JP3260851B2 - Exposure mask and manufacturing method thereof - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an exposure mask,
In particular, the present invention relates to an exposure mask having a shape different from a desired pattern in order to form a desired pattern.

When the wavelength of an exposure beam approaches the resolution of a pattern to be imaged, diffraction and interference of the exposure beam become problems. Therefore, if a desired device pattern is used as a mask pattern as it is, a desired device pattern cannot be obtained.

[0003]

2. Description of the Related Art In semiconductor integrated circuit devices and the like, device patterns tend to be increasingly miniaturized. In light exposure using light, diffraction and interference of light become a problem along with miniaturization of device patterns.

In exposure using an optical system, the resolution is given by a function such as a numerical aperture (NA) and a wavelength. In order to increase the resolution, it is desirable to increase the numerical aperture and shorten the wavelength. The increase in the numerical aperture is limited by the depth of focus, the design of the optical system, and the like. Shortening of the wavelength has progressed from the emission line of a mercury lamp to excimer laser light, and is approaching the limit due to restrictions on optical materials and the like.

As a technique for exposing a fine device pattern even when light diffraction or interference occurs, there is a method using a mask pattern in which a device pattern to be imaged is changed. One is a phase shifter that gives a phase difference to light, and the other is a method of designing a mask pattern that leaves a device pattern and obtains an optimum light intensity distribution on an imaging plane. In addition, both are not mutually opposite and can be compatible.

[0006]

The method of obtaining the optimum light intensity distribution on the image plane largely depends on empirical rules. Even if a device pattern is designed with much time and effort, the obtained result is not obtained. It often falls into a local solution.

SUMMARY OF THE INVENTION It is an object of the present invention to manufacture an exposure mask having a shape different from a desired pattern, which can provide an exposure beam intensity distribution that can approximate a desired pattern with high accuracy on an image plane. Is to provide a way.

Another object of the present invention is to provide a method of manufacturing an exposure mask which is less affected by empirical rules and is less likely to fall into a local solution.

[0009]

[0010]

According to the method of manufacturing an exposure mask of the present invention, a desired design pattern is formed by providing a uniform exposure beam transmission characteristic to each of a plurality of cell regions. Preparing a departure pattern that approximates, and for the departure pattern, based on a random number,
Providing a change (mutation) in the exposure beam transmission characteristic for any of the cell regions to form a plurality of child patterns; and causing the plurality of child patterns to cross or mutate with other child patterns. A genetic algorithm step of evaluating the imaging characteristics of the obtained pattern to leave a predetermined number of grandchild patterns; a step of repeating the genetic algorithm step using the grandchild pattern as a child pattern; As a device pattern.

[0011]

An area for forming a pattern is divided into a large number of cell regions, and an optimum exposure beam transmission characteristic is selected for each cell region to form an optimum exposure beam intensity distribution on an image plane. A mask is easily obtained.

A desired design pattern is used as a starting pattern,
An optimal pattern can be obtained with high probability by creating a deformed pattern using a genetic algorithm including a mutation, evaluating the imaging characteristics, and converging the deformed pattern.

[0013]

FIG. 1 shows the basic structure of an exposure mask according to the present invention. The mask 1 is formed to include a number of cell regions 3. The cell region 3 has, for example, a rectangular shape as shown. An optimum exposure beam transmission characteristic is set for each cell region 3. For example, two states of blocking and transmission are set as the exposure beam transmission characteristics, and a blocking or transmission state is set for each cell region 3.

By structuring the mask with finite elements, optimization of the mask pattern can be rationally advanced. Each cell region 3 depends on a desired device pattern and an exposure beam wavelength to be used, but is preferably about 0.001 to 1 μm on a side.

Instead of blocking the transmission characteristics of the exposure beam of each cell region 3 between the two states of blocking and transmission, a plurality of states may be set as transmission states, and the phases may be different. According to such a design, a mask pattern including a phase difference, such as a phase shift reticle, can be formed.

Here, an example of a mask pattern according to the prior art will be described for comparison. FIG. 2 shows a conventional mask and its intensity distribution on the image plane. FIG.
(A) shows an optical mask having two patterns 15 and 16 at the center.

These patterns 15 and 16 have a shape proportional to a desired device pattern to be formed on the image plane. For example, it is a pattern obtained by enlarging the device pattern by a factor of 5 or 10. On the image plane, pattern 1
It is assumed that 5, 16 and the distance therebetween are of the same order as the wavelength of the light used.

In such a case, the light that passes through the mask and forms an image on the image plane is diffracted and interferes with each other. As a result of diffraction and interference, the light intensity distribution on the image
Deviates from the shapes of 5 and 16.

The mask 1 shown in FIG. 2A is 2.0 μm square, and the patterns 15 and 16 are 1.6 μm each.
m × 0.6 μm, NA = 0.5, wavelength WL =
An image is formed on an image plane through an optical system of 0.365 μm.

FIG. 2B shows an intensity distribution on the image plane when the optical system forms an image of the mask just on the image plane.
The curve in the figure represents a contour line of the light intensity distribution. FIG.
(A) and (B) similarly show images obtained when the mask image shown in FIG. 4 shows a light intensity distribution on a surface.

As apparent from FIGS. 2B, 3A and 3B, the corners of the patterns 15 and 16 on the mask 1 are rounded, and the contrast between the patterns 15 and 16 is reduced. Is declining. As the defocus advances, the contrast between the patterns 15 and 16 further decreases.

In a semiconductor device manufacturing process, when a resist mask is applied on the surface of a semiconductor wafer and light exposure is performed as shown in FIGS. 2 and 3, a desired pattern cannot be obtained, and the corners of the pattern are rounded. It can be seen that there is a high risk that adjacent patterns will be continuous.

FIG. 4 shows a mask according to an embodiment of the present invention. FIG. 4 (A) shows a large number of cell regions 3 as shown in FIG.
1 shows a mask in which light transmission patterns 5 and 6 are formed on a mask 1 having the following.

The mask 1 has a size of 2 μm × 2 μm on the image plane, as in the case of FIG. 2, and each cell region 3 corresponds to a rectangle of 0.1 μm × 0.1 μm on the image plane. Shall be. In each cell region 3, the light transmission characteristics are set uniformly. In the case shown, the cell regions arranged in the light transmission patterns 5 and 6 are in a light transmission state, and the cell regions outside these are in a light blocking state.

The mask shown in FIG. 4A is for exposing a rectangular pattern. Unlike a desired rectangle, a portion corresponding to each corner protrudes outward, and a portion corresponding to a long side has a protrusion in the middle.

FIGS. 4 (B), 5 (A) and 5 (B) show FIG.
(A) mask, numerical aperture NA = 0.5, wavelength for exposure WL
11 is a graph showing a light intensity distribution formed on an image plane by an optical system of = 0.365 μm. FIG. 4B shows the mask 1.
FIG. 5A shows a case where the image of FIG.
(B) shows a case where the image of the mask 1 is formed at 0.25 μm and 0.5 μm above and below the image plane.

It is clear that the light intensity distribution obtained in this embodiment is closer to a desired rectangular pattern than in the case of FIGS. 2 and 3 having the same pattern as the design pattern on the mask. Also, many contour lines are formed in the region between the adjacent patterns 5 and 6, indicating that sufficient contrast can be obtained.

Next, a process of designing a mask pattern capable of exposing a desired pattern on an image plane using a mask having a basic structure as shown in FIG. 1 will be described below.

FIG. 6 is a flowchart of an example of a mask design process. First, in step S1, the mask pattern is divided into meshes, and the arrangement of the divided cell regions is regarded as an individual gene. For example, as shown in FIG. 1, it is assumed that the mask 1 is divided into a large number of cell regions 3, and each of the divided cell regions 3 is a gene.
The sequence of the gene determines the genetic characteristics of the individual.

A light intensity distribution is formed on the image plane by the mask 1, and this light intensity distribution is assumed to be a character of an individual expressed based on a gene. The individual includes the mask 1 and the light intensity distribution formed by the mask 1.

Next, in step S2, the initial sequence is set as a parent, and m children mutated from the parent are generated by asexual reproduction. For example, when the design patterns are patterns 15 and 16 shown in FIG. 2A, this design pattern is used as an initial arrangement as it is. Mutant children
For example, in the arrangement of FIG. 1, the horizontal direction is x, the vertical direction is y, coordinates (x, y) are designated by random numbers, and the state of the cell area 3 corresponding to these coordinates is inverted. The mutation is not limited to this method, but may be any mutation that can cause any change in any cell region in the mask.

After m individuals are generated in this way, m grandchildren are generated for m children in step S3. In this case, since a plurality of individuals already exist, the offspring to be generated can be generated by mating or mutation.

In the case of mating, a mating partner is selected, a logical operation such as an AND operation, an OR operation, or a knot operation is performed between the corresponding cell area of the individual and the counterpart individual to generate a newly generated mask cell. The light transmission state of the region is determined.

Whether to perform crossing or mutation can be determined by random numbers. For example, one of three types of AND by mating, OR by mating, and mutation is selected by random numbers, and a cell area of a new mask is determined.

In step S4, if the number of individuals is increased without limit, an extremely long time is required for arithmetic processing. Therefore, individuals having excellent genetic traits are selected from the large number of individuals generated, and the rest are selected. Let them cull.

That is, m × m generated from m children
The traits of the grandchild individuals, that is, the light intensity distribution on the image plane are examined by simulation, and the external constraints,
That is, as compared with an ideal light intensity distribution corresponding to a desired pattern, only m excellent light intensity distributions are present and the rest are eliminated.

In exposing a photoresist film in a semiconductor device or the like, the photoresist film is not necessarily formed on a plane. When exposing a photoresist film on a stepped surface, the light intensity distribution at the defocus position is also important. If necessary, the light intensity distribution at the focus and defocus positions is calculated and compared with the ideal light intensity distribution.

The determination of the superiority or inferiority of a trait is made, for example, by sampling a region where the light intensity distribution is the most important, integrating the deviation from the ideal light intensity distribution with weight, and setting the obtained value as energy E. It is assumed that the smaller the E, the better the trait. In this way, m individuals with excellent traits are genetically generated from the m individuals.

In step S5, the above steps S3 and S4 are repeated to converge until surviving individuals have substantially uniform genes. The sequence of the best gene obtained in this way is adopted as the best individual.

By such a procedure, for example, FIG.
A device pattern as shown in FIG. 4A can be obtained from a design pattern as shown in FIG. In addition,
In order to quickly converge the process shown in FIG. 6, it is preferable to incorporate a miscarriage process in step S3 of FIG.

FIG. 7 shows the process of miscarriage using temperature parameters. The temperature parameter T corresponds to a threshold for determining whether the trait is suitable for surviving. When a new individual is generated, step B1 is performed by first identifying the mating partner or performing mutation.
Of conception. Although a new individual is specified by conception, it is investigated whether the gene of the offspring generated in step B2 before delivery is excellent. That is, the light intensity distribution (trait) created by the individual that has occurred is determined, the deviation E from the ideal trait is determined, and the degree of the trait is expressed as exp.
Expressed by (-E / T).

Here, the criterion is a uniform random number r (0 <r <
Determined by 1). That is, if exp (−E / T)> r, it is determined that the child has a trait that can withstand childbirth.

When exp (−E / T) <or = r, it is determined that the trait cannot be tolerated. If it is determined that the trait is endurable, the process proceeds to step B3, and the new individual is counted as a child.

If it is determined that the trait of the new individual cannot withstand childbirth, the process proceeds to step B4, where the new individual is selected. As described above, by judging the trait of the newly generated individual, individuals with poor traits are selected (miscarriage), and only individuals with excellent traits are generated as new individuals (delivery).

After steps B3 and B4, it is determined in step B5 whether or not the number of generated children is m. When the number of offspring does not reach m, the process returns to step B1 according to the arrow of no to perform a new conception. When the number of children reaches m, the process proceeds to the next process according to the arrow of yes.

As described above, in step S3 of FIG. 6, in the process of producing m grandchildren from m children, the traits of new individuals that occur are determined, and individuals with inferior traits are selected. Thus, all individuals proceeding to the next step S4 can be made excellent with uniform traits. Such selection promotes convergence according to the flowchart of FIG.

It should be noted that in the process of judging the superiority of the gene shown in step B2 in FIG. 7, the process of finding the deviation E from the ideal trait is not necessarily performed on the entire surface of the mask.

For example, in the case of a pattern of a contank hole, it is important that an opening is reliably formed at the center of the pattern, and there is no problem even if the peripheral shape is slightly deformed. In such a case, the character to be generated may be determined at the center of the opening pattern.

Conversely, in the case of a pattern of a capacitor electrode of a semiconductor memory device, the area of the capacitor electrode has an important meaning. In such a case, sampling can be performed in the outer peripheral portion of the pattern to determine the superiority or inferiority of the trait.

In the step B2, it is preferable to set the temperature parameter T to a relatively high value at the beginning of the process and to gradually reduce the value as the calculation proceeds. When the temperature parameter T is set low, individuals with relatively large deviations are eliminated, and convergence is promoted. If the temperature parameter is set low from the beginning of the process, it becomes difficult to escape from the local solution when it falls into a locally large solution.

FIG. 8 shows an example of a mask according to another embodiment of the present invention. FIG. 8A is a plan view showing a pattern on a mask, and FIG. 8B shows a light intensity distribution on an image plane at the time of focusing obtained by the mask of FIG. 8A.

FIGS. 9A and 9B are likewise shown in FIG.
2 shows the light intensity distribution formed on the image plane when the mask of (1) forms an image at 0.25 μm and 0.5 μm on the image plane. FIG. 8 (A)
The mask 1 has a thickness of 2.0 μm
× 2.0 μm, and each cell region 3 has a size of 0.05 μm.
It is a rectangle of m × 0.05 μm.

When observing the mask pattern shown in FIG. 8A, a light-shielding region is present at the center, and a light-transmitting region and a light-shielding region are intermingled and distributed in the peripheral portion. FIG. 8 (A)
The numerical aperture NA = 0.5 and the exposure wavelength WL
An image was formed on the image plane using an optical system of 0.365.

With such a configuration, the light intensity distribution actually obtained on the image plane is as shown in FIGS. 8 (B), 9 (A), and 9 (B).
As shown in (1), the central portion has a substantially uniform light intensity and a high contrast.

For comparison, FIGS. 10 and 11 show the case where the patterns corresponding to FIGS.
Shown in FIG. 10A shows a mask having a shape corresponding to a design pattern to be obtained on an image plane. Width 0.3
This is a shape in which a rectangle of 0.5 μm × 0.5 μm overlaps a line of μm. As described above, when a mask having the same shape as the design pattern is used, FIGS. 10B, 11A,
A light intensity distribution as shown in (B) is obtained on the image plane.

FIG. 10B shows the case where the image of the mask is formed on the image plane, and FIGS. 11A and 11B show that the image of the mask is 0.25 μm and 0 mm below or above the image plane. .50 μm
The case shown in FIG.

Although the mask of FIG. 10A has a rectangular pattern at the center, the obtained light intensity distribution is shown in FIG.
0 (B), FIG. 11 (A), and FIG. 11 (B), the corners are rounded, and the central region where the light intensity distribution is flat is narrow. In FIG. 11B, a region having a low light intensity appears instead of the central portion of the pattern.

As described above, a desired pattern as shown in FIG. 10A is different from that of FIG.
Exposure with higher precision can be performed by the mask (A).
As described above, the mask is divided into a large number of cell regions, and the light transmission characteristics are set for the cell regions. A pattern can be obtained with high probability.

The light exposure has been described above as an example, but is not limited to light as long as the exposure beam has a wave property. However, the present invention is particularly excellent in correcting the deformation of the imaging pattern due to the diffraction and interference of light by using light having a wavelength at which an optical system can be easily designed.

The cell area dividing the mask surface is not limited to a rectangle. Any shape may be used as long as the entire area of the mask can be equally divided.

[0061]

As described above, according to the present invention,
It becomes easy to obtain an exposure mask having a pattern different from the design pattern and capable of generating an optimum pattern close to the design pattern on the image plane.

[Brief description of the drawings]

FIG. 1 is a schematic plan view showing a basic configuration of an exposure mask of the present invention.

FIG. 2 shows a conventional technique. FIG. 2A is a plan view showing a planar shape of the mask, and FIG. 2B is a graph showing a light intensity distribution on an image plane.

FIG. 3 shows a conventional technique. FIGS. 3A and 3B are graphs showing the light intensity distribution on the image plane.

FIG. 4 shows an embodiment of the present invention. FIG. 4A is a plan view showing the planar shape of the mask, and FIG. 4B is a graph showing the light intensity distribution on the image plane.

FIG. 5 shows an embodiment of the present invention. FIG. 5 (A), (B)
5 is a graph showing a light intensity distribution obtained on an image plane by the mask of FIG.

FIG. 6 is a flowchart showing a mask design process.

FIG. 7 is a flowchart showing a miscarriage process using a temperature parameter.

FIG. 8 shows an embodiment of the present invention. FIG. 8A is a plan view showing the shape of the mask, and FIG. 8B is a graph showing a light intensity distribution on an image plane obtained by the mask of FIG. 8A.

FIG. 9 shows an embodiment of the present invention. FIG. 9 (A), (B)
9 is a graph showing a light intensity distribution obtained on an image plane by the mask of FIG.

FIG. 10 shows a conventional technique. FIG. 10A is a plan view showing the shape of the mask, and FIG. 10B is a graph showing the light intensity distribution obtained on the image plane by the mask of FIG.

FIG. 11 shows a conventional technique. FIG. 11 (A), (B)
Is a graph showing a light intensity distribution obtained on the image plane by the mask of FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Mask 3 Cell area 5, 6 Light transmission pattern

Claims (4)

(57) [Claims] 1. A step of preparing a starting pattern which is constituted by a plurality of cell regions and gives a uniform exposure beam transmission characteristic to each cell region to approximate a desired design pattern; On the other hand, based on random numbers, changes in the exposure beam transmission characteristics for any cell area (mutation)
Forming a plurality of child patterns, crossing or mutating the plurality of child patterns with other child patterns, and evaluating the imaging characteristics of the obtained pattern to obtain a predetermined number of grandchildren. A method of manufacturing an exposure mask, comprising: a genetic algorithm step of leaving a pattern; a step of repeating the genetic algorithm step using the grandchild pattern as a child pattern; and a step of selecting an optimal one of the remaining patterns as a device pattern. . 2. The method of manufacturing an exposure mask according to claim 1, wherein the selection of the hybridization or mutation in the genetic algorithm is determined based on a random number, and the hybridization includes at least a plurality of types including an AND process and an OR process. . 3. The evaluation of the imaging characteristics in the genetic algorithm step includes converting a difference between an exposure beam intensity distribution obtained on an image plane and an ideal exposure beam intensity distribution into energy E, and converting a threshold into temperature T. 3. The method of manufacturing an exposure mask according to claim 1, further comprising a step of determining pass / fail based on a value of exp [-E / T]. 4. The method of manufacturing an exposure mask according to claim 3, wherein said temperature T is gradually reduced in said genetic algorithm step.