JP3148770B2 - Photomask and mask pattern data processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mask pattern forming method, a mask and a pattern forming method for forming a fine pattern such as an LSI on a wafer (substrate) using a projection lens.

[0002]

2. Description of the Related Art Conventionally, a projection exposure technique for forming a fine pattern such as an LSI has been required to have a high resolution. Therefore, the projection lens of the projection exposure apparatus has a resolution close to a theoretical limit determined by the wavelength of light. In recent years, the phase shift method has been studied as a method for transferring finer patterns.
In the phase shift method, it is possible to increase the resolution by adding a transparent film that generates a phase difference of about π in the transmitted light to a part of the light transmission part on the mask placed on the object plane of the projection lens Becomes This phase shift method can be classified into several methods. As a method for forming an isolated fine line pattern, there is a pattern formation method using a phase difference of about π generated at an end of a phase shift film. . This method will be described with reference to FIG. FIG. 17A shows a part of a cross section of the phase shift mask, where 60 is light for illuminating the mask, 61 is a mask substrate, 1500 is a phase shift film, and the presence of the phase shift film 1500 is shown in the figure. There is a boundary side (shifter edge) between the portion that does and the portion that does not exist. The complex amplitude distribution of the light passing through the mask is as shown in FIG. 17B, and the corresponding light intensity distribution is as shown in FIG. Here, FIGS. 17 (a), (b) and (c) assume that the horizontal axes correspond. These figures show that the light intensity of the corresponding portion becomes 0 because the phase changes by π at the edge portion. As described above, the edge of the phase shift film forms a very thin light-shielding portion. For example, when the g-line (wavelength: 436 nm) of a high-pressure mercury lamp is used, a light-shielding pattern made of ordinary chrome or the like is considered in consideration of the depth of focus. Is limited to a width of about 0.5 μm, but using this shifter edge, a pattern having a width of 0.2 μm can be formed. This line width is called a critical resolution line width. Thus, the use of the shifter edge is expected to be applied to the formation of fine patterns.

In using this shifter edge, two points will be additionally described. One is the limitation of the pattern density. When a shifter edge is used, a light-shielding portion having a fine width can be formed. For this purpose, light having a phase difference needs to be transmitted on both sides of the light-shielding portion. This indicates that there is a limit to the distance between two light-shielding portions having a fine width so that they can be formed close to each other. FIG.
Although the case where the light shielding portions having a width of μm are continuously formed at the interval p is shown, the interval p between the light shielding portions needs to be close to 1 μm.

Another point is related to control of the line width of the light shielding portion. To obtain the finest line width, a shifter edge having a phase difference of π may be used. In this case, a line width of 0.2 μm is possible. However, in actual LSI pattern formation, the minimum line width obtained by transferring a normal light-shielding body, for example, smaller than 0.5 μm, the dimension obtained by the shifter edge of 0.2 μm,
The formation of a line thicker than m may be required. FIG.
(A) and (b) show a method of forming a line width therebetween.
(A) has shown the cross section of the phase shift mask. By inserting a light shield made of chrome or the like into the shifter edge portion and adjusting the line width of the light shield, a pattern width larger than the critical resolution line width can be realized. (B) shows a planar structure of the phase shift mask. By providing a fine turn at the edge portion of the phase shift film, a pattern larger than the critical resolution line width can be realized. In this manner, if the pattern is smaller than the line width obtained by transferring a normal light-shielding body and is larger than the critical resolution line width, it can be formed using the edge of the shifter. FIG.
Such a pattern forming method including (a) and (b) is herein referred to as a shifter edge method.

The biggest problem of pattern formation using the shifter edge method is that a fine pattern is formed on all edges. FIG. 19 is an explanatory view showing a phase shift mask including a phase shift pattern 1500 for forming a fine pattern 1100 to be formed in (a) and a fine pattern shown in (a) by a shifter edge method in (b). (C) to (b)
Shows a pattern 15 formed using the mask shown by. It can be seen that an unnecessary pattern is formed at the edge of the shifter.
As a method of not generating the unnecessary pattern, a method using a multi-stage phase shift film and a multiple transfer method using a plurality of masks have been proposed.

[0006] The technique using a multi-stage phase shift film means that a plurality of phase shift films having different phase differences are used, and a portion requiring a fine pattern has a phase difference of π on both sides thereof, but a portion requiring no pattern formation. Is a method in which a pattern is formed by setting the phase difference on both sides to, for example, 60 degrees. This method uses 1
A desired pattern can be formed with a single phase shift mask,
It is necessary to form a multi-stage phase shift film, which has a fatal drawback that mask fabrication is difficult.

FIGS. 20 (a), 20 (b) and 20 (c) are views for explaining a multiple transfer method using two masks, and FIG. 19 (a) is a pattern to be formed. FIG. 20A shows the same mask 1 with a phase shift film as in FIG.
(B) shows the mask 2 composed only of the light shield. (C) shows a state in which these two masks are overlapped to form a pattern. When exposure is performed using two masks and development is performed, only the portions shielded by both masks are patterned as a final light-shielding portion. (C)
The pattern shown by the broken line corresponds to this. That is, a fine pattern is formed on the boundary side where the phase difference of the mask 1 becomes π, and unnecessary portions are exposed and removed by the mask 2. This multiple transfer method requires only one type of phase shift film, which is equivalent to π in phase difference of transmitted light, and has an advantage that a mask can be easily manufactured.

[0008]

Conventionally, as a method of removing unnecessary patterns using the multiple transfer method, it has been possible to manually arrange a shifter or a light shielding portion for each pattern, but it has been complicated. No method has been proposed for automatically generating and arranging those patterns when they are mixed. Further, in the actual pattern forming process, when an attempt is made to form a pattern of one layer by two exposures using different masks, an error in alignment between the two exposures cannot be avoided. FIG. 21 shows an example of a deformation that a pattern undergoes when a relative displacement occurs in two exposures. FIG. 21A shows a pattern 8 to be formed, and FIG. 21B shows a first mask (M
1) is composed of a light-shielding portion 13 and a phase shift pattern 1500, and (c) shows the pattern 10 obtained as a latent image by exposure using M1 with no displacement of the second mask (M2). Shows the state of being overlapped. Since the latent image existing in the area corresponding to the light transmitting portion 14 of M2 is removed, the target pattern shown in FIG. (D)
Indicates a case where the image is transferred while M2 is displaced upward with respect to M1. The light transmitting portion on M2 for removing the unnecessary pattern formed by M1 overlaps the latent image of the pattern to be formed, and the normal pattern portions on both sides of the fine pattern undergo deformation to lose part, The pattern is also constricted.

The present invention has been made in view of the above points, and has as its object to form a circuit pattern of an LSI or the like using a phase shift method, and to use it in forming a single layer pattern using a plurality of masks. The present invention provides a method of creating a mask pattern and a mask for reducing the influence of an alignment error between a plurality of masks, and further provides a shifter which adds a part of the pattern limitation to an LSI or the like to be applied. Another object of the present invention is to provide a method for automatically generating a mask pattern including the following patterns.

[0010]

First, terms used herein will be explained. Of the patterns to be formed, narrow patterns formed using shifter edges are fine patterns 110.
0, a pattern formed by transferring a light-shielding body existing on a mask in the conventional manner is called a normal pattern 1200. Pattern to form,
That is, the given pattern 1000 is always a fine pattern 1100
Alternatively, it is classified into a normal pattern 1200, for example, a width of 0.5
Patterns of μm or more can be classified as normal patterns 1200, and patterns smaller than 0.5 μm can be classified as fine patterns 1100. In addition, the length of the fine pattern 1100 is sufficient for its width,
For example, if it is longer than 4 times,
The axial direction and a direction perpendicular to the axial direction will be referred to as a width direction . Further, in order to form a fine pattern using the shifter edge method, the phase is approximately π on both sides of the fine pattern.
Requires a region through which distant light can pass, and specifies the phase for that region
It is called region R1 or R region 1. A region 4 where a pattern is to be formed, for example, a chip region other than the R region is called an S region 3. The entire pattern forming region 4 includes an R region 1 and an S region 3, and the R region 1 and the S region 3 do not overlap. The boundary between the R region and the S region is defined as the RS boundary.
Call it 2. These are shown in FIG. 22, FIG. 23, and FIG. The fine pattern 1100 is originally a wide pattern.
In order to avoid complicating the drawing, it is described as one line segment.

First, the present invention will be schematically described. Although the multiple transfer method is used, here, the exposure is limited to two times using two masks. The first (M1) is a mask having a phase shift film (shifter), and the second (M2) is a mask having a shifter. No mask. That is, in the first exposure, a fine pattern using a shifter edge is formed, and in the second exposure, a necessary portion of the formed fine pattern is protected, and an unnecessary portion is exposed and removed.

Here, the following three prerequisites are introduced to facilitate explanation and the like. (Precondition 1) All the fine patterns 1100 are parallel to the X axis. (Precondition 2) The distance between the center lines of the fine patterns 1100 that are adjacent to each other is equal to a constant value p or an integer multiple of p.

(Precondition 3) The line width of the fine pattern 1100 is 1
It is assumed that only the types are used, and the line width d is equal to the width of the pattern formed by the light-shielding portions formed when the shifters having the width p are arranged at the interval p. The pattern of FIG. 22 satisfies these conditions. Now, one of the objects of the present invention is to obtain M1 from a given pattern.
Is to automatically generate a shifter pattern and a light shielding pattern for the M2 and a light shielding pattern for the M2. From the given pattern 1000, the phase designation area R1 is calculated,
The normal pattern 1200 includes a pattern 1300 existing in the R region,
It is separated from the pattern 1400 existing outside the region, that is, in the S region.

When the pattern 1000 shown in FIG. 22 is given, the phase designation area R 1 is, for example, as shown in FIGS.
The area shown in FIG. Although omitted in this figure, a phase shift film that generates a phase difference π in either case is:
They exist periodically as shown in FIG. FIG. 23, FIG.
In both cases, the R region 1 is defined at the upper part of the uppermost fine pattern and at the lower part of the lowermost fine pattern from the center line of the fine pattern to a region at a distance p from the center line.

FIG. 23 shows only the light transmitting portion required for forming the fine pattern 1100 as the R region 1.
1 is an area of a complicated shape corresponding to the fine pattern 1100. On the other hand, in FIG. 24, the R region 1 is a rectangular region, and its boundary in the X direction is determined so as to cover the widest region from among the boundaries required for forming the fine pattern 1100. Here, a rectangular area whose calculation is easy is adopted as the R area 1. A specific procedure for obtaining the R region will be described in Examples.

If the R region 1 is found, the given pattern 10
It is easy to separate 00 into a pattern 1300 existing in the R region and a pattern 1400 existing in the S region. As a result, the first mask M1 has a planar structure as shown in FIG. 26, forms a pattern including a fine pattern in the R region 1 (only the fine pattern in FIG. 26), and protects the S region 3 by shielding light. I do. In FIG. 26, a region where dots are scattered is a shifter 1500, and a fine pattern 1100 is formed at the edge thereof. An area 1700 covered by a diagonally downward-sloping line is a light-shielding portion. Meanwhile, 2
In the second mask M2, the normal pattern of the S region 3 is formed as shown in FIG. 27, the R region 1 is protected by shielding light, and the unnecessary fine pattern formed by M1 is exposed and removed. The area covered by the downward-sloping oblique lines is the light-shielding portion.

As described above, the introduction of the R region 1 makes it possible to separate the region where the fine pattern is to be formed from the other region, and to form the pattern in each region independently. Is automatically generated from a given pattern 1100. Although FIGS. 26 and 27 still have drawbacks as they are, the introduction of the R region 1 plays a large role in the automatic generation of patterns even in the improved method described below.

Another object of the present invention is to make it possible to transfer a pattern as faithfully as possible even when there is a relative displacement between a plurality of masks. FIG.
6. If M1 and M2 shown in FIG. 27 can be transferred without error, the intended given pattern 1100 can be formed.
Actually, there is a relative displacement between M1 and M2. That is, RS boundary 2 of M1 does not overlap with RS boundary 2 of M2. As a result, for example, considering the case where M2 is transferred with its position shifted to the upper right with respect to M1, the upper side and the right side of the R region 1 of M2 overlap with the S region of M1,
Both M1 and M2 are shielded from light, are not exposed, and form unnecessary patterns. This indicates that the portion to be exposed near the RS boundary 2 must be exposed from both M1 and M2.

In M2 of FIG. 27, in order to protect the latent image of the necessary fine pattern in the R area 1, the entire area except for the unnecessary latent image is made a light shielding portion. Only the pattern portion in the R region. Therefore, in order to prevent the pattern portion from being exposed even if the mask is misaligned, only the region where the pattern portion in the R region is slightly enlarged by an amount that allows for the misalignment is set as the light shielding portion, and the other portions are set as the transmission portions. . In this way, even if a misalignment occurs in a portion where no pattern exists at the boundary of the region R (for example, the upper side of the R region in FIG. 27), an unexposed pattern does not occur.

Further, the present invention will be described with respect to a method for coping with a mask misalignment when a pattern exists at the RS boundary portion 2. First, the case where the end of the fine pattern is at the RS boundary 2 will be described. The portion 51 encircled in FIG. 24 corresponds to this. This part is enlarged to show the latent image 10 formed by M1 and the part protected by M2.
FIG. 28A shows a case where there is no misalignment of the mask with respect to the overlap with 12 and the resulting pattern 15.
A case where M2 is shifted to the left is shown in FIG. In FIG. 28A where there is no misalignment, the pattern is formed as expected, but in the case where there is misalignment, the tip of the obtained fine pattern 15 bulges. In order to solve this phenomenon, when determining the right and left ends of the R region 1, as shown in FIG. 25, a region which is extended by α on both sides in the axial direction from the section where the fine pattern exists is defined as the R region 1. Also, M1
Protection pattern in M2 against fine pattern formed in step 25
In the case of 00, the width is increased in the width direction of the fine pattern 1100 in consideration of misalignment, but is not increased in the axial direction. FIG. 29 shows a graphic obtained by performing the above-described processing. FIG. 29A shows a fine pattern 1100 to be formed.
(B) and (c) show the overlap between the latent image 10 formed by M1 and the portion 12 protected by M2, and FIG. A faithful pattern can be transferred regardless of misalignment. Fine pattern 1100
In the formation of the pattern R, the region R1 is expanded by α on both sides in the axial direction, so that even when the mask is misaligned, the pattern is not deformed, and the position of the fine pattern in the Y direction is M
According to 1, the position in the X-axis direction can be determined by M2.

Next, by expanding the region R by α in the axial direction, the normal pattern 1200 exists at the RS boundary portion in FIG.
The case (a) will be described. First, a description will be given of how the pattern is deformed in the conventional case in which no countermeasures are taken against misalignment. FIGS. 30 (b) and (c)
The patterns of M1 and M2 are shown, respectively. In M2, in R area
It is enlarged and thickened evenly for the normal pattern of 1.
This is to protect the latent image formed by M1 even if the mask is misaligned, as described above. If the pattern is transferred without any alignment error, the pattern shown in FIG. 30A is obtained.
As shown in FIG. 1D, the thickened portion in the R region 1 intrudes into the S region 3 to form a projection in the pattern 15, thereby hindering other adjacent patterns.

In the present invention, in order to avoid such a problem, in the width direction (y
Direction), and a process of increasing the thickness by one side δ only in the axial direction (x direction) in M2. FIG.
FIGS. 31A and 31B show patterns of M1 and M2 to which the present invention is applied with respect to the pattern of FIG. When M1 and M2 are transferred without alignment error, when M2 is shifted to the upper right,
FIG. 31C shows a case where M2 is shifted to the lower left.
(D) and (e). In any case, the formed pattern shape does not cause constriction or protrusion, and almost exactly reflects the design pattern. This is because the normal pattern in the R region is separated and thickened in the vertical direction in M1 and in the left and right direction in M2. Therefore, even if misalignment occurs, regardless of the inside and outside of the region, the left and right sides of the normal pattern in the region ( Only the right side (in the example of FIG. 30) is defined by M1, and the remaining sides are defined by M2.

[0023]

An exposure using a mask including a pattern for shifting the phase so that the phase difference of light passing through both sides of the fine pattern to be formed is approximately π, and an unnecessary pattern for erasing the resulting pattern are eliminated. In pattern formation in which a pattern of one layer is formed by a plurality of exposures including exposure, a region where exposure is commonly blocked by a plurality of masks remains as a pattern. By dividing the exposure region into a region having a fine pattern required by the phase shifter and a region not having the fine pattern, one is provided with a shifter, and in the region having the fine pattern, a mask shifter pattern having an edge corresponding to the fine pattern and another are provided. A mask having a mask pattern corresponding to the pattern, a mask pattern that does not have a fine pattern is used as a mask pattern for shielding, and another mask is a normal mask that does not include a shifter and forms a pattern existing in a region where no shifter is provided. A shifter pattern and a mask pattern in two masks can be easily and automatically generated such that a pattern and a pattern belonging to a region where a shifter is formed have a mask pattern for removing unnecessary patterns. Further, since the pattern to be formed is a portion which is shielded by two masks, the mask pattern is thickened in one axis direction in one mask, and the mask pattern is masked in the direction perpendicular to this axis in another mask. By increasing the thickness of the pattern, the influence of misregistration between masks on the formed pattern can be significantly reduced, and the alignment margin can be increased.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A case where the present invention is applied to a gate layer which requires formation of a fine pattern in an LSI will be described. FIG. 1 shows the processing flow of the embodiment, and FIGS. 2 to 6 show the given input pattern data and the pattern data of the mask making operation created by using the present invention. FIG. 2 shows that input pattern data 1000 including a fine pattern satisfies the three preconditions described in “Means for Solving the Problem”. 3 and FIG.
The pattern shown in FIG. 2 is formed by using the masks M1 and M2. M1 has a shifter, and M2 has no shifter. As described above, the normal pattern is separated into the R region and the S region and the pattern is generated. Therefore, in order to manufacture the M1 mask, the shifter pattern data 1500, the R region pattern data 1600, and the S region protection pattern are required. Three types of pattern data of data 1700 are required. Similarly, to manufacture the mask of M2, the pattern data 2500 for protecting the fine pattern created by the shifter of M1, the R area pattern data 2600, and S
Three types of pattern data of area pattern data 2700 are required. 3 and 4 show these patterns.

FIG. 1 shows that, based on input pattern data 1000, M1
And three patterns for M2 and a total of six patterns 1500 and 16 respectively.
The flow of processing for creating 00, 1700, 2500, 2600, and 2700 is shown. The squares with rounded corners represent data processing, and the input and output are pattern data shown by ordinary squares. Nine types of data processing are required from 101 to 109, and the details of each processing and a specific implementation method are described below. Note that the realization method described here is merely one method, and the present invention is not limited to this.

The data processing described below handles mask data which is two-dimensional graphic data. The contour extraction for two-dimensional graphic data, thickening or thinning of a uniform width called resizing, black-and-white inversion processing for a chip area, or the intersection set operation when there are a plurality of input data are L
SI mask pattern data design rule check (DRC,
Design rule check) or data processing for drawing a mask using an electron beam, which is already generally performed. Therefore, here, the description of the specific method of the process is omitted if the process is generally performed. For specific methods, refer to, for example, “Inspection of LSI Mask Pattern Data with Advanced Automation”, Nikkei Electronics, 1980.4.28., Pp. 90-107.

Fine pattern extraction processing 101 A given pattern 1000 is divided into a fine pattern 1100 and a normal pattern.
Separated into 1200. If the width d of the fine pattern is known,
It is determined by the following processing. (1) The given pattern 1000 is reduced by the distance d / 2. (2) Extract the contour. (3) The result of (2) is increased by the distance d / 2. The result obtained is usually a pattern 1200. (4) If the normal pattern 1200 is subtracted from the given pattern 1000, the obtained result is a fine pattern 1100 having a width d.

FIG. 2 is taken as a given pattern 1000, and figures 1010-1 to 1010-4 resulting from contour extraction with the distance d / 2 narrowed together with the pattern 1000 given in FIG. 7 are shown in FIG. Shape
Normal patterns 1200-1 to 1200-4 obtained by increasing the distance d / 2 from 1010-1 to 1010-4, and fine patterns obtained based on these
1100-1 to 1100-4 are shown. Normally, the line width of the fine pattern is known, but if unknown, the above-described processing is first performed with the maximum size considered as the fine pattern, for example, 0.49 μm as d. The obtained fine pattern 1100 has a width of 0.
All patterns of 49 μm or less are included. Next, the same processing is executed with d slightly reduced. This processing is sequentially performed by d
Is repeated with a slight reduction, the fine pattern disappears someday, resulting in a zero set. D immediately before that corresponds to the width of the fine pattern.

Center Line Extraction Process This is a process for obtaining the center line segment 1250 of the 102 fine pattern. According to the premise, the line width of the fine pattern is one type, and since the line width is known from the fine pattern extraction processing 101, the processing is easy. Fine pattern 1100-1 to 1100-3 and center line segment 1250-1 to 1250
FIG. 9 shows the relationship of -4.

R, S area determination processing This is processing for determining the R area 1 and the S area 3 based on the center line data 1250 of the 103 fine pattern. It is assumed that there are n pieces of center line segment data in all, and the interval p between adjacent center line segment data is known. This is because even unknown cases can be known by analyzing the data. (1) In the width direction of the center line segment data 1250, a rectangle having a width of 2L is formed by increasing the distance by L on both sides. (2) Perform contour extraction processing based on the n rectangles obtained in (1), number the polygons formed by connecting the rectangles, and assign the polygons to each rectangle constituting the polygon. Assign a number. (3) Sort n pieces of rectangular data by polygon number,
A rectangle that forms the same polygon is extracted, and a minimum rectangular area that can cover the polygon area is obtained. (4) The rectangular area obtained in (3) is extended by α to both sides in the axial direction to obtain a new rectangular area. If the new rectangular areas overlap or touch each other, they are assigned the same polygon number, and the process returns to (3). (5) The width direction of the rectangular area obtained in (4) is reduced by L on both sides, and then increased by p. The obtained region is the R region 1. (6) The S region 3 is obtained by subtracting the R region 1 obtained in (5) from the pattern formation region.

This is a process in which center lines existing near each other are collected to obtain an R region. The description of the processing in this embodiment is shown in FIGS. Center line 1250-1 to 1250-4
If each is a rectangle having a width of 2L, they are all connected to form one polygon. Since there is no other polygon, the R region 1 in FIG. 11 is obtained by the above processing (3), (4), and (5). (1)
If the interval between the adjacent center line segment data is 2L or less, the two rectangles can be connected to form one polygon. Therefore, if 2p <2L <3p, as shown in FIG. 10, if the interval between adjacent center line segment data is up to 2p, one R region can be regarded as another R region if it is 3p apart. .

The normal pattern extraction process 104 in the R region and
S area normal pattern extraction processing 105 R area normal pattern extraction processing 104 includes R area determination processing
The normal pattern data in the R region is easily obtained by the intersection set operation of the R region data and the normal pattern data obtained in 103. The same applies to the normal pattern extraction processing 105 in the S region. FIG. 12 shows normal patterns 1300-1 to 1300-4 in the R region,
And normal patterns 1400-1 to 1400-2 in the S region.

Thickening process in the width direction 106 The normal pattern 1300 in the R area is δ
Fat. FIG. 13 shows the results. Normal pattern in R area 13
00-1 to 1300-4 are R region patterns for M1 1600-1 to 1600
It becomes -4. δ is an amount determined by the relative alignment accuracy of the two masks, and may be, for example, 0.2 μm.

Axial thickening process 107 The normal pattern 1300 in the R region is thickened in the axial direction by δ in the R region. FIG. 14 shows the results. Normal pattern in R area
1300-1 to 1300-4 are R region patterns for M2 2600-1 to 26
00-4. In the figure, the S area patterns 2700-1 and 2700 for M2 are shown.
-2 is also shown. δ has the same concept as the width-direction thickening process 106 and has the same value.

Shifter pattern generation processing 108 The R area data and the corresponding center line segment data are input to generate a shifter pattern. If the center line segment data 1250 is sorted in the width direction (for example, from top to bottom in the y direction in FIG. 15) for each R region, the center line segment data 1250 is sorted.
May be sequentially taken out and assigned to a region without a shifter and a region with a shifter. FIG. 15 shows the result of the assignment. Fine pattern protection processing 109 The center line segment data 1250 is increased by ε in the width direction to create a fine pattern protection pattern 2500 for M2. FIG. 16 shows an embodiment. The value of ε may be the same as the value δ used in the width-thickening process 106 in consideration of the alignment between masks, but ε> δ in the sense that the fine pattern serving as the gate electrode is completely shielded to improve the dimensional accuracy. And, for example, 03. A value of μm is used.

By executing the nine types of arithmetic processing described above, it is possible to automatically generate patterns M1 and M2. M1 is, as shown in FIG. 3, a shifter pattern 1500-
1, 1500-2, an S area protection pattern 1700, and R area normal patterns 1600-1 to 1600-4. M2
As shown in FIG. 4, the pattern is composed using fine pattern protection patterns 2500-1 to 2500-4, R region patterns 2600-1 to 2600-4, and S region patterns 2700-1 and 2700-2.

FIG. 5 shows a case where two masks overlap without error. If a positive resist in which the portion exposed to light is dissolved by development is used, the intersection of the latent image area 10 formed by M1 and the area 12 protected by the light-shielding portion of M2 becomes unexposed, and the pattern becomes It is formed. It can be seen that the pattern of FIG. 2 is formed. FIG. 6 shows a case in which M2 is transferred to the lower left with respect to M1. Although the relative position of the fine pattern and the normal pattern are shifted, there is no fatal defect in the pattern shape. In the gate pattern, the overlay accuracy with another layer, for example, a contact hole layer, becomes a problem, but in this regard, there is a significant difference between the normal method using one mask and the method according to the present invention. Absent.

Now, on the premise of the gate layer pattern using the standard cell design method, a supplementary explanation will be given on the restrictions of the three preconditions described in Means for Solving the Problems. Precondition 1 to be parallel
It goes without saying that even if all are parallel to the Y-axis, and furthermore, if there are a plurality of first regions whose phases are designated, for each region, all the fine patterns included in that region May be parallel to the X axis or the Y axis.

The condition 2 that the interval between the center line portions of the fine pattern is p or an integer multiple of p can be considered to be satisfied in the standard cell system. Regarding the third condition for the line width of the fine pattern, if the shifter edge method described with reference to FIG. 18 is applied, there is no restriction on the line width.
For the type of line width, fine pattern extraction processing 101, determination of the width of the light shield required for forming the fine pattern, for example, the light shield required for the shifter edge portion, pattern generation processing, and protection of the fine pattern by M2 Considering that the pattern width in the process 109 is plural, it is not necessary to limit to one type. Actually, the gate pattern of CMOS is P-type MOS.
Although the width of the gate pattern may be different between the transistor and the N-type MOS transistor, it is easy to cope from the above.

Therefore, the method of the present invention can be applied to the formation of a gate layer pattern using the standard cell design method, and further applicable to the formation of a pattern in which fine patterns are arranged at equal intervals. High in nature.

[0041]

Since the present invention is configured as described above, it has the following effects.
By dividing the exposure region into regions having a fine pattern required by the phase shifter and regions not having the fine pattern, the region forming the fine pattern and the other region are separated, and the pattern in each region is formed independently. And it is possible to automatically generate a pattern in each area including the shifter from a given pattern to be formed.

When the normal pattern in the region where the phase shifter is required is separated and thickened in the vertical direction in the mask M1 and in the horizontal direction in the mask M2, the M2 is transferred with a misalignment with respect to the M1. However, the shape of the formed pattern does not cause constriction or protrusion, and the influence of the displacement can be reduced so that a pattern that reflects the design pattern almost exactly can be formed.

[Brief description of the drawings]

FIG. 1 is a flow of a process for automatically generating a pattern for a phase shift mask including a pattern of a phase shift film from a given mask pattern.

FIG. 2 is a pattern to be formed.

FIG. 3 is a plan view of a first mask for forming the pattern of FIG. 2 when the present invention is implemented.

FIG. 4 is a flat layout view of a second mask for forming the pattern of FIG. 2 when the present invention is implemented.

FIG. 5 is an illustration of a pattern when two masks of FIGS. 3 and 4 are overlapped and transferred without displacement.

FIG. 6 illustrates a pattern when the mask of FIG. 4 is transferred with a positional shift in a lower left direction with respect to the mask of FIG. 3;

FIG. 7 illustrates a fine pattern extraction process 101;

FIG. 8 illustrates a fine pattern extraction process 101.

FIG. 9 is an illustration of a center line segment extraction process 102 of a fine pattern.

FIG. 10 illustrates an R / S area determination process 103;

FIG. 11 illustrates an R / S area determination process 103;

FIG. 12 illustrates a normal pattern extraction process 104 in an R region and a description of a normal pattern extraction process 105 in an S region.

FIG. 13 is an illustration of a process 106 for making the normal pattern in the R region thicker in the width direction.

FIG. 14 is a diagram illustrating an axial thickening process 107 for a normal pattern in an R region.

[15] Shifutapatan description of generation processing 108.

FIG. 16 is an illustration of a fine pattern protection process.

17A and 17B are explanatory diagrams of a phase shift method using an edge of a phase shift film. (A) Cross-sectional view of a phase shift mask. (B) Complex amplitude of light transmitted through the mask of (a). phase shift mask cross-sectional view for forming the intensity distribution of the transmitted light (d) is a fine pattern using the edge of the phase shift film at a period p

FIG. 18 illustrates a method for controlling a pattern width by forming a fine pattern using an edge of a phase shift film. (A) a cross-sectional view of a mask in which a light-shielding body is provided at an edge of a phase shift film,
(B) A plan view of a mask in which the edge of the phase shift film is formed into a folded shape having a small period.

FIG. 19 illustrates an example in which an unnecessary pattern is formed by using a phase shift film.

FIG. 20 is a view for explaining a method of removing unnecessary patterns shown in FIG. 19 using two masks .

FIG. 21 illustrates an adverse effect on a formed pattern when there is a misalignment between two masks.

FIG. 22 is a form form you want pattern

FIG. 23 shows an example of forming an R region .

FIG. 24 is an example in which an R region is formed in a rectangular shape .

FIG. 25 shows an example of an R region in which the region is extended by a distance α in the axial direction of the fine pattern in determining the R region in order to cope with misalignment between masks.

FIG. 26 shows an example of a mask pattern of a mask provided with a shifter when the misalignment between the masks is not dealt with.

FIG. 27 shows an example of a mask pattern of a mask without a shifter when the misalignment between the masks is not dealt with.

[Figure 28] Pattern formation example when there is Re without adjusting the inter-mask.

FIG. 29 is an example of pattern formation when there is a mask misalignment when the distance α area is widened in determining the R area.

FIG. 30 is a diagram illustrating a case where misalignment occurs when the misalignment between masks is not sufficiently dealt with (a) a pattern to be formed (b) a plan view of the mask 1 (c)
(D) Pattern formed when misalignment occurs

FIGS. 31A and 31B are explanatory diagrams in a case where misalignment between masks is dealt with by the present invention. FIG. 31A is a plan view of a pattern of a mask 1 FIG. 31B is a plan view of a pattern of a mask 2 FIG. ) Forming pattern when M2 is transferred to the upper right with respect to M1. (E) Forming pattern when M2 is transferred to the lower left with respect to M1.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-4-76551 (JP, A) JP-A-4-337732 (JP, A) JP-A-4-186244 (JP, A) 369823 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G03F 1/08 H01L 21/027

Claims (2)

(57) [Claims] An exposure using a mask including a pattern for shifting the phase so that the phase difference of light passing through both sides of a fine pattern to be formed is approximately π, and an unnecessary pattern generated by the exposure is erased. In a pattern formation for forming a pattern of one layer by a plurality of exposures including a light exposure, a rectangle required for transmitting light having a phase difference on both sides of the fine pattern to form the fine pattern A first step of setting a first region, and a second step of separating all patterns to be formed or all patterns except the fine pattern into and out of the first region. Pattern data processing method. 2. The length of a fine pattern formed by setting the phase difference of light passing through both sides to approximately π in determining the first region is at least longer than the length of a fine line finally formed. The boundary of the first region is set so as to be longer on both sides in the axial direction of the fine pattern by a distance equivalent to an overlay error caused by mutual overlay between a plurality of masks. 2. The mask pattern data processing method according to 1.