U.S. patent application number 13/712611 was filed with the patent office on 2013-05-09 for pattern shape evaluation method, pattern shape evaluation device, pattern shape evaluating data generation device and semiconductor shape evaluation system using the same.
This patent application is currently assigned to HITACHI HIGH-TECHNOLOGIES CORPORATION. The applicant listed for this patent is HITACHI HIGH-TECHNOLOGIES CORPORATION. Invention is credited to Ryoichi MATSUOKA, Akiyuki SUGIYAMA, Yasutaka TOYODA.
Application Number | 20130117723 13/712611 |
Document ID | / |
Family ID | 40382213 |
Filed Date | 2013-05-09 |
United States Patent
Application |
20130117723 |
Kind Code |
A1 |
TOYODA; Yasutaka ; et
al. |
May 9, 2013 |
PATTERN SHAPE EVALUATION METHOD, PATTERN SHAPE EVALUATION DEVICE,
PATTERN SHAPE EVALUATING DATA GENERATION DEVICE AND SEMICONDUCTOR
SHAPE EVALUATION SYSTEM USING THE SAME
Abstract
A pattern shape evaluation method and semiconductor inspection
system having a unit for extracting contour data of a pattern from
an image obtained by photographing a semiconductor pattern, a unit
for generating pattern direction data from design data of the
semiconductor pattern, and a unit for detecting a defect of a
pattern, through comparison between pattern direction data obtained
from the contour data and pattern direction data generated from the
design data corresponding to a pattern position of the contour
data.
Inventors: |
TOYODA; Yasutaka; (Hitachi,
JP) ; MATSUOKA; Ryoichi; (Yotsukaido, JP) ;
SUGIYAMA; Akiyuki; (Hitachinaka, JP) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
HITACHI HIGH-TECHNOLOGIES CORPORATION; |
Tokyo |
|
JP |
|
|
Assignee: |
HITACHI HIGH-TECHNOLOGIES
CORPORATION
Tokyo
JP
|
Family ID: |
40382213 |
Appl. No.: |
13/712611 |
Filed: |
December 12, 2012 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12192317 |
Aug 15, 2008 |
8355562 |
|
|
13712611 |
|
|
|
|
Current U.S.
Class: |
716/112 |
Current CPC
Class: |
G06K 2209/19 20130101;
G06T 7/0004 20130101; G06F 30/398 20200101; G06K 9/00 20130101 |
Class at
Publication: |
716/112 |
International
Class: |
G06F 17/50 20060101
G06F017/50 |
Foreign Application Data
Date |
Code |
Application Number |
Aug 23, 2007 |
JP |
2007-216576 |
May 30, 2008 |
JP |
2008-141828 |
Claims
1-13. (canceled)
14. A pattern shape evaluation data generation device comprising:
means for generating pattern shape range data from semiconductor
design data; means for displaying said pattern shape range data;
means for inputting data regarding a change in said pattern shape
range data; and means for updating said pattern shape range data in
accordance with said data regarding a change.
15. A pattern shape evaluation data generation device comprising:
means for generating pattern direction data from one of
semiconductor design data of the electronic device or from the
image obtained by photographing the circuit pattern of the
electronic device, or pattern shape data obtained through process
simulation; means for displaying said pattern direction data; means
for inputting data regarding a change in said pattern direction
data; and means for updating said pattern direction data in
accordance with said data regarding a change.
16. A pattern shape evaluation data generation device comprising:
means for generating pattern direction data and pattern shape range
data from one of semiconductor design data of the electronic device
or from the image obtained by photographing the circuit pattern of
the electronic device, or pattern shape data obtained through
process simulation; means for displaying said pattern direction
data and said pattern shape range data; means for inputting data
regarding a change in said pattern direction data and said pattern
shape range data; and means for updating said pattern direction
data and said pattern shape range data in accordance with said data
regarding a change.
17. The pattern shape evaluation data generation device according
to claim 14, further comprising: means for inputting a parameter of
data representative of anyone of a pattern width, a pattern
direction, a diameter or radius of a circle pattern for forming a
curve pattern, and a positional relation of apex coordinate circle
patterns of said design data, to the pattern of said design data;
and means for generating said pattern shape range data from one of
said design data of the electronic device or from the image
obtained by photographing the circuit pattern of the electronic
device, or pattern shape data obtained through process simulation
by using said parameter.
18. The pattern shape evaluation data generation device according
to claim 15, further comprising: means for inputting a parameter of
data representative of anyone of a pattern width, a pattern
direction, a diameter or radius of a circle pattern for forming a
curve pattern, and a positional relation of apex coordinate circle
patterns of one of said design data of the electronic device or
from the image obtained by photographing the circuit pattern of the
electronic device, or pattern shape data obtained through process
simulation, to the pattern of said design data; and means for
generating said pattern direction data from said design data by
using said parameter.
19. The pattern shape evaluation data generation device according
to claim 16, further comprising: means for inputting a parameter of
data representative of anyone of a pattern width, a pattern
direction, a diameter or radius of a circle pattern for forming a
curve pattern, and a positional relation of apex coordinate circle
patterns of one of said design data of the electronic device or
from the image obtained by photographing the circuit pattern of the
electronic device, or pattern shape data obtained through process
simulation, to the pattern of said design data; and means for
generating said pattern direction data and said pattern shape range
data by using said parameter.
20-30. (canceled)
Description
BACKGROUND OF THE INVENTION
[0001] 1. Field of the Invention
[0002] The present invention relates to a pattern shape evaluation
method of inspecting a pattern formed on a photomask or wafer by
using an image obtained by photographing a semiconductor device and
design data of the semiconductor device, a pattern shape evaluation
device, and a semiconductor inspection system using the method and
device.
[0003] 2. Description of the Related Art
[0004] In recent years, there have been situations where
manufacture of semiconductor devices is difficult because of
minituarization and multilayering thereof have been progressed and
because logic thereof has been more complicated. As a result, there
has been a tendency that defects to be attributed to manufacturing
processes occur frequently, and it has been important to detect
detects efficiently and accurately and identify problems of
manufacturing processes.
[0005] The defects to be attributed to manufacturing processes
include a deformed pattern, a broken pattern, a shorted pattern and
the like, and these defects can be detected through comparison with
a standard pattern having an ideal shape. More specifically, an
operator selects a pattern having an ideal shape from patterns
formed on a photomask or wafer, and the selected pattern is
photographed to visualize it (i.e., to use it as a standard image).
Next, a pattern of an inspection object is photographed, positions
of the inspection object image pattern and standard image pattern
are adjusted and a difference calculation is performed. If the
inspection object image pattern contains a defect, luminance
information on a defect position becomes different from that of the
standard image at a corresponding position, and a difference amount
becomes large. By using this phenomenon, a position having a
difference value not smaller than a predetermined value is detected
as a defect position.
[0006] With this inspection approach, however, since an operator is
required to perform a registration work for the standard pattern,
there arises a problem that the standard pattern registration work
takes a long time if patterns of various shapes are to be
inspected. To solve this problem, it has been tried to shorten an
inspection time by automating the registration work by detecting a
defect through comparison between a design pattern and an
inspection pattern by using as the standard pattern the design
pattern of a semiconductor device. Detecting a defect through
comparison between a design pattern and an inspection pattern is
disclosed in Japanese Patent Publication 2001-338304. The
Publication discloses the contents that design data is changed to a
shape estimated to be actually formed in an inspection object
pattern, this shape is photographed, edges of the design data and
an inspection object pattern are correlated to each other if a
direction and an inter-edge distance of edges of the inspection
object pattern existing along a normal direction of each edge of
the design pattern are in a predetermined range, and edges of the
inspection design pattern not correlated to those of the design
pattern are detected as defects.
SUMMARY OF THE INVENTION
[0007] With this method of searching edges of the inspection object
pattern in one direction using each edge of the design pattern as a
judgment criterion, however, if a pattern shape is deformed more or
less relative to the design data, a pattern area may exist which is
difficult to be correlated to the design data, resulting in a
problem of erroneously detecting a normal pattern area as a defect.
Further, if a correlating direction range is broadened, it takes a
time to search edges.
[0008] In view of these issues, the present invention provides a
pattern shape evaluation method capable of reducing erroneous
judgments of normal pattern areas which have been the problems of
conventional techniques, and detecting a pattern defect position at
high speed by simplifying the process sequence, a pattern shape
evaluation device, a pattern shape evaluation data generation
device, and a semiconductor shape evaluation system using the
same.
[0009] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation method comprising
steps of: extracting contour data of a pattern from an image
obtained by photographing a circuit pattern of an electronic
device; generating direction data of the pattern, in accordance
with pattern shape data extracted from design data of the
electronic device; and detecting a defect of the pattern through
comparison between direction data of the pattern obtained from the
contour data, and direction data of the pattern generated from the
contour data of the pattern extracted from the design data
corresponding to a pattern position of the contour data.
[0010] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation method comprising
steps of: extracting contour data of a pattern from an image
obtained by photographing a circuit pattern of an electronic
device; generating shape range data of the pattern, in accordance
with pattern shape data extracted from design data of the
electronic device; and detecting a defect of the pattern through
comparison with the shape range data of the pattern extracted from
the pattern shape data extracted from the design data corresponding
to a pattern position of the contour data.
[0011] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation method comprising
steps of: extracting contour data of a pattern from an image
obtained by photographing a circuit pattern of an electronic
device; generating pattern direction data and pattern shape range
data, in accordance with pattern shape data extracted from design
data of the electronic device; and detecting a defect of the
pattern through comparison between direction data of the pattern
obtained from the contour data, and the pattern direction data and
the pattern shape range data generated from the pattern shape data
extracted from the design data corresponding to a pattern position
of the contour data.
[0012] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, when the
contour data of the pattern is extracted from the image, contour
pixels of the pattern are detected from the image in a pixel unit,
and new contour pixels obtained through an approximation process
for adjacent contour pixels are used as the contour data.
[0013] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, wherein
when a defect of the pattern is to be detected, the pattern
direction data of the contour data is compared with the pattern
direction data generated from the design data corresponding to the
pattern position, and a contour data area having a difference value
obtained from the two sets of pattern direction data outside a
rated range is judged as a defect.
[0014] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, when a
defect of the pattern is to be detected, a pattern position of the
contour data is compared with the pattern shape range data
generated from the design data corresponding to the pattern
position, and if the pattern position of the contour data is
outside the pattern shape range, a contour data area is judged as a
defect.
[0015] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, the
pattern direction data includes pattern direction range data, and
when a defect of the pattern is to be detected and if the pattern
direction data of the contour data is outside the pattern direction
range, a contour data area is judged as a defect.
[0016] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, a defect
having a continuous contour is detected from each pattern defect,
and an area containing the defect having the continuous contour is
detected as defect area data.
[0017] In order to settle the above-described issues, the pattern
shape evaluation method of the present invention further comprises
a unit for detecting a contour having an opposing relation from
each pattern defect, and detecting an area including the contour
having the opposing relation as defect area data.
[0018] In order to settle the above-described issues, in the
pattern shape evaluation method of the present invention, a defect
type of the defect area data is judged from a state of the pattern
of the design data at a position of the defect area data.
[0019] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation device comprising:
means for extracting contour data of a pattern from an image
obtained by photographing a circuit pattern of an electronic
device; means for generating direction data of the pattern, in
accordance with pattern shape data extracted from design data of
the electronic device; and means for detecting a defect of the
pattern through comparison between direction data of the pattern
obtained from the contour data, and direction data of the pattern
generated from the contour data of the pattern extracted from the
design data corresponding to a pattern position of the contour
data.
[0020] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation device comprising:
means for extracting contour data of a pattern from an image
obtained by photographing a circuit pattern of an electronic
device; means for generating shape range data of the pattern, in
accordance with pattern shape data extracted from design data of
the electronic device; and means for detecting a defect of the
pattern through comparison with the shape range data of the pattern
extracted from the pattern shape data extracted from the design
data corresponding to a pattern position of the contour data.
[0021] In order to settle the above-described issues, the present
invention provides a pattern shape device comprising: means for
extracting contour data of a pattern from an image obtained by
photographing a circuit pattern of an electronic device; means for
generating pattern direction data and pattern shape range data, in
accordance with pattern shape data extracted from design data of
the electronic device; and means for detecting a defect of the
pattern through comparison between direction data of the pattern
obtained from the contour data, and the pattern direction data and
the pattern shape range data generated from the pattern shape data
extracted from the design data corresponding to a pattern position
of the contour data.
[0022] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation data generation
device comprising: means for generating pattern shape range data
from semiconductor design data; means for displaying the pattern
shape range data; means for inputting data regarding a change in
the pattern shape range data; and means for updating the pattern
shape range data in accordance with the data regarding a
change.
[0023] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation data generation
device comprising: means for generating pattern direction data from
semiconductor design data; means for displaying the pattern
direction data; means for inputting data regarding a change in the
pattern direction data; and means for updating the pattern
direction data in accordance with the data regarding a change.
[0024] In order to settle the above-described issues, the present
invention provides a pattern shape evaluation data generation
device comprising: means for generating pattern direction data and
pattern shape range data from semiconductor design data; means for
displaying the pattern direction data and the pattern shape range
data; means for inputting data regarding a change in the pattern
direction data and the pattern shape range data; and means for
updating the pattern direction data and the pattern shape range
data in accordance with the data regarding a change.
[0025] In order to settle the above-described issues, the pattern
shape evaluation data generation device further comprises:means for
inputting a parameter of data representative of any one of a
pattern width, a pattern direction, a diameter or radius of a
circle pattern for forming a curve pattern, and a positional
relation of apex coordinate circle patterns of the design data, to
the pattern of the design data; and means for generating the
pattern shape range data from the design data by using the
parameter.
[0026] In order to settle the above-described issues, the pattern
shape evaluation data generation device further comprises: means
for inputting a parameter of data representative of any one of a
pattern width, a pattern direction, a diameter or radius of a
circle pattern for forming a curve pattern, and a positional
relation of apex coordinate circle patterns of the design data, to
the pattern of the design data; and means for generating the
pattern direction data from the design data by using the
parameter.
[0027] In order to settle the above-described issues, the pattern
shape evaluation data generation device further comprises: means
for inputting a parameter of data representative of any one of a
pattern width, a pattern direction, a diameter or radius of a
circle pattern for forming a curve pattern, and a positional
relation of apex coordinate circle patterns of the design data, to
the pattern of the design data; and means for generating the
pattern direction data and the pattern shape range data by using
the parameter.
[0028] In order to settle the above-described issues, the present
invention provides a semiconductor inspection system comprising a
computer equipped with a scanning electron microscope and control
means for the scanning electron microscope.
[0029] In order to settle the above-described issues, in the
pattern shape evaluation device, the pattern direction range data
takes a different value depending upon a pattern area.
[0030] As described above, according to the present invention, it
is possible to reduce erroneous judgments of normal pattern areas
which have been the problems of conventional techniques and to
detect a pattern defect position at high speed by simplifying the
process sequence, through comparison between contour data extracted
from image data and direction distribution data representative of a
pattern direction and tolerance range using design data as a
criterion.
BRIEF DESCRIPTION OF THE DRAWINGS
[0031] FIG. 1 is a block diagram illustrating a pattern shape
evaluation method according to a first embodiment of the present
invention.
[0032] FIG. 2 is a flow chart illustrating a contour data
generation sequence.
[0033] FIG. 3 is a flow chart illustrating a direction distribution
data generation sequence.
[0034] FIG. 4 is a diagram showing data to be used by pattern shape
evaluation of the present invention.
[0035] FIG. 5 is a diagram showing a computer system for executing
pattern shape evaluation of the present invention in a software
manner.
[0036] FIG. 6A to 6C are diagrams showing direction distribution
data.
[0037] FIG. 7 is a diagram showing patterns of contour data.
[0038] FIG. 8 is a diagram showing template processing results.
[0039] FIG. 9 is a flow chart illustrating an operation of a
comparison unit.
[0040] FIG. 10 is a block diagram illustrating a pattern shape
evaluation method utilizing pattern matching.
[0041] FIG. 11 is a diagram showing defect data.
[0042] FIG. 12 is a diagram showing a screen displaying pattern
shape evaluation results of the present invention.
[0043] FIG. 13 is a flow chart illustrating a pattern matching
sequence.
[0044] FIG. 14 is a flow chart illustrating a direction
distribution generation sequence according to a second
embodiment.
[0045] FIGS. 15A and 15B are diagrams illustrating a direction
distribution generation sequence according to a third
embodiment.
[0046] FIG. 16 is a flow chart illustrating the direction
distribution generation sequence according to the third
embodiment.
[0047] FIG. 17 is a block diagram illustrating a pattern shape
evaluation sequence according to a fourth embodiment of the present
invention.
[0048] FIGS. 18A to 18F are diagrams illustrating the pattern shape
evaluation according to the fourth embodiment of the present
invention.
[0049] FIG. 19 is a flow chart illustrating a defect judgment
sequence according to the fourth embodiment of the present
invention.
[0050] FIGS. 20A to 20C are diagrams illustrating the defect
judgment according to the fourth embodiment of the present
invention.
[0051] FIG. 21 is a diagram showing a semiconductor shape
evaluation system equipped with a pattern shape evaluation method
of the present invention.
[0052] FIG. 22 is a diagram showing a template group to be used for
calculating a pattern direction value of contour data.
[0053] FIG. 23 is a table to be used for judgment of a corrupt
pattern.
[0054] FIGS. 24A to 24D are diagrams illustrating a first detection
method for a corrupt pattern.
[0055] FIGS. 25A to 25A-2 are diagrams illustrating a second
detection method for a corrupt pattern.
[0056] FIGS. 26A to 26C are diagrams illustrating the details of
the second detection method for a corrupt pattern.
[0057] FIG. 27 is a flow chart illustrating the first detection
method for a corrupt pattern.
[0058] FIG. 28 is a flow chart illustrating the second detection
method for a corrupt pattern.
[0059] FIG. 29 is a block diagram illustrating a pattern shape
evaluation method according to a sixth embodiment of the present
invention.
[0060] FIG. 30 is a block diagram illustrating a pattern shape
evaluation method according to a seventh embodiment of the present
invention.
[0061] FIG. 31 is a block diagram illustrating a pattern shape
evaluation method according to an eighth embodiment of the present
invention.
[0062] FIGS. 32A to 32D are diagrams illustrating the outline of a
process of identifying a measurement point from defect data.
[0063] FIG. 33 is a flow chart illustrating a sequence of
generating direction distribution data from an image.
[0064] FIG. 34 is a diagram illustrating a general pattern
measurement method.
[0065] FIG. 35 is a flow chart illustrating a sequence of measuring
a defect area from contour data.
[0066] FIG. 36 is a flow chart illustrating a sequence of measuring
a defect area from image data.
[0067] FIGS. 37A and 37B are diagrams illustrating a relation
between a defect shape and a measuring box size.
[0068] FIG. 38 is a diagram showing the outline of determining a
measuring box size from a defect shape.
[0069] FIG. 39 is a diagram illustrating a corrupt pattern
detection method.
DESCRIPTION OF THE EMBODIMENTS
[0070] Embodiments of the present invention will be described with
reference to the accompanying drawings.
(1) First Embodiment
[0071] FIG. 1 is a block diagram illustrating a basic principle of
the present invention. FIG. 4 is a diagram showing data to be used
in each block shown in FIG. 1.
[0072] The present invention is reduced in practice by a computer
system such as shown in FIG. 5. Connected to a system bus 0507 of
the computer system are a central processing unit (CPU) 0501, a
memory 0502, an input/output (I/O) controller 0505, a network
interface 0508 and a frame memory 0510. The I/O controller 0505
connects a disc 0506 storing design data of semiconductor and a
photographed image of semiconductor, and receives from an external
a signal such as parameters to be used for inspection of the
present invention via a data input unit such as a mouse 0503 and a
keyboard 0504. The frame memory 0510 stores therein data to be
displayed on a display 0511, such as photographed images and design
data of semiconductor and results of pattern shape evaluation of
the present invention. The system bus 0507 is connected to a local
area network (LAN) 0509 via the network interface 0508, and CPU
0501 executes a pattern shape evaluation process illustrated in
FIG. 1 in a software way. As in the embodiment shown in FIG. 21, a
computer system 2102 having a function equivalent to that of the
above-described computer system is also mounted on a semiconductor
inspection system mounting a semiconductor photographing device
2101 such as an optical microscope and a scanning electron
microscope (SEM). The computer system 2102 of this semiconductor
inspection system 2103 can also perform pattern shape evaluation of
the present invention.
[0073] The details of each block shown in FIG. 1 will be now be
described.
[0074] A contour extraction unit 0103 detects a contour of a
pattern from image data 0101 obtained by photographing a circuit
pattern of an electronic device manufactured on a silicon wafer
such as shown in a portion (a) of FIG. 4 or a circuit pattern of an
electronic device formed on a mask to be used for exposure of the
pattern, and generates contour data 0107 such as shown in a portion
(b) of FIG. 4. This contour extraction unit 0103 removes
information unnecessary for pattern shape evaluation such as noises
and background contained in the image data 0101 when photographing
the pattern.
[0075] FIG. 2 is a flow chart illustrating a contour extraction
sequence. First, image data the disc or memory is read at Step
0201. Next, an edge detection process is performed for the image
data 0101 at Step 0202. Image data to be used for semiconductor
inspection is generally gray information having a luminance signal
of 8 bits per pixel, 16 bits per pixel or the like which signal has
a slightly different luminance value distribution state (such as a
luminance distribution having peak pattern positions and a
luminance distribution having stepwise pattern positions) depending
upon a semiconductor photographing device and a sample material
quality. As disclosed in the image characteristic extraction
paragraph in "Computer Image Processing" written by Hideyuki
TAMURA, an edge can be detected by utilizing an edge detection
operator suitable for a luminance value distribution state of a
pattern area. Since edge data is generally gray information,
contour data of a pattern is generated by subjecting the edge data
to a binarizing process at Step 0203 by binary image processing
disclosed in "Computer Image Processing" and to a line thinning
process at Step 0204. These processes are executed for all pixels.
The contour data generated finally is written in the memory or disc
at Step 0205. In this embodiment, although description will be made
on an example wherein a pattern defect is detected by using the
contour data 0107 expressing a pattern position on the pixel unit
basis, a pattern position may be expressed by a pixel not larger
than by one pixel, by subjecting edge data to a subpixel position
estimation process using a fitting function disclosed in detection
techniques for patterns and figures in "Digital Image Processing"
written by CG-ARTTS Association. In this case, as compared to
processing contour data on the pixel unit basis, a defect position
can be detected more precisely.
[0076] A direction distribution generation unit 0104 generates
direction distribution data 0108 such as shown in a portion (d) of
FIG. 4 from design data 0102 such as shown in a portion (c) of FIG.
4. The direction distribution data 0108 is data for judging an
abnormal or defective pattern through comparison with the contour
data 0107, and is characterized in that this data has values
related to the following two items as its pixel values.
(1) Pattern Tolerance Area
[0077] All pixels of the direction distribution data 0108 have data
for judging whether pattern deformation is in an allowable range
(hereinafter called a tolerance area) or an unallowable range
(hereinafter called an abnormal area), by using as a criterion a
pattern position of the design data 0102.
(2) Pattern Direction
[0078] All pixels in the tolerance area have data for judging, from
comparison with the data regarding a pattern direction contained in
the contour data 0107, whether the contour data 0107 is normal or
abnormal.
[0079] The direction distribution data shown in a portion (d) of
FIG. 4 has a negative value "-1" as a pixel value corresponding to
the pattern abnormal area, and a positive value "0.degree. " to
"135.degree. " representative of the normal pattern direction at
each pixel position. By comparing this direction distribution data
0108 with the contour data 0107 as shown in a portion (e) of FIG.
4, it is possible to judge as a defect a defect pattern 0401 in the
tolerance area whose pattern direction value of the contour data
0107 existing in the tolerance area, e.g., such as shown in a
portion (f) of FIG. 4 is considerably different from the direction
value of the direction distribution data 0108, and a defect pattern
0402 of the contour data 0107 in the abnormal area/
[0080] FIG. 3 is a flow chart illustrating a process of generating
direction distribution data 0108.
[0081] The direction distribution data 0108 can be generated, for
example, by the sequence illustrated in the flow chart of FIG. 3.
It is however to be noted that a criterion for judging a pattern
tolerance and direction whether it is abnormal or normal is to be
determined by know how of a semiconductor developer. The generation
sequence for the direction distribution data 0108 to be described
hereunder is not intended to limit a distribution state and its
generation method for the direction distribution data 0108
representative of the pattern tolerance and direction.
[0082] First, an initial value "-1" representative of an abnormal
area is written as the direction distribution data 0108 at Step
0301. The direction distribution data 0108 is data capable of
expressing a pattern area contained in the image data 0101 at the
same resolution as that of the contour data 0107. Next, design data
(corresponding to the pattern contained in the image data 0101) is
read from the disc or memory at Step 0302. The semiconductor design
data 0102 is generally written with the number of apexes and the
coordinates of each apex on a wafer, as information constituting
the pattern. The apex coordinates are converted at Step 0303 into
apex coordinates on the direction distribution data 0108, in
accordance with a photographing magnification of the image data
0101, a pattern direction value for each pixel between apex
coordinates is obtained at Step S0304 by a method to be describe
later, and the pattern direction value is written in the direction
distribution data 0108 as a one pixel width pattern at S0305. This
pattern write is performed for all patterns contained in the image
data 0101at Step 0306.
[0083] Next, description will be made on a method of obtaining a
pattern direction value from the design data 0102. A direction
value of a straight line pattern between apexes can be obtained
from apex coordinates constituting the start and end points of the
straight line pattern. This will be described by using an example
shown in FIG. 6.
[0084] FIG. 6A is a diagram showing an image obtained by converting
a rectangle pattern of design data constituted of apex coordinates
A to D into image coordinates.
[0085] Straight lines between a point A (x=50, y=50) and a point B
(x=100, y=50) and between a point C (x=50, y=200) and a point D
(x=100, y=200) are straight lines along a 0.degree. direction
because the values of Y-coordinates of two points are equal, and
straight lines between the points A and C and between the points B
and D are straight lines along a 90.degree. direction because the
values of X-coordinates of two points are equal. A direction value
of an apex position, and hence a pattern direction, is judged from
the relation between two sets of apex coordinates interposing the
first-mentioned apex. It is very difficult to reproduce a shape of
an apex of design data, i.e., a right angle pattern, on a wafer, An
apex area 0404 of design data shown in the portion (c) of FIG. 4
becomes an apex area 0403 of a pattern formed on a wafer
corresponding to the apex area of design data having an oblique
direction angle shown in the portion (a) of FIG. 4. Therefore, a
pattern shape near an apex area is permitted to have displacement
from the pattern shape of design data, more than other pattern
shapes to some degree. Therefore, the area near an apex area of the
pattern of the direction distribution data 0108 has preferably a
direction value at which an oblique direction pattern is judged
normal at the comparison process 0105 to be later described. For
example, the apex A shown in FIG. 6A is judged as a pattern in a
45.degree. direction from the relation between apex coordinates X
and Y of the A, B and C points (X coordinate of point A=X
coordinate of point C, Y coordinate of point A=Y coordinate of
point B, X coordinate of point A<X coordinate of point B), and
the apex D shown in FIG. 6A is judged as a pattern in a 135.degree.
direction from the relation between apex coordinates X and Y of the
D, B and C points (X coordinate of point D=X coordinate of point B,
Y coordinate of point D=Y coordinate of point C, X coordinate of
point D<X coordinate of point C).
[0086] Next, an expansion process is performed for the direction
distribution data 0108 at Step 0307, and a result of conversion of
a direction identification value into a pattern direction value is
written in the direction distribution data at Step 0308. The
expansion process is general image processing of referring to
several peripheral pixels of a processing object pixel, and
replacing the value of the processing object pixel with the highest
pixel value among the referred pixels, and is called also a maximum
value filter process or a MAX filter process. Next, the direction
value of the pattern is converted into the direction identification
value which is in turn written in the direction distribution data
0108. For the use convenience of the expansion process, the
direction value of the pattern obtained by the above-described
method is once replaced with the direction identification value
such as described below. For example, the direction identification
value takes "1" for the pattern area in the 0.degree. direction,
"2" for the pattern area in the 90.degree. direction, "3" for the
pattern area in the 45.degree. direction, "4" for the pattern area
in the 135.degree. direction, and "-1" for the area other than the
pattern. By utilizing this direction identification value, the
direction distribution data 0108 representative of an allowable
range of a pattern shape having a width of several pixels can be
generated from the direction distribution data 0108 of a pattern
having a one pixel width, by a simple method. It is an important
point to set the direction identification value of the apex area to
a lager value than that of the straight line area.
[0087] The direction identification value of the pattern area is
larger than the direction identification value "-1" of the area
other than the pattern. Therefore, for example, if a reference
pixel area is 3.times.3 pixels and one expansion process is
performed for the whole plane of the direction distribution data
0108, the pattern area having a one-pixel width can be converted
into a pattern having a three-pixel width area. By repeating this
expansion process, it is possible to broaden the pixel width of a
pattern. For example, if a pattern tolerance range corresponds to a
five-pixel width in image coordinates, for the direction
distribution data 0108 having a tolerance range of a 5.times.5
pixel width, direction distribution data 0108 having a tolerance
range of 5.times.5 pixel width can be generated by the expansion
process with reference pixels of 3.times.3 pixels and two
repetitions. Further, the direction identification value of the
apex area of a pattern is larger than that of the straight line
area of the pattern, so that the apex area will not be reduced by
the expansion process. With this expansion process, it becomes
possible to generate the direction distribution data 0108 having
the direction identification values of "-1", "1" to "4" shown in
FIG. 6B. FIG. 6C shows the direction distribution data whose
direction identification values are converted into the pattern
direction values. An area of "-1" indicates an abnormal area, and
the areas of 0.degree., 45.degree., 90.degree. and 135.degree. are
data representing the tolerance range of the pattern.
[0088] In the comparison process 0105, a pattern position in the
contour data 0107 is compared with the direction distribution data
0108 corresponding to the pattern position to detect defect data
0106 in the contour data 0107. FIG. 9 is a flow chart illustrating
the comparison process 0105. First, the contour data 0107 and
direction distribution data 0108 are read from the memory or disc
at Steps 0901 and 0902. Next, a pattern position in the contour
data 0107 is detected at Step 0903, and the direction distribution
data 0108 corresponding to the pattern position is referred to.
[0089] Next, description will be made on a pattern defect detection
sequence by using as an example, comparison between the direction
distribution data such as shown in a portion (g) of FIG. 4 and the
contour data such as shown in a portion (h) of FIG. 4. Pixels in a
thick black frame in the direction distribution data shown in the
portion (g) of FIG. 4 corresponds to pattern positions of the
contour data shown in the portion (h) of FIG. 4.
[0090] At pattern positions A2 and A3, since the direction value of
the direction distribution data corresponding to the pattern
positions is "-1", it is judged at Step 0904 that the pattern is in
the abnormal area, and is a defect. At pattern positions A0 and A1,
since the direction value of the direction distribution data is
positive, the pattern direction value of each contour pixel is
detected at Step 0905. The pattern direction value can be detected
by using, for example, templates shown in FIG. 22. Each template is
constituted of a matrix of 3.times.3 pixels each having a weight
coefficient. Coefficient layouts indicate pattern directions of
0.degree., 22.5.degree., 45.degree., 67.5.degree., 90.degree.,
110.5.degree., 135.degree., and 157.5.degree.. These fourteen
matrices are convolved into 3.times.3 pixel areas around each
pattern position of the contour data 0107, and the direction value
of the template having a highest output value is used as the
direction value of the pattern position. For example, FIG. 8 shows
the results of convolution of fourteen matrices shown in FIG. 22
upon each of three types (0.degree., 90.degree. and 45.degree.
directions, contour pixel=1, non-contour pixel=0) of 3.times.3
pixel areas around each contour pixel such as shown in FIG. 7. It
can be understood that the largest values are given for the
convolution results by No. 1 (direction value 0.degree.) relative
to the pattern position in the 0.degree. direction, for the
convolution results by No. 8 (direction value) 90.degree.) relative
to the pattern position in the 90.degree. direction, and for the
convolution results by No. 4 (direction value 45.degree.) relative
to the pattern position in the 45.degree. direction. By the
template process of this kind, the direction value of the pattern
position can be detected. If the pattern position direction cannot
be judged precisely for a small area such as 3.times.3 pixels,
because of influence of noises contained in the direction
distribution data 0108, the matrix size shown in FIGS. 6A to 6C may
be enlarged using similar coefficient layouts, and the
above-described template process is executed so that a precise
pattern direction value can be detected.
[0091] By applying a broken line approximation process for a line
figure in the binary image processing described in "Computer Image
Processing" written by Hideyuki TAMURA, to the contour data 0107,
an approximate straight line of a pattern can be formed from
information on a plurality of pattern positions consecutive in the
contour data 0107. By replacing the straight line position with
pixel positions of the pattern, it is possible to suppress a local
shape variation of a pattern to be caused by influence of noises or
the like contained in the image data 0101, and to use the direction
value of the straight line itself as the direction value of the
pattern position. This calculation of the direction value of the
contour data 0107 may be performed during a sequence by the contour
extraction unit 0103.
[0092] Next, an absolute difference value is calculated at Step
0906 between the direction value of the pattern of the contour data
0107 and the direction value of the direction distribution data
0108. For example, since the pattern position AO shown in the
portion (h) of FIG. 4 has a pattern direction value of 67.5.degree.
and a direction distribution data direction value of 90.degree., an
absolute direction difference value is ABS(67.5-90)=22.5. Since the
pattern position A1 has a pattern direction value of 22.5.degree.
and a direction distribution data direction value of 90.degree., an
absolute direction difference value is ABS(22.5-90)=67.5. This
absolute direction difference value is compared with a threshold
value which is used for judging whether a pattern is a detect or
not. For example, if a pattern deformation allowable range is
0.degree. to 30.degree. and a threshold value is 30, then a pattern
having an absolute direction difference value larger than 30 is
judged as a defect at Step 0907. Since a contour line pixel A1 has
an absolute direction difference value of 22.5, this pixel is
judged normal, whereas since a contour pixel A2 has an absolute
direction difference value of 67.5, this pixel is judged as a
defect. Lastly, data representative of a pattern judged as a defect
is written in the memory or disc, The threshold value to be used
for defect judgment may be set as an inspection parameter by a user
via the mouse or keyboard.
[0093] This comparison process is executed for all contour pixels
in the contour data 0107 so that pattern defects contained in the
contour data such as shown in the portions (f) and (i) of FIG. 4
can be detected.
[0094] In the present invention, a pattern shape evaluation is
performed by comparing the pattern of the contour data 0107 such as
shown in the portion (e) of FIG. 4 with the direction distribution
data 0108 corresponding to the pattern position. It is therefore
preferable to generate the direction distribution data 0108 in such
a manner that the position of the pattern to be compared and the
position of the direction distribution data indicate the same
position on a wafer. However, if the positional relation between
the contour data 0107 and direction distribution data 0108 is
unknown, as shown in FIG. 10 a position shift amount of the image
data 0101 relative to the design data 0102 or a position shift
amount (hereinafter called a shift amount 1002) of the design data
0102 relative o the image data 0101 can be calculated through
pattern matching 1001 between the design data 0102 and image data
0101, and pattern shape evaluation becomes possible by the
comparison process 0105 considering the shift amount 1002. However,
with a general pattern matching method such as a normalized
correlation method and a residual minimum method, a precise shift
amount 1002 may not be detected if the pattern shape of the design
data 0102 is different from a pattern shape contained in the image
data 0101. Even if pattern shapes are different, a precise shift
amount can be detected by an approach disclosed in
JP-A-HEI-6-96214. According to this approach, a shift amount is
detected by a normalized correlation method after the shape of
design data is adjusted to be aligned with the pattern shape on a
wafer.
[0095] FIG. 13 is a flow chart illustrating the pattern matching
1001. First, the design data 0102 and image data 0101 are read from
the disc or memory at Steps 1301 and 1302, and pattern matching
adaptive to the different pattern shapes is executed to detect a
shift amount 1002 at Step 1303. If a shift amount is detected by
pattern matching, a direction value of the direction distribution
data is obtained from coordinate values obtained by adding the
shift amount 1002 to the coordinate values of the pattern position
of the contour data 0107 to thereby execute the comparison process.
In this manner, the pattern shape evaluation can be made even if
the positional relation is unknown between the image data 0101 and
design data 0102.
[0096] In the above-described comparison process, although a defect
position in the contour data 0107 is detected on the pixel unit
basis, defects consecutive as a pattern may be detected as
coordinate data (x0, y0).fwdarw.(x1, y1) representative of a
rectangular area containing a set of defect pixels as shown in FIG.
11, by performing a labeling process for the consecutive defects
disclosed in the binary image processing of "Computer Image
Processing" by Hideyuki TAMURA in combination with a process of
searching a rectangular area containing the same label.
[0097] The design data 0102 and image data 0101 used for inspection
and the defect data 0106 may be written in the frame memory in the
format such as shown in FIG. 12 to display these data on a display
and provide a user with the inspection results.
[0098] In the above description, a direction value of a normal
pattern is input to the direction distribution data 0108, and a
pattern whose absolute direction difference value is outside a
certain range, this pattern is detected as the defect data 0106.
Instead, for example, an allowable pattern direction range may be
input to the direction distribution data 0108, and the comparison
process 0105 judges whether the direction value of the contour data
0107 is inside or outside the allowable pattern direction range to
thereby detect the defect data 0106. In this case, the direction
value for detecting a defect and an abnormal state can be adjusted
depending upon a pattern position.
[0099] As described above, according to the embodiment, erroneous
judgment of defects which have been a problem of a conventional
method can be reduced by comparing the contour data extracted from
the image data with the direction distribution data indicating a
pattern direction and a tolerance range using the design data as a
reference. Further, since the process sequence is simple, detecting
a defect position of a pattern at high speed can be realized.
(2) Second Embodiment
[0100] This embodiment describes a method of inspecting an abnormal
state and a defect of a pattern by utilizing direction distribution
data 0108 defined by a user. This invention is also reduced in
practice by the computer system shown in FIG. 5. Processes other
than those of the direction distribution generation unit 0104 are
similar to those of the first embodiment, and the description
thereof is omitted.
[0101] FIG. 14 is a flow chart illustrating the process to be
executed by the direction distribution generation unit 0104 of the
present invention. The design data 0102 is first read at Step 1401,
direction distribution data 0108 is generated by a simplified
method described in the first embodiment at Step 1402, and
thereafter the direction distribution data 0108 is written in the
frame memory and displayed on the display at Step 1403. Next, a
user is urged to input a correction value of the direction
distribution data 0108 via the keyboard or mouse. In accordance
with the correction value designated by the user, the direction
distribution generation unit 0104 updates the direction
distribution data 0108 at Step 1404, and thereafter the defect
detection is performed in the manner described in the first
embodiment. User definition of the direction distribution data 0108
can thus be performed, and it is possible to detect the abnormal
state and defect of a pattern intended by the user.
(3) Third Embodiment
[0102] This embodiment shows a method of inspecting an abnormal
state and a defect of a pattern by utilizing parameters designated
by a user for generating direction distribution data 0108 and
efficiently generating the direction distribution data 0108. This
invention is also reduced in practice by the computer system shown
in FIG. 5. Processes other than those of the direction distribution
generation unit 0104 are similar to those of the first embodiment,
and the description thereof is omitted.
[0103] FIG. 16 is a flow chart illustrating a generation sequence
for direction distribution data.
[0104] First, a parameter input screen for generating the direction
distribution data 0108 is displayed on the display at Step 1601. As
shown in FIG. 15A, the display also displays data allowing a user
to judge the relation among respective parameters, design data 1501
and a tolerance area 1502. Examples of a parameter to be designated
by a user are shown below.
[0105] (1) H: a parameter for setting a tolerance range of a
straight line area in the 0.degree. direction.
[0106] (2) W: a parameter for setting a tolerance range of a
straight line area in the 90.degree. direction.
[0107] (3) R1, a1(x, y): a parameter for generating a tolerance
range outside a pattern corner area.
[0108] (4) R2, a2(x, y): a parameter for generating a tolerance
range inside a pattern corner area.
[0109] Next, the design data 0102 corresponding to the image data
0101 to be inspected is read at Step 1602, and the direction
distribution data 0108 is generated by using the parameters at Step
1603.
[0110] The direction distribution generation unit 0104 writes a
direction value 90.degree. in the W range of the direction
distribution data 0108, for the pattern straight line area in the
0.degree. direction of the input design data 0102, by using the
pattern position of the design data as a reference, and writes a
direction value 0.degree. in the H range of the direction
distribution data 0108, for the pattern straight line area in the
90.degree. direction, by using the pattern position of the design
data as a reference.
[0111] A tolerance range of the pattern apex area is determined by
utilizing two circle patterns constituting the inner and outer
frames of the direction distribution data, the shape of the pattern
corner area constituting part of the circle pattern, as shown in
FIG. 15A. a1(x, y) and a2(x, y) are parameters indicate relative
positions between the pattern apex positions and the center
positions of the circle patterns, and R1 and R2 are parameters
indicate the radiuses of the circle patterns. A direction value is
written in the tolerance range constituted of the tolerance range
of the two circle patterns of the pattern apex areas and pattern
straight line areas. More specifically, as shown in FIG. 15B,
straight lines are drawn on the tolerance range from the center
coordinates of the pattern inner frame circle pattern 1504
positioned inside the pattern, and normal directions of the
straight lines are used as the pattern direction values. For
example, since a position b1 has coordinates on the straight line
in the 157.5.degree. direction in the circular coordinates, a
direction value is 67.5.degree., since a position b2 has
coordinates on the straight line in the 135.degree. direction in
the circular coordinates, a direction value is 45.degree., and
since a position b3 has coordinates on the straight line in the
112.5.degree. direction in the circular coordinates, a direction
value is 22.5.degree.. These parameters are designated by using the
mouth or keyboard. After the direction distribution data 0108 is
generated, defect detection is performed in the manner described in
the first embodiment.
[0112] As described above, according to the present invention, a
user is urged to input a very small number of parameters regarding
generation of the direction distribution data 0108, and the
direction distribution data intended by the user can be generated
efficiently. An abnormal state and a defect of pattern can
therefore be inspected.
(4) Fourth Embodiment
[0113] In this embodiment, description will be made on a pattern
shape evaluation method of identifying a defect type by using the
defect data 0106 detected by the pattern shape evaluation described
in the first embodiment.
[0114] FIG. 17 is a block diagram of the present invention. The
sequence up to generating the defect data 0106 is similar to that
described in the first embodiment, and the description thereof is
omitted. FIGS. 18A and 18B are images showing a design data pattern
(broken line) and a contour data pattern (solid line) superposed
each other. FIG. 18A shows a shorted contour data pattern, and FIG.
18B shows a broken contour data pattern. As the comparison process
with the direction distribution data described in the first
embodiment is performed for the two contour patterns, a defect
pattern shown in FIG. 18C can be detected for the pattern shown in
FIG. 18A, and a defect pattern shown in FIG. 18D can be detected
for the pattern shown in FIG. 18B. Since these defect patterns are
made of dot train coordinate data, a pattern shorted area and a
pattern broken area are detected by using the dot train coordinate
data. A defect existing inside the design data pattern has a high
possibility that the defect is a shorted pattern, whereas a defect
existing outside the design data pattern has a high possibility
that the defect is a broken pattern. Further, as shown in FIGS. 18C
and 18D, the pattern broken area and shorted area have each
commonly opposing defect areas. The pattern shorted area and broken
area are detected by using these characteristics.
[0115] A specific embodiment will be described with reference to
the flow chart shown in FIG. 19. First, in order to discriminate
consecutive defect coordinates from the dot train coordinate data
of a defect such as shown in FIG. 20A obtained by the comparison
process 0105, the labeling process publicized as the binary image
processing techniques in "Computer Image Processing" by Hideyuki
TAMURA is executed at Step 1901, and the broken line approximation
process disclosed as the binary image processing techniques in
"Computer Image Processing" by Hideyuki TAMURA is executed at Step
1902 to convert the defect pixel dot train into line segment data
such as shown in FIG. 20B. Next, a line segment having a different
label existing in a vertical direction to a direction of each line
segment or in a range of a vertical direction .+-.L.degree. is
searched at Step 1903, and if it is judged at Step 1904 that a line
segment direction difference is in an M range and a distance
between lines is in an N range, then this line segment section is
judged as a broken or shorted pattern area, such as shown in FIG.
20C, at Step 1905. This judgment process is performed for all
labels. Next, a design data pattern corresponding to this area is
read at Step 1906, the shorted area or broken area is judged at
Step 1907 through judgment of whether the area is inside or outside
the pattern constituted of apex coordinates of the design data.
Lastly, coordinate information on a rectangle containing the
shorted area or broken area and shorted/broken area judgment
information are written as defect data in the memory or disc at
Step 1908. Parameters such as L, M and N used for judgment can be
designated by a user via the mouth or keyboard.
[0116] As described above, according to the present invention, a
broken area and a shorted area of a pattern can be detected by
using the coordinate data of a defect pattern expressed by dot
train and by using the design data, and a user can be supplied with
information arranged as defect data.
(5) Fifth Embodiment (Detection of Corrupt Pattern)
[0117] This embodiment shows a method of detecting an area where a
pattern which should exist does not formed (hereinafter called a
pattern corrupt area), through comparison between direction
distribution data and contour data. FIG. 39 is a diagram showing a
tolerance area 3903 and an abnormal area 3902 of direction
distribution data of a corrupt pattern area and contour data 3901,
superposed each other. Since a contour does not exist in the
pattern corrupt area, the pattern corrupt area can be detected by
detecting a discontinuous section 3904 of the contour data on the
direction distribution data. Description will be made on two
methods of detecting the discontinuous section 3904 of the contour
data.
[0118] (1) Corrupt Pattern Detection Method A
[0119] This corrupt pattern area detection method will be described
with reference to FIGS. 24A to 24D. FIG. 24A shows design data,
FIG. 24B shows direction distribution data generated from the
design data shown in FIG. 24B, and FIG. 24C shows contour data. The
contour data has a discontinuous section 2303 of the contour to be
detected. FIG. 27 is a flow chart illustrating the operation to be
executed by the corrupt pattern detection method A. First, a length
of each line segment (a line segment length in the pixel unit
obtained from the photographing conditions such as an image
photographing magnification factor) constituting the pattern of
design data is obtained from the design data, together with its
serial number at Step 2701. For example, the pattern of the design
data shown in FIG. 24A is constituted of four line segments shown
in FIG. 24A-1 to 24A-4. Line segment serial Nos. (1) to (4) are
assigned to the line segments, and line segment lengths L1 to L4
are obtained. The line segment serial Nos. and line segment lengths
are stored in the memory 0502 or disc 0506 in the table format such
as shown in FIG. 23. Next, data is generated at Step 2702, with the
tolerance area of the direction distribution data being correlated
to the segment serial number. For example, the line segment serial
number (1) is correlated to a tolerance area (1) of the direction
distribution data. Similarly, the line segment serial number (2) is
correlated to a tolerance area (2), the line segment serial number
(3) is correlated to a tolerance area (3), and the line segment
serial number (4) is correlated to a tolerance area (4).
Correlation is performed by searching a line segment having the
design data most nearest to the pixels of the tolerance area. Data
correlated to the line segment serial number may be stored as
special layer information of the direction distribution data. Next,
the defect judgment is performed at Step 2703 by using the
direction distribution data in the manner similar to that of the
above-described embodiments. If contour data exists in the normal
area during the defect judgment, the number of edge pixels of the
line segment serial number corresponding to the contour pixel
position is counted up, and stored in the table such as shown in
FIG. 23 at Step 2704. After defect judgment of contour data, a line
segment length at each line segment serial number is compared with
the edge pixel number by using the table information shown in FIG.
23. If the edge pixel number is small as compared to the line
segment length, position information 2304 of the line segment or
line segment coordinates are written in the memory 0502 or disc
0506 as corrupt area data. Since a pattern shape may be distorted,
expanded or contracted, a rated value such as the number of pixels
several per cent or smaller relative to the line segment length may
be set to suppress erroneous detection of a corrupt pattern
area.
[0120] (1) Corrupt Pattern Detection Method B
[0121] Another method of detecting a corrupt pattern area will be
described with reference to FIGS. 25A to 25A-2, FIGS. 26A to 26C
and FIG. 28. FIG. 28 is a flow chart illustrating the operation to
be executed by the corrupt pattern detection method B. First, by
referring to pattern direction information of the direction
distribution data, a plurality of sets of direction distribution
data are generated indicating the tolerance area of each direction,
at Step 2801. For example, generated from the direction
distribution data (tolerance area 2501 and abnormal area 2502)
shown in FIG. 25A are direction distribution data sets (FIG. 25A-1,
hereinafter called direction distribution data A) indicating the
tolerance areas in the pattern directions of 0.degree. to
45.degree. and 135.degree. to 180.degree. and direction
distribution data set (FIG. 25A-2, hereinafter called direction
distribution data B) indicating the tolerance areas in the pattern
direction of 46.degree. to 134.degree.. Next, the direction
distribution data A is scanned in the horizontal direction and the
direction distribution data B is scanned in the vertical direction
to search whether the contour data is present or absent in the
normal area at Step 2802. With reference to FIGS. 26A to 26C,
description will be made on a specific contour data search method.
FIG. 26A is a diagram showing part of the direction distribution
data in the pattern directions of 0.degree. to 45.degree. and
135.degree. to 180.degree. shown in FIG. 25A-1, and showing a
tolerance area 2601, an abnormal area 2602, a pattern of design
data 2603 and contour data 2604 superposed each other. Description
will be made on an example of detecting a discontinuous section
2308 as a corrupt pattern area. The direction distribution data is
scanned in the horizontal direction in the order of line (1) to
line (6). FIG. 26B is a graph showing a presence/absence of the
tolerance area, abnormal area and contour in each of the lines (1)
to (6). In the lines (1), (2), (5) and (6) without a corrupt
pattern, contour data exists in the tolerance area, whereas in the
lines (3) and (4) with a corrupt pattern, contour data does not
exist in the tolerance area. By detecting whether the contour data
exists in the tolerance area during the line scanning in this
manner, a pattern area 2607 of design data corresponding to the
defect area of the design data 2603 corresponding to the pattern
corrupt area can be identified. However, in the case wherein
contour data is lost due to noises contained in a photographed
image of the contour data, there is a possibility that this lost
area is erroneously detected as a corrupt pattern area. To avoid
this, after candidates for the corrupt pattern area are identified
by the above-described method, the number of consecutive pixels of
the corrupt area candidate is calculated at Step 2803, and only the
area having not smaller than a rated value is judged as a corrupt
pattern area, and pattern position information 2609 of the pattern
area 2607 of the design data is written in the memory 0502 or disc
0506 as corrupt area data at Step 2804.
[0122] As described above, according to the present invention,
through comparison between the direction distribution data and
contour data, an area where a pattern which should exist is not
formed can be detected as a corrupt pattern area.
(6) Sixth Embodiment (Inspection Using Contour Data of Real Pattern
Image as Reference)
[0123] In this embodiment, description will be made on a method of
generating direction distribution data from contour data generated
from a photographed pattern image and comparing the direction
distribution data with contour data extracted from another
photographed image, with reference to FIG. 29. In the first to
fifth embodiments, the design data is used for generating the
direction distribution data. There may exist an inspection style
wherein a difference between semiconductor manufacture devices and
a secular change are checked through comparison between a pattern
shape formed normally and a pattern shape formed during actual
manufacture processes. In this inspection style, it is desired to
extract a pattern contour and detect an abnormal shape different
from the contour shape. In this embodiment, therefore, as shown in
FIG. 29, a process 2902 similar to the above-described contour
extraction is performed for image data 2901 photographed from a
normal pattern, to generate contour data 2903 of the normal
pattern. A direction distribution unit 2905 generates direction
distribution data from the contour data of the normal pattern. The
direction distribution data is compared with contour data 0107
extracted from an image 0101 to be inspected to thereby detect an
abnormal area or a defect area of the pattern shape.
[0124] The direction distribution generation unit 0104 of the first
embodiment can generate direction distribution data by the sequence
shown in FIG. 3. However, since the direction distribution
generation unit 2905 of this embodiment uses contour data as an
input, it is necessary to modify the sequences at Step 0301 to 0306
shown in FIG. 3. The flow sequence of the direction distribution
unit 2905 is illustrated in FIG. 33.
[0125] First, contour data is read at Step 3301. Next, a direction
value of a contour is calculated at Step 3302 for each pixel value
constituting the contour data in accordance with the relation of
peripheral consecutive contour pixels, and the calculated direction
value is written in the memory 0502 or disc 0506 at Step 3303. The
direction value of the contour can be detected by using the
templates shown in FIG. 22 described in the comparison process of
the first embodiment. Alternatively, by using the broken line
approximation process disclosed in "Computer Image Processing"
written by Hideyuki TAMURA or the like, contour data as dot train
information of an image is once approximated to straight line
information, and a pattern direction can be detected from
inclination information of the straight line. In this manner, a
pattern direction value of all contour pixels of the contour data
is detected at Step 3304. Steps 0307 and 0308 after the expansion
process are similar to those in the flow chart of FIG. 3, and the
detailed description thereof is omitted.
[0126] As described above, according to the present invention, the
direction distribution data is generated from the contour data of
an actually manufactured pattern and inspected. It is therefore
possible to realize inspection of an abnormal state and a defect
area of a pattern shape at high precision, through comparison with
the normally formed pattern shape.
(7) Seventh Embodiment (Inspection Using Contour Data of Simulation
Image as Reference)
[0127] In this embodiment, description will be made on a method of
generating direction distribution data from contour data extracted
from a manufactured pattern estimation image formed by process
simulation, and comparing the direction distribution data with
contour data extracted from another photographed image. In recent
semiconductor development processes, it is essential to perform a
work of estimating how a pattern shape of design data is formed on
a wafer by manufacture process simulation and adjusting the pattern
shape and layout of design data. This is because the pattern shape
and layout of design data difficult to be manufactured can be
prevented in advance.
[0128] With this inspection method, as shown in FIG. 30, direction
distribution data is generated at Step 3003 from contour data 3001
of a simulation pattern formed by process simulation, and compared
with contour data extracted from a photographed image of a
manufactured pattern to inspect an abnormal state and a defect of a
pattern relative to the simulation pattern. An input to a direction
distribution generation unit 2905 is contour data of the simulation
image so that it is possible to adopt the method of generating
direction distribution data of the sixth embodiment.
[0129] As described above, according to the present invention, the
direction distribution data is generated from the contour data of a
simulation image formed by process simulation and inspected. It is
therefore possible to realize inspection of an abnormal state and a
defect area of a pattern shape at high precision, by using
simulation data. Further, since defect data output as inspection
result indicates a difference between the simulation pattern shape
and an actual pattern shape, the defect data can be used also for
verification of simulation performance.
(8) Eighth Embodiment (Method of Generating Measuring Recipe from
Defect Information)
[0130] This embodiment shows a method of, in accordance with
information on an abnormal pattern shape area detected through
comparison between direction distribution data and contour data,
deciding a pattern measuring sequence for measuring an abnormal
shape area, and measuring the abnormal shape area in accordance
with the measuring sequence. As described in the first to seventh
embodiments, it is possible to detect abnormal areas such as a
shorted or broken pattern and a broad or narrow pattern distance,
through comparison between the direction distribution data and
contour data. A shorted or broken pattern is an issue to be settled
at semiconductor process development stages. An abnormal pattern
distance is also an issue to be settled during semiconductor mass
production, and strict inter-pattern distance measurement is
required in some cases for management of semiconductor processes.
In this case, an inter-pattern distance in an abnormal shape area
can be measured in accordance with a sequence illustrated in FIG.
31. Description will be made on the pattern measuring sequence
shown in FIG. 31. Since the sequence described in the first to
seventh embodiments can be used as the sequence up to detecting the
defect data 0106 representative of an abnormal inter-pattern
distance area, the description thereof is omitted. Data to be input
to a direction distribution generation unit 0104 shown in FIG. 31
may be any one of design data, contour data generated through
process simulation, and contour data extracted from a photographed
image of a manufactured pattern.
[0131] A measuring sequence determining unit 3101 determines a
sequence (hereinafter called measuring sequence data 3103) of
measuring an abnormal shape area and inter-pattern distances
between design data and the abnormal shape in accordance with the
defect data 0106. More specifically, obtained from information on
the abnormal shape area and design data are a measuring point, a
direction along which an inter-pattern distance is measured and
image photographing conditions if an image is required to be
photographed again. For example, the measuring sequence determining
unit corresponds to a unit for forming a measuring recipe for
controlling a measuring device if difference devices are used for
inspecting and measuring an abnormal shape area.
[0132] Lastly, in accordance with the measuring sequence data 3103,
a measuring unit 3102 measures an inter-pattern distance to obtain
measuring data 0106. Description will be made on a measuring
sequence determining method and a measuring method by using a
specific example.
[0133] FIG. 32A is a diagram showing direction distribution data
(tolerance area 3201), contour data 3203 and design data 3202
superposed each other, showing an abnormal shape area with an
extremely narrow inter-line width. This area can be detected as an
abnormal shape area 3204 of the direction distribution data, by the
pattern shape evaluation described in the first to seventh
embodiments (refer to FIG. 32B). Description will be made on two
approaches to the measuring sequence determining method and
measuring method of measuring a distance 3206 between adjacent
patterns from the data of the abnormal shape area 3204.
[0134] (1) Measuring Distance between Contour Data
[0135] The first approach is to determine a contour position to be
measured in accordance with a pattern layout state of an abnormal
shape area and design data, and to measure the distance. For
example, in FIG. 32A, since patterns are disposed in a horizontal
direction, it can be known that a contour measuring direction is a
vertical direction. Therefore, a point in the abnormal shape area
3204 remotest from the design data 3202 and a contour of an
opposing pattern relative to the point are searched, and a distance
therebetween is measured so that the inter-pattern distance in the
abnormal shape area can be measured. A search direction 3205 of the
opposing pattern becomes different depending upon whether a convex
pattern or a concave pattern is measured. It is therefore desired
that a user is urged to designate whether a convex pattern or a
concave pattern is measured, via the mouse 0503 or keyboard
0504.
[0136] FIG. 35 is a flow chart illustrating a measuring sequence
generation method and a measuring method for a distance between
contour data in an abnormal shape area.
[0137] First, a measuring point is determined from data of an
abnormal shape area at Step 3501. For example, the measuring point
is a contour position in the abnormal shape area remotest from
adjacent design data. Next, a pattern of the design data nearest to
the measuring point in the abnormal shape area is searched at Step
3502. At Step 3503 a direction of the pattern of the design data is
detected to determine a direction along which the opposing contour
as another measuring point exists. For example, if a pattern of the
design data is continuous in the horizontal direction, the opposing
contour should exist along the vertical direction. Next, a contour
is searched along the direction along which the opposing contour
exists to determine the measuring point of the opposing contour at
Step 3504. Lastly, at Step 3507 a distance is measured between the
measuring point in the abnormal shape area and the measuring point
of the opposing contour detected by search. If a distance between
the measuring point in the abnormal shape area and the design data
pattern detected at Step 3502 is measured, a shape deformation
amount of the pattern relative to the design data can be
obtained.
[0138] (2) Measuring Inter-Pattern Distance Using Photographed
Pattern Image
[0139] The second approach is to measure an abnormal shape area at
high precision from information on the abnormal pattern area and a
photographed pattern image, by using a pattern length measuring
method. If an extraction position precision of contour data to be
used for pattern shape evaluation is low, the first approach is
associated with a measurement precision problem. By using the
pattern length measuring method, a pattern can be measured at high
precision.
[0140] The pattern length measuring method is a high precision
pattern measuring method adopted by length measuring SEM mainly
used for pattern line width management. FIG. 34 illustrates the
outline of a pattern length measuring method and shows an image of
a pattern photographed with a SEM. SEM images a sloped portion and
a projecting portion of an object as white stripe image
information. Therefore, by measuring a distance between white
stripes, a pattern can be measured. However, random noises are
superposed upon a SEM image because of SEM device characteristics.
In order to increase an S/N ratio of a measuring point, a length
measuring box 3403 is disposed on a pattern to be measured, a
luminance value of each pixel in the length measuring box is
integrated along a pattern direction to form a line profile 3404,
and thereafter a length line profile 3405 is measured between
peaks, or between points corresponding to a predetermined threshold
value at the sloped portion. Since SEM has a high resolution along
a scan direction, if the pattern exists along a vertical direction
as shown in FIG. 34, an image scanned in the horizontal direction
as scan direction 3401 is used.
[0141] In order to adopt the pattern length measuring method of
measuring a pattern in an abnormal shape area, the measuring
sequence data becomes necessary including (1) a position of a
measuring pattern (a layout of the length measuring box), (2) a
size of the line measuring box and (3) a scan direction.
[0142] With reference to FIGS. 32A to 32D, a sequence of generating
the measuring sequence data will be described. FIG. 36 is a flow
chart illustrating an operation of a measuring sequence method.
First, an abnormal shape area 3204 is detected through comparison
between the direction distribution data and contour data, and
thereafter, similar to the first approach, a measuring point 3207
is determined from the data of the abnormal shape area at Step
3601. The measuring point is, for example, a contour point in the
abnormal shape area remotest from adjacent design data. Since the
length measuring box is disposed surrounding the measuring point
3207, the measuring point indicates a rough position of the length
measuring box. Next, in accordance with a contour shape of the
abnormal shape area existing around the measuring point 3207, a
size and position of the length measuring box in the abnormal shape
area is determined at Step 3602. The length measuring box is used
for integrating a luminance value of each pixel along a pattern
direction. Therefore, for example, as shown in FIGS. 37A, if a
pattern is deformed greatly and a large size length measuring box
is disposed, a line profile 3702 as an integration result becomes
broad, posing a problem that an incorrect point is measured. By
adjusting the size of the length measuring box in accordance with
the pattern deformation such as shown in FIG. 37B, a line profile
3703 can be suppressed from being broadened so that correct
measurement can be performed.
[0143] The size of the length measuring box in the abnormal shape
area can be determined from a contour shape, by measuring distances
between a straight line centering the abnormal shape area and a
nearby contour. This will be described with reference to FIG. 38.
First, a straight line 3803 is drawn passing a length measuring
point 3802 in the abnormal shape area selected from a contour line
3801 in the abnormal shape area. A straight line direction is the
same direction as the pattern direction of adjacent design data.
Next, distances between the contour positions and straight line are
measured in a range not larger than the maximum size of the length
measuring box size, by using the length measuring point in the
abnormal shape area as a center. A length of contour positions
having a distance to the straight line not larger than a rated
value is set to a height 3806 (relative to the pattern direction)
of a length measuring box 3805. A width 3807 (relative to the
pattern direction) of the length measuring box is set to a target
value of a twofold of the rated value. However, if an adjacent
contour is mixed in the length measuring box, the measuring box
size is reduced in order to avoid erroneous detection of the
measuring point. Rated values to be used for judging the
maximum/minimum range of the length measuring box and a distance
between the contour and straight line are desired to be designated
by a user via the mouse 0503 and keyboard 0504.
[0144] Next, a measuring point of an opposing pattern in the
abnormal shape area is determined. By using the method described in
the first approach, a search direction 3205 (FIG. 32B) of the
opposing pattern is obtained at Step 3603, and an opposing contour
3212 is searched to determine a measuring point 3213 on the
opposing contour at Step 3604. A position and size of a length
measuring box 3206 of the opposing pattern is determined by using
the measuring point 3213 as a center. The method of determining the
position and size of the length measuring box 3206 is fundamentally
the same as the method of determining the position and size of the
length measuring box in the abnormal shape area. If a search
direction 3205 of the opposing pattern is different from the scan
direction of a photographed pattern image, the image scan direction
is determined as the search direction 3205 to photograph an image
by changing the scan direction.
[0145] Next, if it becomes necessary to photograph an image again
because the pattern measuring direction and the scan direction are
different or the device for pattern shape evaluation is different
from a pattern measuring device, an image of a pattern in the
abnormal shape area to be measured is photographed at Step 3608.
This image is photographed in accordance with the measuring
sequence data of position information of the abnormal shape area
and a scan direction determined by the above-described sequence.
Positioning a pattern to be measured can be realized through
pattern matching with the photographed image by forming templates
of the design data near the pattern. In accordance with the
positioning result and the measuring sequence data of the position
and size of the length measuring box, the length measuring box 3206
is disposed on the image. Finally, a line profile 3210 of the
abnormal shape area and a line profile 3209 of the opposing pattern
are generated, and a distance 3211 between points of the profiles
is measured to obtain a pattern distance in the abnormal shape area
at Step 3609.
[0146] As described above, according to the present invention, the
measuring sequence data for measuring a distance between patterns
in the abnormal shape area is generated from information (position
and shape) on the abnormal shape area of the pattern detected
through comparison between the direction distribution data and
contour data, and a distance between patterns is measured. It is
therefore possible to correctly measure the distance between
patterns in the abnormal shape area necessary for management of
semiconductor processes.
[0147] The present invention described so far can be applied to a
pattern shape evaluation method and a semiconductor inspection
system of inspecting a defect position of a pattern formed on a
photomask or wafer at high speed.
[0148] It should be further understood by those skilled in the art
that although the foregoing description has been made on
embodiments of the invention, the invention is not limited thereto
and various changes and modifications may be made without departing
from the spirit of the invention and the scope of the appended
claims.
* * * * *