurement profile. However, as long as the edge is detected for the purpose, it is included in the present invention.

<Effects of the Invention> As described above, according to the present invention, in the process of detecting a target edge in line width measurement for measuring the width and pitch of a pattern on a substrate, the type of edge of an image signal is actively identified and the edge type is detected. Since the edge whose type matches the edge type of the specified measurement profile is detected as the target edge, the measurement pattern can be sandwiched between two cursors as in the conventional method, and the edge can be simply created using these cursors as a reference. Compared to the case where the operator selects , line width measurement can be performed easily and quickly without the need for cursor operations by the operator. Moreover, misidentification caused by unnecessary edges caused by Fresnel diffraction or dust adhesion can also be eliminated.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram of an apparatus using the measuring method according to the present invention, FIG. 2 is a waveform diagram of an image signal and its differential signal detected in an embodiment of the present invention, and FIG. A flowchart showing the processing procedure for measuring the line width of the pattern closest to the center of the visual field, which is an embodiment of the present invention. A flowchart showing a part of the processing procedure when measuring, Fig. 5 is a schematic block diagram of an apparatus using the conventional measuring method, Fig. 6 is an explanatory diagram of cursor operation in the conventional method, and Fig. 7 is the same as in the book. A waveform diagram of a measurement profile common to an embodiment of the present invention and a conventional example, and a cross-sectional view of a sample corresponding to each. FIG. 9 shows an example of a measurement profile specified for pitch measurement common to the present invention and the conventional example.
FIG. 10 is an explanatory diagram of a conventional edge detection method. 1... Microscope, 2... Sample, 3... Binocular eyepiece, 4, 5... Light source, 6... CCD, 8...
CRT, 9...keyboard, 10...printer,
11... Parallel-serial converter, 12... A/D converter, 13... RAM, 14... Timing pulse generator, 15... Address specifier, 16...
CPU, 17...ROM, 18...I/O interface.

Claims (1)

[Claims] 1. Measurement profile information corresponding to a plurality of samples is stored in advance, an optical image obtained by enlarging and projecting a fine pattern on the sample is converted into a multi-tone image signal, and this multi-level image signal is converted into a multi-tone image signal. From the gradation image signal, two target edges that match the specified measurement profile information and are on the measurement target line are detected, and the width or pitch of the fine pattern is calculated by calculating the dimension between the two edges. In the line width measurement method, the direction in which the edge changes and the size of the edge are stored as the profile information, and the edge is detected and the edge changes based on the multi-tone image signal. detecting the direction and size of the edge, and detecting the edge as a target edge if the direction in which the edge changes and the size of the edge match information in a specified measurement profile with respect to the detected edge; A line width measurement method characterized by: