JPS631744B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is a figure dividing method in electron beam exposure, particularly in pattern formation using an electron beam, for creating drawing data from design pattern data of an integrated circuit, etc. This relates to the method of

[Conventional technology]

Electron beam exposure is used because the wavelength of electrons is short and is suitable for forming fine patterns, and because electrons have a charge, it is easy to control.
It has attracted particular attention as a pattern forming technology for semiconductor integrated circuits, and has already begun to be applied to the production of photomasks. As integrated circuits become even larger in the future,
A wide variety of LSIs are required in small quantities, and as an alternative to the conventional process using masks, direct writing technology on wafers using electron beams, which does not require the production of masks, is seen as a promising manufacturing technology. In order to establish direct writing technology using an electron beam, the most important requirement was to improve the productivity of exposure equipment, but an exposure equipment that uses a variable shaped beam was proposed, and the writing speed was significantly improved. The reason is,
In the conventional point beam method, the pattern is filled with points, whereas in the variable shaping beam method, the graphic pattern is divided into rectangles with sides of up to 10 μm, for example, and each rectangle is filled in with a surface. be. It is important to note here that in the point beam method, the drawing time is determined by the area of the pattern to be filled, whereas in the variable shaped beam method, the drawing time is determined by the number of surfaces to be filled, that is, the number of shots. . In other words, in the variable shaped beam method, even when forming the same pattern, the productivity of the exposure apparatus is improved by changing the method of dividing the pattern and reducing the number of shots.

Now, when forming an integrated circuit pattern by electron beam exposure, the graphic data obtained from the design cannot be used as drawing data for electron beam exposure in its original form. The reason for this is that while design figures are generally described as polygons, electron beam writing data only allows trapezoids at most, and the figure data output in design is expressed only in rectangles. Even so, this is because rectangles overlap each other and result in multiple exposure, which is not suitable for drawing data. Therefore, the design data is subjected to, for example, discursive processing to remove multiple exposures, and then divided into basic figures such as rectangles and trapezoids or drawing unit figures. Here, I would like to touch on basic figures and drawing unit figures.Basic figures are the figures obtained when various patterns are divided into simple patterns such as rectangles and trapezoids, and drawing unit figures are figures that cannot be drawn. This is the unit figure used to create the pattern. In a point beam type exposure apparatus, a basic figure is usually a drawing unit figure. However, in the shaped beam method, since the basic figure generally exceeds the maximum beam size, the basic figure is further divided into figures of less than the maximum beam size to form drawing unit figures. In the shaped beam method, in order to obtain drawing unit figures, there is a method of directly dividing the contoured figure into drawing unit figures, but this method tends to damage the repeatability of drawing unit figures and is not suitable for data compression. For this reason, it is usually divided into basic figures and then divided into drawing unit figures. The following explanation will be based on the latter method.

Conventionally, processing speed has been considered important in the division method into basic figures. An example of the conventional method will be explained using FIG. P represents one contour figure that can be expressed as a collection of rectangles with no slope. Among the sides representing the contour of this contour figure P, the side parallel to the X axis will be written as =(x, x', y). Here, x and x' are the X coordinate values at both ends of the side, x<x', and y is the Y coordinate of the side. Once determined in this way, it can be processed and divided into basic figures as follows.

(1) Among the sides parallel to the X coordinate of P, choose the one with the maximum Y coordinate value.

Let = (x ₁ x ₁ ′, y ₁ ).

(2) Among the sides whose projection in the Y-axis direction overlaps (when moving along the Y-axis), choose the one with the largest Y coordinate.

= (x ₂ , x ₂ ′, y ₂ ), (x ₁ x ₁ ′)∩(x ₂ x ₂ ′)≠0 (3) Line drawn perpendicularly from the sides of rectangle L to points A and B and the area enclosed by a part of the side or an extension. That is, x ₁ ′≦x≦x ₁
Let the set of points (x, y) satisfying y ₂ <y < y ₁ be a rectangle.

(4) If L=P, the process is finished; if not, the figure obtained by removing L from P is set as a new P, and the process returns to (1) and the same process is repeated.

The above method was described as a method of focusing on the side with the maximum Y coordinate value and projecting it downward, but it is also possible to reverse the idea of up and down and project from the bottom to the top,
The same is true if we consider it as a left-right projection. Here, we will take up the above-mentioned top-to-bottom projection, and hereinafter this division method will be referred to as a conventional division method and compared with the present invention.

There are other conventional methods for dividing outline figures into basic figures, but either method has problems in pattern formation. First of all, when a variable shaping beam is used, the number of shots increases significantly and productivity decreases, and when a point beam is used, the beam position is susceptible to fluctuations, so-called beam drift, resulting in pattern distortion. Quality tends to deteriorate. Second, when shaped beams are used, pattern quality deteriorates at the joints of each shot because beams of various sizes are mixed.

The first problem will be explained using FIGS. 2A to 2C. This pattern actually exists in the gate pattern of 64k-bit MOS memory, and is approximately 2μm wide and 2μm long.
It extends over 100 mm and is repeated over 100 times per chip. When the above-described dividing method is applied to the contour figure in FIG. 2A, the contour figure in FIG. 2B is obtained. In the point beam method, this figure becomes the drawing unit figure, but if you draw it twice as an extremely thin and adjacent figure like this, it will be affected by beam drift, and the patterns will tend to separate at the contact area between the two patterns. Pattern quality tends to deteriorate. In the variable shaping beam method, the figure shown in FIG. 2C, which is obtained by further dividing the figure shown in FIG. 2B by the maximum dimension of the beam used, becomes the figure of the drawing unit, but in this case, the number of shots increases.

Next, the second problem will be explained using FIGS. 3A to 3C. As is already known, due to space charge effects, the shaping beam edge droop is proportional to the beam current. Figure 3A shows the drooping of this beam edge. Here, the vertical axis represents the current density of the shaped beam, and the horizontal axis represents the beam dimensions. Since the current density of the beam is constant, as the beam size increases, the beam current increases and the width of the beam edge dip becomes larger. That is, in FIG. 3A, the droop widths S ₁ and S ₂ satisfy S ₂ >S ₁ . FIG. 3B shows a case where beams with the same beam size are connected, and FIG. 3C shows a case where beams with different beam sizes are connected. In FIGS. 3B and 3C, the vertical and horizontal axes are the same as in FIG. 3A. If the beam dimensions are the same, the overall current density, that is, the charge distribution, will be flat as shown in Figure 3B, as shown by the dotted line, but if the beam dimensions are different, it will be flat as shown in Figure 3C. As shown by the dotted line, the charge distribution becomes non-uniform. This bias in charge distribution becomes more noticeable when the beam dimensions differ greatly, and the pattern becomes more likely to be interrupted, resulting in a deterioration in pattern quality. Therefore, in pattern formation, it is not desirable to connect shots with greatly different beam currents, and the use of beams with extremely large or small beam currents must be avoided. For these reasons, it is necessary to set the upper and lower limits of the beam area to be used, but although it is easy to set the upper limit of the area, it is extremely difficult to specify the lower limit of the beam area to be used for a variety of LSI patterns. It is difficult and no special treatment has been applied until now.

[Problem that the invention seeks to solve]

The purpose of the present invention is to reduce the number of shots in an exposure device that uses a shaped beam, improve productivity, and improve pattern quality, and to reduce the number of connections between patterns in an exposure device that uses a point beam. The object of the present invention is to provide a figure division method in electron beam exposure that solves the first problem and can improve pattern quality.

[Means for solving problems]

The present invention provides a figure dividing method in electron beam exposure in which a figure to be formed is divided into basic figures consisting of rectangles when forming a pattern using an electron beam.
The present invention is characterized in that it includes the steps of determining the magnitude relationship between the length of the maximum side of the shaped beam and the length of the division path required for division, and the step of determining the division path based on the determination.

[Effect]

In the present invention, the magnitude relationship between the length of the side of the figure, the length of the maximum side of the shaped beam, and the length of the dividing path is determined, and the dividing path is determined. This makes it possible to avoid unnecessary divisions, and the number of drawing unit figures (number of shots) is greatly reduced. This reduction in the number of shots not only increases the productivity of electron beam lithography, but also prevents separation of connected portions of patterns due to the influence of beam drift, thereby improving pattern quality.

〔Example〕

Next, the present invention will be explained based on examples.

In order to explain the embodiments of the present invention in an easy-to-understand manner, it is assumed that the graphic pattern of the integrated circuit is composed of a collection of rectangles with no slope. If there is a figure including a diagonal line such as a rectangle with an inclination, additional processing is required in addition to the embodiments of the present invention described below, but the present invention is still applicable. The electron beam exposure apparatus will be described below as one of a variable shaping beam type, which is particularly effective in achieving the effects obtained by the present invention.

Before describing the embodiments of the present invention in detail, an outline of the embodiments will be described below.

(1) Determine whether there is a short side smaller than the minimum allowable side length ε of the shaped beam among the sides of the contoured figure, and if so, divide the line extending the short side. By dividing it as a route, there is no short side smaller than ε in the outline of the divided figure (corresponding to FIG. 4B shown later).

(2) In the case of (1), if a new short side less than ε is created in the outline of the new figure generated by one division, this short side is extended. Further divide the line as a dividing route (4th
(corresponds to Figure C).

(3) In the case of (1), in order to avoid a figure with a width less than ε occurring as a result of division, it is searched to see if an edge exists from the division route to an area ε apart, and if it exists, The dividing path extends up to the point where the edge begins to exist, and does not extend beyond that point (corresponding to Figure 4D).

(4) Classify the figures divided by (1) to (3) into two types.

Type: When the projection of the top side parallel to the X axis in the Y direction (downward) overlaps sides other than the top side (corresponding to Fig. 5A described later).

Type: When the projection of the topmost side parallel to the X-axis in the Y direction (downward) touches the side parallel to the X-axis (corresponding to Figure 5B).

(5) The mold and the shape of the mold are further defined by the length of the top side w, the length of the maximum side of the forming beam l, and the dividing path.
Classification is based on the size relationship _{of the} length _{of C}
(corresponds to the figure).

(6) When classifying the figures classified in (5), if you divide them using the conventional division method, the number of divisions will increase when you further divide them into drawing unit figures, making them undesirable. In the type, the first
(corresponding to (4) in Figure 0) is divided by a dividing path different from the conventional method (C _y for the type, C _x for the type). Figures in other categories are divided using the conventional division method.

(7) In the division in (6), for those divided using the conventional division method (C _x for the mold, C _y for the mold), the result of division is a figure with a width less than ε. In order to avoid this, we search to see if there is an edge to a region ε away from the dividing path,
If it exists, the division path is defined as the point where that side begins to exist, and is not extended any further (corresponding to FIG. 9 and FIG. 12A, which will be described later).

(8) Repeat steps (4) to (7) to divide all shapes into basic shapes.

(9) Divide the basic figure into drawing unit figures.

The above is an outline of the figure division method in the embodiment of the present invention, and the essential steps in the embodiment of the present invention are the steps (4) to (6) above. If the above steps (1) to (3) and (7) are not performed, the division will include edges less than ε, but even in that case, the effect of reducing the number of shots can be obtained as an effect of the present invention.

Next, embodiments of the present invention will be described in more detail.

First, a description will be given of processing when a short side with a length less than ε exists among the sides of the outline of a graphic pattern. Fourth
An example is shown in Figure A. The length indicated by ε in the figure is
This is the minimum allowable length of one side of the shaped beam. The value of ε varies depending on the maximum dimension of the beam, beam current density, etc., but for example, ε=0.5 μm is considered. The broken line indicates a dividing path according to the conventional dividing method, and if this is followed, a basic figure with a short side whose side length is less than ε will be generated. In this way, in the case of a figure that has a short side with a length less than ε among the edges representing the outline, dividing this figure into basic figures as is will likely result in a basic figure with sides whose side length is less than ε. . Therefore, in the present invention, if there is a short side with a length less than ε, the short side is extended and used as a dividing path, and the original figure is divided in advance.
An example is shown in FIG. 4B. Here, the broken line is the dividing route when the present invention is implemented. Two cases need to be considered when extending the short side to form a dividing path. One is as shown in FIG. 4C, where as a result of the division by extending the short sides, a step called ABCD, which includes the short sides less than or equal to ε, occurs in the divided figure. In this case, extend the newly generated short side BC,
It is sufficient to use CE as the next division route. The other one is
As a result of division by extending the short side as shown in Figure 4D,
This is the case when a peninsula-like figure called ACDE whose width is less than ε is generated. It is desirable to avoid such division because pattern formation will not be possible unless a beam with a side of less than ε is used. Therefore, when extending a division route, we search for the existence of a pattern boundary including neighboring regions that are ε away from the route, and if a pattern boundary exists, the extension of the division route is performed until the pattern boundary newly exists in the vicinity. Let's just get started. In the example shown in FIG. 4D, ABE becomes the new division route. The handling of the newly generated BE level difference less than ε is the same as in the case of FIG. 4C, as described above.

Note that if a peninsular pattern whose width is less than ε exists in the first place, the beam size for forming this pattern must also be less than ε. Also, in the case of the pattern shown in Fig. 4E, unless multiple exposure is allowed,
No matter what division path is chosen, a basic figure with one side less than ε is required.

As a result of the above processing, there is no short side with a length less than ε among the sides of the contour figure, except in unavoidable cases as shown in FIG. 4E. I will explain.

First, graphic patterns are classified into two types. Representative patterns of these two classifications are shown in FIGS. 5A and 5B. Here, w is the length of the top edge, C _X , C _y
indicates candidates for dividing routes in each pattern.
Focusing on the top edge parallel to the X-axis = (x ₁ , x ₁ ′, y ₁ ), consider the downward projection of this edge in the Y-axis direction, and if the projection first overlaps with another edge parallel to the X-axis is a type, and the case where the projection first contacts another side parallel to the X axis at x = x ₁ or x = x ₁ ' is a type. In other words, as shown in Figure 5A, when the downward projection of side AB first overlaps with side CD, it is a type, and as shown in Figure 5B, when it first comes into contact with side EF, it is a type. . On the other hand, as shown in FIG. 5C, if it applies to both types, it is classified as type. In this way, all graphic patterns can definitely be classified into one of the categories.

First, a method of dividing a type graphic pattern will be explained. The dividing path is either C _x or C _y in FIG. 5A. The length of the top side w, the length of the maximum side of the shaped beam l, the length of the dividing path C _x , l _CX , the dividing path C _y
The 6th classification of patterns according to the size relationship of the length l _Cy
As shown in the figure. w, l _Cy was compared with l, and l _cx was compared with l _Cy ×n. l _cx may be compared with l, but in order to make the pattern shape dependence more obvious, it is compared with n times l _Cy (n is a positive integer, for example, n=2).

In the conventional division method, _CX is used as the division route, but among the various patterns shown in FIG. 6, the number of shots increases and the division becomes undesirable in case (3). In such cases, w≧l, l _cx ≧nl _Cy , l _Cy <
In the case of l, the present invention performs the division along the path of C _y .
In the case of figures of other classifications, that is, (1) (2) (4) (5) (6) and (7), division is performed using the conventional C _X route.

FIG. 7 is a diagram showing the effect when the dividing method of the present invention is applied to the graphic pattern in the case (3) above.
B shows the situation when dividing into drawing unit figures by the conventional method, and B shows the situation when dividing into drawing unit figures by the dividing method of the present invention. From FIGS. 7A and 7B, it is clear that the division method of the present invention is effective in reducing the number of shots.

Now, when C _y is a dividing path, no edge with a side length less than ε can occur, but when C _X is a dividing path, an edge less than ε may occur, so In order to eliminate the occurrence of edges less than ε, it is sufficient to divide as follows. Such a case will be explained using FIG. 8. In the same figure, w, C _X ,
C _y is already shown in FIGS. 5A and 5B.
Here, the dashed-dotted line m located at a distance p below the dividing path C _x indicates that the graphic pattern always exists up to that point. As shown in FIG. 8A, if p≧ε, there is a sufficiently large graphic pattern below the dividing path C _x , so when dividing along C _X , no problem will occur with the remaining graphic patterns below. However, when p<ε as shown in Figure 8B,
Since a drawing unit figure whose side length is less than ε occurs in the figure pattern below C _x , division at C _X is inappropriate as it is.

Therefore, when extending the dividing route, we search for the existence of a pattern boundary, including the region ε away from the lower side of the route, and if a pattern boundary exists, the extension of the dividing route will be performed until the pattern boundary newly exists in the vicinity. Let's just get started.

In the pattern shown in FIG. 9, the division route does not extend from point D to G, but ends at point H, and is divided into rectangles whose sides are .

Next, a method of dividing a graphic pattern of the type shown in FIG. 5B will be explained.

The dividing path is either C _x or C _y in FIG. 5B. As in the case of molds, the classification of various patterns is shown in FIG. w, l, l _Cy , and n are as already shown in the explanation of FIG. In the shape pattern of the type, the division path in the X direction _C
Since the length _lCX of is equal to w, in the classification shown in FIG. 10, there is no need to classify by _lCX , and only the magnitude relationship between l, _lCy and w needs to be considered.

In the conventional division method, the division path is C _y , but among the various patterns shown in FIG. 10, the number of shots increases and the division becomes undesirable in case (4). That is, in the case of w<l, _lCy ≧nw, the present invention divides along the division path _CX . Other (1)～
In the case of (3), division is performed using the conventional division route C _y . It is clear that by dividing in this way, when divided into drawing unit figures, the effect of reducing the number of shots can be obtained, as in the case of the figure pattern of the type shown in FIG.

Now, _{if C} To eliminate the occurrence of edges, you can divide it as follows. Regarding such cases, the first
This will be explained using Figures A and B. w, C _x , C _y , p,
Regarding m, it is as shown in FIG.

If p≧ε as shown in Figure 11A, the division path
Since there is a sufficiently large figure to the right of C _y ,
When dividing by C _y , there will be no problem with the remaining figure pattern on the right. As shown in FIG. 11B, when p<ε, the figure pattern on the right side of C _y does not produce a figure whose side length is less than ε, so the pattern boundary is searched including the area ε away from the dividing route, and the dividing route is Determine. FIG. 12A shows a case where the pattern shown in FIG. 11B is divided according to the present invention. The shaded area is the figure to be divided. Special examples of molds are shown in Figures 12B and C. Here, if the Y value of the side in contact with the downward projection of side AB in the Y-axis direction is equal on the left and right sides, for example, if w≧l as shown in Figure 12B, then divide by C _y ', and as shown in Figure 12C As shown in w<l
If so, divide by C _x .

A comparison of division between the embodiment of the present invention described above and the conventional method is shown in FIGS. 13A and 13B. Figure 13A shows the case where the present invention is applied, and Figure 13B shows the case where the conventional method is used. According to the present invention, a pattern can be formed with about half the number of shots compared to the conventional method, increasing the productivity of the device. will improve. Actually 64k bit DRAM
As a result of applying the present invention to a gate pattern, it became possible to form a pattern with 55% fewer shots compared to the conventional example.

〔Effect of the invention〕

As explained above, when pattern formation is performed using a variable shaped beam according to the figure division method in electron beam exposure according to the present invention, the number of shots can be reduced and productivity can be improved. Furthermore, the present invention can be effectively applied to pattern formation using a point beam, and can improve pattern quality by reducing the adverse effects of beam position fluctuations.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram of the conventional figure division method, Figure 2A
~C is an explanatory diagram of the problems of the conventional division method, Figure 3A
~C is a diagram showing the width of the beam edge droop, Figure 4A
~E, FIGS. 5A to C, 8A and B, FIGS. 9, 11A and B, and 12A to C each illustrate the dividing method of the present invention based on various forms of patterns. Figures 6 and 10 are diagrams for explaining the application of the present invention, Figures 7A and B and Figures 13A and B are comparisons with conventional division methods. These are plot diagrams for each. P...Contour figure, L...Rectangle, _S1 , _S2 ...Dip width, ε
...Minimum allowable width, w...Length of the top side, l...Length of the maximum side of the shaped beam, C _x , C _y ...Dividing path, l _cx ,
l _Cy ...Division length, p...Distance, m...Position below the division path _CX or outside of _Cy by a distance p.

Claims (1)

[Claims] 1. In a figure dividing method in electron beam exposure that divides a figure to be formed into basic figures consisting of rectangles when forming a pattern using an electron beam, the length of the side of the outlined figure and , a step of determining a magnitude relationship between the length of a maximum side of a shaped beam and a length of a dividing path necessary for dividing; and a step of determining the dividing path based on the determination. Figure division method in beam exposure.