JPH02166717A - Exposing method - Google Patents

Abstract

PURPOSE:To enable a finer pattern to be transcribed without seeking extreme increase of number of openings of a projecting optical system and extreme lengthening of short wave of illumination light by decomposing the whole pattern to be formed on a photosensitive substrate into a plurality of patterns according to the partial shape of the pattern or pattern density, and laying one on top of the other after aligning them with each other so as to do exposure. CONSTITUTION:In each reticle R1, R2, and R3, a light shielding belt SB in accordance with a chip province CP is formed in the circumference, and in each inside pattern PTA1, PTA2 and PTA3, which are made by decomposing a pattern PA, and patterns PTB1, PTB2, and PTB3, which are made by decomposing a pattern PB, are formed, respectively. Also, in each reticle R1, and R2, and R3, marks RM1, RM2, RM3, and RM4 for alignment are provided, which are used for alignment with marks WM1, WM2, WM3 and WM4 which are provided incidentally in the chip province CP. After aligning the patterns PTA1 and PTB1 with the chip province CP and exposing it to the light, it is changed with the reticle R2, and the patterns PTA2 and PTB2 are aligned with the chip province CP and are exposed to the light, and next the reticle R3 is aligned so as to expose the patterns PTA3 and PTB3.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention provides a method for manufacturing semiconductor devices, liquid crystal devices, etc.
The present invention relates to an exposure method for transferring an original pattern formed on a mask onto a sensitive substrate.

[Conventional technology]

In the manufacture of semiconductor devices, miniaturization and higher integration are progressing year by year, and lithography processes with increasingly narrower line widths are required, such as I Mbit memories and 4 Mbit memories.

In order to meet this demand, reduction projection type exposure apparatuses (steppers) are currently the mainstream exposure apparatuses used in lithography processes. In particular, the reticle with the original pattern is 115
1i! A method is often used in which the resist layer on the wafer is exposed to light after reducing the size to about 15×15 angles using a small projection lens.

Year by year, the projection lens of this stepper has been made to have a higher numerical aperture (N, A,) to increase its resolution, and when the wavelength of the exposure illumination light is 436 nm (g-line), N, A.

= about 0.48 is in practical use.

Increasing the numerical aperture of the projection lens in this way means that the effective depth of focus decreases accordingly, and the depth of focus of the projection lens with N, A, = 0.48 is, for example, ±0. .8 μm or less.

In other words, one shot area on the wafer is 15×15
If it is m square, the surface of this entire area (resist layer) is
It must be accurately positioned within ±0.8 μm (preferably within ±0.2 μm) with respect to the best imaging plane of the projection lens.

Therefore, in order to cope with the lack of depth of focus of the projection lens,
While displacing the wafer in the optical axis direction relative to the projection lens,
A method of multiple exposure of the same reticle pattern has been proposed.

This method increases the apparent depth of focus of the projection lens and is an effective exposure method.

[Problem that the invention seeks to solve]

This multi-focus exposure method attempts to guarantee contrast over a wide focal range, although the contrast at the best focus is slightly lowered. Based on the results of experiments, etc., this method is suitable for patterns for so-called contact hole processes, where the pattern surface of the reticle is mostly dark areas (shielding areas) and rectangular openings (transparent areas) are scattered therein. However, it is currently not effective for other patterns, especially reticle patterns such as wiring layers where bright and dark linear patterns are repeated, and in the case of contact holes.

In such a reticle pattern such as a wiring layer, when the focus position is changed, the light intensity due to the defocused image of the bright line portion is given to the portion of the wafer that should originally be a dark line.
This is because the contrast rapidly decreases and the resist film decreases. Furthermore, in the projection exposure method, the period of the repeatable pattern that can be transferred is limited to a certain value or more due to the performance of the projection lens. This value is also called the resolution limit of the projection lens, and the one currently in practical use is 1
15w1 small, N, A, = 0.4517) (!-The line width of the bright and dark lines of the repeated Shiba turn is 0.8μ on the wafer.
m (4 μm on the reticle).

Therefore, even if the line width of the pattern on the reticle is made thinner, patterns with line widths smaller than that will not be exposed properly, and the limit of lithography using the projection exposure method is primarily the performance (resolution power) of the projection lens. It is thought that it is determined by

Also, in the proximity exposure method, it is difficult to make the repetition period of bright lines and dark lines on the mask smaller than a certain value due to the diffraction phenomenon that occurs depending on the wavelength of illumination light.
This is handled by making the wavelength as short as possible. Therefore, soft X
required a special energy beam such as a wire.

The present invention was made in view of these problems, and aims to dramatically increase the numerical aperture of the optical system for projecting finer patterns.
The first objective is to enable transfer without having to make the wavelength of illumination light extremely short.

Furthermore, a second object of the present invention is to provide a method that enables the transfer of finer patterns, regardless of whether the projection exposure method or the proximity exposure method is used.

Furthermore, a third object of the present invention is to provide a method in which the effects of the multiple point exposure method can be sufficiently obtained for most patterns other than contact holes.

[Means for solving problems]

In order to achieve the above object, the present invention decomposes the overall pattern to be formed on a sensitive substrate (a substrate having a layer sensitive to energy rays) into a plurality of patterns according to the local shape and pattern density of the pattern. Then, the separated patterns were aligned with each other and exposed overlapping each other.

Here, an outline of the present invention will be explained based on FIG. 1st
In the figure, the overall pattern to be formed on the sensitive substrate is
Chip (or pattern P created within the short area CP)
ASPB, the pattern PA is a line and space (L/S) pattern bent at 90°,
Pattern PB is a simple L/S pattern.

Patterns PA and PB are each divided into three decomposition patterns, and each decomposition pattern consists of three reticles R+ and R.
1, formed in Rx. Each reticle R+, Rt, R
2, a light-shielding band SB corresponding to the chip area CP is formed around the periphery, and inside it there are three patterns PTA, , PTAx, and PTAs obtained by decomposing the pattern PA, and three patterns PTB+ and PTBI obtained by decomposing the pattern PB.
, PTBs are formed. Also, each reticle R+
, Rz, R3 have alignment marks RM, , R
Mt, RM3, RM are provided, and marks WM, , WMtSWMs are provided along with the chip area CP.
, used for alignment with WMa.

Patterns P TA+ , P TAt , P TAs
, P TB+ 5PTBt , and PTBI are shown as dark lines in the figure, but they are actually bright lines due to the light transmitting part. After positioning the patterns PTA, , and PTBI in the chip $I area CP and exposing them, change to the reticle R8 and print the patterns PTA,
*, PTBz is positioned in the chip area CP and exposed, and then the reticle R is positioned to form the pattern PTAs.
, expose the PTBs.

Each of the patterns PTB, , PTBt, and PTB is a collection of every two linear patterns corresponding to bright lines from the L/S pattern of pattern PB, and the line and space pitch is It is three times that of the entire pattern (duty is 1/3). The same goes for each of the patterns PTA, , PTA, , PTAs, but each pattern is decomposed so that it does not bend at 90 degrees and create a continuous line, like each line of pattern PA. . The 90° bend consists of two lines that are perpendicular to each other (each line is formed on a separate reticle).
The edges are set so that they partially overlap. In this way, in the case of line-and-space patterns, adjacent bright lines are formed on separate reticles, and the pattern density of bright lines is reduced in one reticle (as shown in Figure 1). (1/3) in order to isolate the bright line.

(Function) FIG. 2(A) shows the case where the entire line-and-space pattern P is formed as it is on one reticle R, and FIG. This shows the case of a decomposed pattern P, in which every other bright line of the pattern P is formed, where the bright lines of P, , P, have the same width and are d.Illumination light is irradiated onto these reticles R. Then, diffracted light is generated in a direction corresponding to each pattern pitch P.

The diffraction angle θ of this n-th order diffracted light is expressed as n λ sin θ=□ (where n-0, ±1, above 2-), where λ is the wavelength of the illumination light. That is, the resolved pattern P having a larger pattern pitch has a smaller diffraction angle for the same diffraction order, and as a result, the amount of diffracted light that contributes to image formation of -order or higher increases, resulting in a larger image contrast. An example is shown below.

Figures 2 (C), (D), and (E) show g-lines, N, and A.

-0.45, σ-0.5 projection lens to project and expose 0.4μmL/S (0.4μm wide repeating pattern of bright lines and dark lines) onto a photosensitive substrate at best focus. It shows a calculated value (simulation) of an aerial image, where the σ value represents the ratio of the area of the entrance pupil of the projection lens to the area of the light source image. Figure 2 (C) shows the intensity distribution of the aerial image when exposed with a single reticle, the horizontal axis is the position (μm) on the photosensitive substrate with the center of a certain bright line as the origin, and the vertical axis is the It is a relative strength. Figure 2 (F) shows the sum of the aerial image intensities of the two reticles separated and exposed respectively.
Figures (D) and (E) each represent the intensity distribution of the spatial image of the separated pattern.As is clear from this simulation, the contrast of the spatial image is improved by exposing the pattern in parts.

That is, in the case of an L/S-shaped pattern, by using two or more separated patterns, even if projection lenses with the same numerical aperture are used, more high-order light can be used for imaging. This means that finer linear patterns,
It means to form an image to the maximum extent possible up to the resolution limit determined by the performance of the projection lens, and to improve the image quality of the pattern (image quality of the resist pattern).

Furthermore, by forming the pattern Ph with a lower ratio of bright areas to the overall pattern P1, even if the best imaging plane of the projection lens and the surface of the photosensitive substrate are defocused, the defocused image of the dark areas of pattern Pb is The dark areas are maintained to the last without becoming bright lines, and only the contrast of the bright line image is reduced. Therefore, if the multi-focus exposure method is performed for each separated pattern, the effect of increasing the apparent depth of focus can be obtained as in the case of contact holes.

[Embodiment] FIG. 3 is a perspective view showing the configuration of a projection type exposure apparatus (stepper) suitable for an embodiment of the present invention. The basic configuration of this stepper is the same as that disclosed in, for example, Japanese Patent Laid-Open No. 145'130/1982, and will therefore be briefly described below.

Illumination light from the exposure light source 2 passes through an illumination optical system 4 having a reticle blind (illumination field stop), etc., and illuminates one reticle on a reticle stage 6. Reticle stage 6 includes four reticles R+, Rt
, Rx, and Ra can be placed simultaneously and move two-dimensionally in the x and X directions. On the reticle stage 6, movable mirrors 8x and 8y that reflect a laser beam from a laser interferometer 10 for position measurement are fixed at right angles to each other. Reticle alignment system 12 is the reticle alignment mark R
In addition to detecting M t to RM a, the marks WM1 to WM on the wafer W are also provided so as to be detectable. For this reason, the alignment system 12 is used when positioning one of the four reticles with respect to the device, or when positioning the mark RM.
It can be used both when detecting r to RM and marks WM1 to WMa at the same time and performing die-to-die alignment. Although the alignment system 12 is provided at only one location in FIG. 3, each mark RM+ shown in FIG.
,RM! , RMs , are arranged at multiple locations corresponding to RM. Mark RM, ~RM, or mark W
Photoelectric detection of M, to WMa is performed by the mark detection system 14.

Now, the image of the pattern area of the reticle is captured by the projection lens system 16.
The image is projected onto a chip area CP previously formed on the wafer W via the wafer W. The wafer W is placed on a wafer stage 26 that moves in the X and X directions;
It is composed of an X stage 26x that moves in the X direction on y, and a Z stage 26z that moves slightly in the projection optical axis direction (Z direction) on the X stage 26x. On the Z stage 262, movable mirrors 28x and 28y that reflect the laser beams from the laser interferometers 30x and 30y are fixed at right angles to each other. Further, a reference mark FM is fixed to the Z stage 26z so as to be at approximately the same height as the wafer W. The X stage 26x and the Y stage 26y are driven in each axial direction by motors 27x and 27y. Here, the projection lens system 16 has an image formation correction mechanism 18 incorporated therein, and the optical characteristics (magnification, focus, certain types of distortion, John, etc.) are automatically corrected from time to time. This image formation correction mechanism 18 is, for example, JP-A-60-78
Since it is disclosed in detail in the No. 454 publication, the explanation will be omitted here. This stepper also includes an alignment optical system 20 that detects marks (WM, ~WM, etc.) on the wafer W from below the reticle stage 6 only through the projection lens system 16; A mark detection system 22 that photoelectrically detects the marked optical information
The lens system includes a TTL (through-the-lens) type alignment system consisting of a TTL (through-the-lens) type alignment system, and an off-axis type alignment system 24 that is separately installed immediately adjacent to the projection lens system 16.

Although not shown in Figure 3, Japanese Patent Application Laid-Open No. 60-7845
Similar to the one disclosed in Publication No. 4, an oblique incident light type focus sensor that detects the height position of the surface of the wafer W with high resolution is provided, and together with the two stages 26z, it is used to detect the best imaging plane of the projection lens system. It operates as an automatic focusing mechanism that always aligns the wafer surface with the wafer surface.

In Section 22, the optical relationship between the illumination optical system 4 and the projection lens system 16 in the configuration shown in FIG. 3 will be explained using FIG. 4. The illumination optical system 4 is configured to project a secondary light source image (surface light source) into the pupil EP of the projection lens system 16, and employs the so-called Koehler illumination method.

The surface light source image is set to be slightly smaller than the size of 111EP. Now, if we focus on one point on the reticle R that has the overall pattern P, the illumination light that reaches this point! A certain solid angle θ, /2 exists in L. This solid angle θ,/2 is preserved even after passing through the entire pattern P, and the luminous flux of the 0th order light is. As projection lens system 1
6.

The solid angle θ, /2 of this illumination light IL is also called the numerical aperture of the illumination light. Further, assuming that the projection lens system 16 is a double-sided telecentric system, the chief ray l, which passes through the center of the pupil EP (the point through which the optical axis AX passes) on the reticle R side and the wafer W side, is parallel to the optical axis AX. become. In this way Hitomi EP
The light beam that has passed becomes an imaging light beam IL on the wafer W side, and is focused on one point on the wafer W. In this case, the projection lens system 1
When the reduction magnification of 6 is 115, the solid angle θ of the luminous flux lLo is
l, /2 has the relationship θ, -5·θ. Solid angle θyu/
2 is also called the numerical aperture of the imaging light beam on the wafer W. Also, the numerical aperture on the wafer side of the projection lens system 16 alone is
It is defined by the solid angle of the luminous flux IL when the luminous flux passes through the pupil EP.

Now, if the overall pattern P is equivalent to that shown in FIG. 2(A), then the higher-order diffraction lights D1, D,! ,・
...occurs. These higher-order lights include the zero-order luminous flux D
The one that spreads out to the outside of 1° and the zero-order luminous flux.

There are some that occur distributed inside the . Especially the 0th order luminous flux. Even if some of the high-order light distributed outside the wafer W enters the projection lens system 16, it will be blocked by the pupil EP and will not reach the wafer W. Therefore, if more high-order diffracted light is to be used for imaging, the diameter of the pupil EP must be made as large as possible, that is, the numerical aperture (N, A,) of the projection lens system 16 must be further increased.

Alternatively, by reducing the numerical aperture (solid angle θ, /2) of the illumination light TL (reducing the diameter of the surface light source image), the spread angle of the higher-order lights D0, D1□, etc. from the pattern P can be kept small. It is also possible.

However, in this case, the zero-order imaging light flux IL on the wafer W side,
If the numerical aperture (solid angle θ, /2) is made extremely small, the original resolution performance will be impaired. Furthermore, originally
The diffraction angle of higher-order light is uniquely determined by the pitch and duty of pattern P, so let's assume that it is illumination light! L
Even if it is possible to bring the solid angle θ1/2 of However, when the entire pattern is divided into a plurality of decomposition patterns as in this example, as is clear from FIG. It becomes possible to easily pass through the pupil BP.

By the way, in FIG. 3, four reticles R to R4 are placed on the same reticle stage 6, and the center of any one of them is aligned with the optical axis AX of the projection lens system 16.
It can be exchanged to be located above. Since the laser interferometer 10 is used, the positioning accuracy of each reticle during this exchange can be extremely high (for example, ±0.02 μm). Therefore, the mutual positional relationship of the four reticles R1 to R1 can be If precise measurements are made in advance, each reticle can be positioned by moving the reticle stage 6 based only on the coordinate measurement values of the laser interferometer 10. Further, even if the mutual positional relationship of each reticle R6 to R4 is not measured in advance, precise positioning can be performed for each reticle using the alignment system 12, mark detection system 14, reference mark FM, etc.

Furthermore, in this embodiment, a multifocal exposure method is used in conjunction with the exposure of each reticle R1 to R4 having a separated pattern. Therefore, when exposing one chip area (shot HJ [area) CP on the wafer W] using a certain reticle, the height position Z on the wafer surface detected by the oblique incident light type focus sensor as the best focus point is , a height position C, for example, about 0.5 μm above this position Z0, and Z. Exposure is repeatedly carried out at each of the three focal positions of the height position Z2, which is, for example, about 0.5 μm below.

Therefore, while exposing a certain chip area CP with one reticle, the height of the wafer W is reduced by 0.5 by the Z stage 26z.
It is moved up and down in μm steps.

Incidentally, instead of moving the Z stage 26z up and down during the exposure operation, the same effect can be obtained by moving the best imaging plane (reticle conjugate plane) of the projection lens system 16 itself up and down using the imaging correction mechanism 18. can get. In this case, JP-A-60-784
As disclosed in Japanese Patent No. 54, the imaging correction mechanism 18
Since this is a method of adjusting the gas pressure in the sealed lens space in the projection lens system 16, an offset is added to the pressure adjustment value for the original correction to move the image plane up and down by about ±0.5 μm. It is only necessary to apply a pressure value during the exposure operation. At this time, only the focal plane is changed by pressure offset,
It is necessary to select a combination of lens spaces that does not change the magnification, distortion, etc.

Furthermore, by taking advantage of the fact that the projection lens system 16 is a double-sided telecenter link, by moving the reticle up and down, it is possible to similarly change the height position of the best imaging plane. Generally, in the case of reduction projection, the amount of defocus on the image side (wafer side) is converted to the amount of defocus on the object side (reticle side).
It is determined by the square of the reduction magnification. Therefore, when a focus shift of ±0.5 μm is required on the wafer side, the reduction magnification is set to 11
5, on the reticle side ±0.5 / (115)
”=±12.5 μm.

Next, although the first drawing has been briefly explained, some examples of dividing the entire pattern into decomposed patterns are shown in FIGS. 5 and 6.
This will be explained with reference to FIGS.

In FIG. 5, the overall pattern has a bright line pattern PLc with a width D1 and a width (D2ζD,) as shown in FIG. 5(A).
The dark line pattern PL is repeated alternately.
In the case of AND SPACE, this is an example in which decomposed patterns as shown in FIGS. 5(B) and 5(C) are formed on each of two reticles. The decomposition pattern in Fig. 5(B) and the decomposition pattern in Fig. 5(C)
), bright line patterns PLc are formed every other line compared to the entire pattern. In the two decomposition patterns, the positions of the bright line patterns PLc are complementary. In this case, the pitch in the overall pattern is DI +Dt (!'i2D+) and the duty is DI / (D, +I)z) - 1/2, but the pitch in the decomposed pattern is 2D, +2Dz (-4D,
), duty is DI / (2DI + 2 Dz)
L:, )/4. Therefore, the bright line pattern PLc on each reticle is isolated.

In Fig. 6, the overall pattern is L/S as shown in Fig. 6 (A).
Instead of dividing each bright line pattern PLc into a separate reticle, each bright line pattern is divided into minute rectangular bright areas PL, as shown in FIGS. 6(B) and (C). This shows how R1 is complementary to each other. In this method, the two decomposition patterns are determined such that the isolated rectangular bright portions PLa are shifted in directions perpendicular to each other in the L/S pitch direction. Therefore any 1
Focusing on the rectangular bright area PL4, there is a dark area with a width of (DI + 2 Di) on both sides of the L/S in the pitch direction, and the duty in the pitch direction is approximately 1/
It is now 4.

Figure 7 shows a linear pattern that bends at right angles in the overall pattern as shown in Figure 7 (A), and is divided into two lines at the bending part according to direction as shown in Figures 7 (B) and (C). It shows how the linear patterns PT, , PT are formed. Here pattern P
The inside of T, PTf is a transparent part, and the surrounding area is a shielding part. Here, if the two patterns PT, PTr are bright parts, it is preferable that they partially overlap at the bending part.

However, the overlapping part has two patterns PT,
, PT, are both about 45'' in the longitudinal direction. Therefore, the patterns PT, , PTf
The connection part should not be made at a right angle (for example, it should be cut at a 45° angle).

In this way, when a linear pattern bent at 90° is decomposed into two patterns PT, , p'rr and exposed in a superimposed manner, image reproduction on the resist at the bent part is particularly good, and even at 90" The shape of a part of the curved inner corner is exposed clearly.The same method can also be applied to linear patterns bent at other angles. In the case of a pattern with curved edges, it is preferable to separate it into two patterns depending on the two directions of the edges.

Fig. 8 shows that the overall pattern that intersects in a T-shape as shown in Fig. 8 (A) is divided into two linear patterns PT, PT, depending on the direction as shown in Fig. 8 (B) and (C). Shown when disassembled. Linear pattern PT.

, PT, are both bright parts, the tip of the linear pattern PT, is an isosceles triangle with an angle of 90" or more, and this triangular part forms the pattern as shown in FIG. 8(C). Make it partially overlap the straight edge of PTh.In this way, the 90° of the T-shaped pattern
A part of the corner of the angle becomes extremely clear on the resist image and becomes less rounded.

Several examples of pattern decomposition have been shown above, but for the entire pattern PA shown in FIG. 1, multiple decomposition patterns PTA,
, PTAz, and PTAs. Note that the number of decompositions is not particularly limited as long as it is 2 or more. however,
If the number of separated patterns (reticles) is large, errors during overlapping exposure will accumulate, which is disadvantageous in terms of throughput.

Each further decomposed pattern has a separate reticle R.
1 to R4, but JP-A-62-145
As disclosed in Japanese Patent No. 730, a plurality of pattern areas of the same size are provided on two large glass substrates,
Each decomposed pattern may be provided within each pattern area.

Next, a typical sequence of this embodiment will be explained with reference to FIG.

[Step 100] First, each reticle R1 to R4 having a decomposition pattern is placed on the reticle stage 6! ! 2 positions, each reticle R3 to R4
is accurately positioned on the reticle stage 6 using the alignment system 12. In particular, the rotation error of each reticle R1 to R4 is made small with sufficient accuracy. For this reason,
A minute rotation mechanism is provided on the reticle stage 6 at a portion that holds each reticle R1 to R4. However, the a-lateral movement in which each reticle R to R# is minutely moved in the X'- and V directions can be omitted. This is because the coordinate position of the reticle stage 6 itself is precisely controlled by the laser interferometer 10, and the marks RM, ~RM, of each reticle R7 to R4,
It is sufficient to store each coordinate value when the reticle stage 6 is positioned so that the alignment system 12 detects the . In addition, the rotation criteria for each reticle R1 to R2 is actually the laser interferometer 30x on the wafer stage side,
Since the coordinate system is defined by 3oy, the fiducial mark FM
It is necessary to detect the marks RMI-RM and RMI-RM by the alignment system 12, and drive the low chiron errors of each reticle R1 to R4 to zero in the coordinate system on the wafer stage side. An alignment method related to such reticle rotation is disclosed in, for example, Japanese Patent Application Laid-Open No. 1868-1868.
This is disclosed in detail in Publication No. 45.

[Step 101J Next, the opening shape and dimensions of the reticle blind as an illumination field stop provided in the illumination optical system 4 are set to match the light-shielding band SB of the reticle.

[Step 102] Subsequently, the wafer W coated with photoresist is loaded onto the wafer stage, and several alignments on the wafer W are performed using the off-axis alignment system 24 or the TTL alignment optical system 20. The mark attached to the chip area CP is detected, the entire wafer is aligned (global alignment), and the wafer W
The positional relationship between the arrangement coordinates of the upper chip area CP and the light mAX (center point of the pattern area of the reticle) of the projection lens system 16 in the X-y plane is defined. Here, when the exposure to the wafer W is the first print, the marks WM, -W
Since M does not exist, step 103 is performed.Next, the pattern number n corresponding to the number of decomposed patterns, that is, the number of reticles, and the chip number m corresponding to the number of chip areas CP to be exposed on the wafer W are determined. Registered in the main controller including the computer. Here, the pattern number n is set to any one of the number A of reticles, and the chip number m is set to 9 at maximum and to l in the initial state.

[Step 104] Next, the reticle stage 6 is precisely positioned so that the reticle corresponding to pattern number n is directly above the projection lens system 16.

[Step 105] Then, the wafer stage is changed based on the chip number m.
m-th chip area CP to be stepped and exposed
is positioned directly below the projection lens system 16. At this time,
The center of the n-th reticle and the center of the m-th chip region CP are usually aligned within a range of about ±1 μm, depending on the result of global alignment.

[Step 106] Next, if the Gui-Pai-Gui alignment is to be performed, the alignment optical system 12 or the alignment optical system 20 is used to align the reticle marks of marks WM to WM4 attached to the chip area CP. RM, ~RM
The positional deviation with respect to a is precisely measured, and either the wafer stage 26 or the reticle stage 6 is slightly moved until the positional deviation falls within an allowable range.

Incidentally, instead of using the TTL type alignment optical system 20 or the alignment optical system 12 to perform the guide-by- guide alignment, as disclosed in Japanese Patent Application Laid-Open No. 61-44429, 3 to 9 points on the wafer W can be aligned. Enhanced global alignment (E, G, A) that measures the positions of the marks WM, ~WM, in the chip area CP, and calculates the stepping positions of all chip areas using a statistical calculation method based on the measured values. Laws, etc. may be adopted.

[Step 107] Next, the m-th chip area CP is exposed using the n-th reticle, but since the multi-focus exposure method is applied to each chip area, first, the chip area is The oblique incident light defocus sensor is used to precisely measure the height position of the chip area surface relative to the best imaging plane.

Then, after adjusting to the best focus position using the Z stage 26z, the pattern on the reticle is exposed with about 1/3 of the normal exposure amount. Next, for example, 0 on the wafer W.
．． If the best focus is the position where the 5 μm L/S pattern is accurately imaged, +0 for this height position.
．． The Z stage 26z is offset to each of two locations changed by about 5 μm, −0, and 5 μm, and exposure is performed with about 1/3 of the exposure amount at each height position. In other words, in this embodiment, triple exposure is performed at a total of three height positions including the best focus point and points before and after the best focus point.The exposure amount for each multiple exposure may be approximately 173, which is the normal exposure amount. When moving the best image plane itself up and down using the image correction mechanism 18, instead of fixing the image plane position in steps, it is better to make slight adjustments. Exposure can also be performed while moving the image plane. In this case, the shutter provided in the illumination optical system 4 only needs to be opened once for one depth region CP, which is extremely advantageous in terms of throughput.

[Step 10] When the exposure of the m-th chip area is completed, the set value of m is incremented by 1.

[Step 109] Here, it is determined whether exposure of all chip areas on the wafer W has been completed. Here, the maximum value of m is set to 9, so if m is 10 or more at this point, the process advances to the next step 110, and if it is 9 or less, the process returns to step 105, and stepping to the next chip area is performed. It will be done.

[Step 110] When the nth reticle is exposed on the wafer W, the wafer stage is set to the exposure position for the first chip area, and the chip number m is set to 1.

[Step 111] When all the reticles of the decomposition pattern prepared here have been exposed, it means that the exposure for one wafer has been completed. If there are still reticles left,
Proceed to step 112.

[Step 112] Next, the pattern number n is changed to a value corresponding to another reticle, and the process returns to step 104 to repeat the same operation.

Of the above steps, it goes without saying that in addition to step 102, step 106 is also omitted during first printing.

As described above, wafers W are processed one after another, but
For example, when processing multiple wafers that have undergone the same process, all wafers in the loft are exposed with the first reticle, the reticle is replaced, and the next reticle is used to expose all the wafers in the loft. The sequence may also be such that several wafers are exposed. Also, in step 106,
When performing pie-gui alignment, one type of mark attached to the chip area CP is placed on each reticle R to R4.
If they are used in common during alignment with each of the reticles, the relative positional deviation between the patterns of each reticle transferred onto the wafer W can be minimized.

Furthermore, when adopting the E, G, A method, drift of each alignment system, drive system, etc. during the exposure sequence may become a problem, but using the fiducial mark FM every time the reticle is replaced or the wafer By checking the drift in each prefecture every time exposure is completed, even if a drift occurs, it can be corrected immediately.

As described above, in this embodiment, since multiple exposure is performed for each isolated separated pattern at a plurality of focus positions, both an increase in the resolution limit and an increase in the depth of focus can be obtained. The resolution limit referred to here means that because the entire pattern on the reticle is dense like an L/S shape, the bright lines and dark lines when the pattern is transferred onto the resist are well separated due to diffraction phenomena, etc. This means the limit at which images cannot be resolved, and has a different meaning from the theoretical resolving power of the projection lens system 16 alone.

In this embodiment, each linear pattern in the overall pattern is separated into isolated patterns, and the isolated patterns are projected, so the theoretical resolving power of the projection lens system 16 is used at once. It is possible to transfer finer linear patterns. This effect can be obtained even when the multi-focus exposure method is not used together, that is, when the reticles R1 to R4 of each separation pattern are exposed in a superimposed manner while the Z stage 26z is fixed at the best focus in step 107 in FIG. The same can be obtained.

Next, a pattern separation method and an accompanying exposure method according to a second embodiment of the present invention will be explained. FIG. 1θ (A) is a cross section schematically showing an example of a circuit pattern configuration formed on a wafer W, and in the latter half of manufacturing, minute irregularities are formed on the wafer surface. Depending on the case, this minute unevenness may be the depth of focus of the projection lens system 16 (for example, ±0.8 μm).
It can sometimes be larger than. FIG. 1O (A) shows a case where a resist layer PR is formed on the wafer surface, and patterns P0, Po, and Pr4 are exposed on the convex portions of the wafer, and patterns Prl are exposed on the concave portions. In this case, conventional exposure In this method, a pattern P as a transparent part is placed on one reticle.
All of rl to P r4 were formed, but in this example, the patterns P□, Po, P,
There is a transparent part Ps on the reticle R2 as shown in Figure 1O (B).
1, P1□, and PI4, and the pattern P to be exposed in the recessed portion is formed as a transparent portion P0 on the reticle R2 as shown in FIG. 10(C).

When performing overlapping exposure using the respective reticles R1 and R1, when the reticle R3 is used, the projection lens system 1
For reticle R2, exposure is performed so that the best image forming surface of No. 6 is aligned with the convex portion side of the wafer W, and in the case of reticle R2, the best image forming surface is aligned with the concave portion side of the wafer W. In this way, all the patterns in the chip area CP are exposed with extremely high resolution, and it is possible to prevent partial defocusing from occurring due to the influence of the convex portions and concave portions.

In this embodiment, the multifocal fogging method described in the first embodiment may also be used in conjunction with the exposure of each reticle R, , R,. Furthermore, when a linear pattern is exposed from concave to convex parts on the wafer W, the linear pattern can be disassembled in the longitudinal direction on the reticle and divided into parts covering the convex parts and parts covering the concave parts. good. Furthermore, the convex portions and concave portions on the wafer W may be divided into three stages, three separated patterns may be created, and exposure may be performed separately at three focal positions. of course,
The decomposition rules explained in FIGS. 5 to 8 may be used together.

FIG. 11 is a diagram illustrating a pattern decomposition method according to the third embodiment.

In recent years, it has been proposed to insert sub-space marks in order to accurately reproduce the shapes of minute isolated patterns (contact holes, etc.) and corner edges formed on a reticle for exposure. FIG. 11(A) shows a minute rectangular opening P e formed on the reticle as a contact hole.
ll and this opening P. When exposed onto a wafer, it becomes about 1 to 2 μm square. When this type of opening p cm is exposed by projection, the 90" corner on the resist often collapses and becomes rounded. Therefore, the size of the opening p cm is so small that it cannot be resolved by the projection optical system (for example, 0.2 μm square on the wafer). Sub-space marks M1 are provided near the four corners of the opening P-.

In this way, the original opening P. In addition, when forming a sub-space mark Msp, if the arrangement pitch of the openings P- becomes narrower, it becomes difficult to insert the sub-space mark M□ with a conventional reticle. However, as in the present invention, if every other opening P- in the overall pattern is formed in separate reticles (or separate separated patterns) together with sub-space marks M0, the surroundings of one opening PC1 can be Since there is enough space (shielding part) for sub space mark M3. It has the advantage of giving you more freedom when it comes to setting it up.

FIG. 11(B) shows a case where linear sub-space marks M- are provided on both sides near the end of the line pattern P-. When the entire pattern is divided into separated patterns, sub-space marks M6. .

It must be formed on the reticle together with the decomposed pattern. Also, one overall pattern (
For example, when a curved linear pattern is decomposed into a plurality of patterns, if a corner edge is generated in each decomposed pattern, a new sub-space mark may be provided near the corner edge.

FIG. 12 is a diagram illustrating a pattern decomposition method according to the fourth embodiment.

In addition to the effects described in each of the previous embodiments, this embodiment provides the effect of achieving fine line width lithography that exceeds the resolution limit of the projection optical system.

FIG. 12(A) shows an example of the cross section of the wafer W, and shows thin line patterns Pr5, Pr6, Pr? This shows the case where the image is left as a resist image.

On the resist layer PR, patterns P, ss P l'6,
Pr? Assuming that the entire surrounding area of
Divide into parts. In FIGS. 12(B) and 12(C), light shielding portions are formed on each of the two reticles so as to overlap each other at the patterns P, 5, Po, and prt. The width ΔD of the overlapping light shielding part is the pattern PrS.
, P rb and Pl line widths are determined. As is clear here, in the conventional method, the patterns PrS, Po, Prv
In order to expose one dark line pattern corresponding to each of
The line width of each pattern PrS to P, . . . is limited by the performance of the projection lens, etc. However, in this embodiment, the width of the dark area on the pattern separated into each of the two reticles is extremely large, and is hardly affected by diffraction. Therefore, the width ΔD can be made extremely small without being limited by the performance of the projection lens, diffraction, etc., and for example, a line pattern of 0.4 μm can be created using an exposure device with a resolution limit of 0.8 μm.

In the case of this embodiment, the dimensional accuracy of the pattern image transferred onto the wafer W is determined by the alignment accuracy of the two reticles (each separated pattern), the alignment accuracy with each chip area CP on the wafer W, and the two reticles. It is conceivable that the problem may deteriorate depending on the error in creating the pattern area between the reticles.

However, alignment accuracy has been improving year by year, and we believe that there will be fewer practical problems if errors in creating the pattern area of each reticle, marking errors, etc. are measured in advance and a sequence is taken to correct the position during alignment. It will be done. Furthermore, as is clear from the pattern decomposition method shown in Figures 12 (B) and (C), the amount of light during exposure for each of the two decomposition patterns is approximately the appropriate exposure amount for both decomposition patterns. Just leave it there. Further, the resist l-layer PR may be of either a positive type or a negative type, and it is also effective to use it in combination with a multifocal exposure method.

Next, a fifth embodiment of the present invention will be described with reference to FIGS. 13(A) and 13(B). In recent years, the use of an excimer laser light source as a light source for the stepper shown in FIG. 3 has attracted attention. An excimer laser light source uses a rare gas halide (XeC2, KrF, ArF, etc.) as a laser medium.
A laser with high gain is used. Therefore, when a high-pressure discharge is generated between the electrodes in the laser duplex, strong light in the ultraviolet region can be stimulated and emitted without a special resonant mirror. In this case, the emitted light has a broad spectrum and low coherency both temporally and spatially. Such broadband light generates significantly large chromatic aberration, depending on the material of the projection lens. In order to efficiently transmit light in the ultraviolet region, projection lenses for excimer lasers are often made only of quartz. For this reason, it is necessary to make the spectral width of the excimer laser light extremely narrow, and it is also necessary to make its absolute wavelength constant. Therefore, in this embodiment, as shown in FIG. A total reflection mirror (rear mirror 201) and a low reflectance mirror (front mirror) 205 that act as a resonator are provided to slightly increase the coherency, and a mirror 20 is provided outside the laser tube 202.
1 and the mirror 205, two Fabry-Perot etalons 203 and 204 with variable t'lJ1 angles are arranged to narrow the band of the teleser light. Here, the etalons 203 and 204 are two quartz plates facing each other in parallel with a predetermined gap, and are used as a kind of band pass filter.
3 is for coarse adjustment, and the etalon 204 is for fine adjustment. By adjusting the head angle of this etalon 204, the wavelength fluctuations are successively monitored so that the absolute value of the wavelength of the output laser light becomes a constant value. Feed pack control.

Therefore, in this embodiment, the configuration of the excimer laser light source and the axial chromatic aberration of the projection lens are actively utilized.
By optically moving the best imaging plane up and down, a multi-focal exposure method is performed. That is, a certain chip area CP
When exposing, the etalon 2.0 inside the excimer laser light source
4 or 203 is irradiated with an excimer laser (pulse or the like) while shifting the head angle by a predetermined amount from the angle required for absolute wavelength stabilization. When the head angle of the etalon is shifted, the absolute wavelength shifts slightly, so the position of the best imaging plane changes in the optical axis direction in response to the axial chromatic aberration of the projection lens. Therefore, if the head angle of the etalon is changed discretely or continuously during exposure with 50 to 100 pulses of excimer laser, the same multi-focus exposure method can be used without any mechanical movement between the reticle and wafer. can be implemented.

FIG. 13(B) shows another configuration of a similar excimer laser, in which a reflective diffraction grating 206 as a wavelength selection element is provided in place of the rear mirror 201 so as to be tiltable. In this case, grating 206
is used for coarse adjustment when setting the wavelength, and etalon 204 is used for fine adjustment. For the multi-focus exposure method, by tilting either the etalon 204 or the filtering 206, the oscillation wavelength changes and the best image plane moves up and down.

As mentioned above, when using an excimer laser, the image plane (focal position) can be changed using the next physical step called chromatic aberration, but chromatic aberration includes longitudinal chromatic aberration (axial chromatic aberration) and lateral chromatic aberration (lateral chromatic aberration). There are two, and each can occur simultaneously due to a change in wavelength. Since lateral chromatic aberration means that the projection magnification is distorted, it is necessary to correct it to a negligible degree. Therefore, for example, in the case of a projection lens that is telecentric on both sides, a field lens group (correction optical system) for maintaining telecentricity provided closest to the reticle in the projection lens is configured to move up and down in the optical axis direction, and the etalon 20 By moving the field lens group up and down in synchronization with the inclination of the lens, chromatic aberration of magnification can be corrected.

Further, even if the image forming correction mechanism 18 shown in FIG. 3 is used in conjunction with the system to apply an offset to the control pressure within the projection lens 16, the lateral chromatic aberration (magnification error) can be similarly corrected. .

Next, another sequence of the multifocal exposure method described above will be described as a sixth embodiment.

For this sequence, the stepper shown in FIG. 3 is equipped with a differential interferometer for measuring the yawing of the wafer stage 26, and two parallel interferometers are arranged at regular intervals on the movable mirror 28X or 2By. A length measurement beam is projected, and a change in the optical path difference between the two length measurement beams is measured. This measured value corresponds to the minute rotational error amount that occurs during movement of the wafer stage 26 or after stepping.

Therefore, first, for all chip areas on the wafer W, 1
Exposure is performed sequentially using a step-and-repeat method at two focal positions. At this time, the amount of yawing of the wafer stage 26 is measured and stored during exposure of each chip area.

Then, by changing the height of the Z stage 26Z or shifting the wavelength of the excimer laser beam, exposure is sequentially performed from the first chip area in the same step-and-repeat manner at the second focal position. At this time, the amount of yawing when stepping to each chip area is compared with the previously stored yawing amount when exposing the chip area, and if the difference is within the tolerance value, exposure is performed as is. If the comparison results show a large difference, either correct the rotation with a θ table that holds the wafer W and slightly rotates it, or adjust the θ table that holds the reticle.
Correct by rotating the table.

At this time, it is preferable to monitor the positional deviation between the reticle and the chip area in the X1y direction using the alignment system 12 or the like using a Gui-Pie-Die method, and perform alignment (correction of positional deviation) in real time. That is, the alignment error in the X1y direction is the mark WM attached to the chip area, ~
While detecting WM4, reticle mark RM, ~RM,
The reticle stage 6 or wafer stage 26 is servo-controlled so that the alignment error becomes zero, and at the same time, the rotation of the reticle or wafer is corrected based on the yawing measurement value from the differential interferometer.

With such a sequence, the alignment time for each chip region is shortened, and the decrease in overlay accuracy due to errors in chip rotation and wafer rotation can be ignored.

Furthermore, since the yawing amount of the wafer stage is memorized, even when using the multi-focus exposure method from the initial exposure (first print), there is no reduction in the accuracy of the overlapping exposure using the disassembled reticle.

As described above, in this embodiment, the focus position is not changed each time each chip area is exposed, but is changed only when the first exposure of one wafer is completed, so that throughput can be expected to be improved.

The embodiments of the present invention have been described above, but since each of the separated patterns has a different backbone shape, the image intensities also inevitably differ. Therefore, the appropriate exposure amount may differ for each separated pattern. For each of the separated patterns, the transmittance of the pattern area of the reticle may be measured to determine the appropriate exposure amount for each separated pattern.Also, the numerical aperture of the imaging light beam during projection exposure may be determined. Reducing the size also helps increase the depth of focus,
The numerical aperture of the imaging light beam can be adjusted by providing a temporary variable aperture diaphragm in the tiEP of the projection lens, and by changing the size of the secondary light source image in the illumination optical system using an aperture, variable magnification optical system, etc. . Furthermore, the light flux passing through the pupil BP may be restricted to a ring shape (zonal shape) with an aperture as shown in Fig. 14, or the secondary light source image may be formed into a ring shape whose diameter and width can be varied or switched. Good too.

〔Effect of the invention〕

As described above, according to the present invention, the multifocal exposure method can be applied to patterns to which it has been difficult to apply the same method in the past. Further, since the spatial frequency of the pattern can be reduced, even when the focus position is not changed, it is possible to form a finer pattern.

Further, by performing multiple exposure by changing the wavelength using excimer exposure or the like, the selection of methods for expanding the depth of focus is expanded.

These are powerful methods for solving the physical limitations of how to increase the depth of focus in sub-0.5 μm lithography using light.

Furthermore, in recent years, a method for dividing a reticle has been developed in which a sub-space mark, etc. is added to each pattern, and this solves the problem of space-related difficulties in inserting sub-space marks along with the main pattern on the same reticle. It also becomes.

[Brief explanation of the drawing]

Figure 1 is a diagram schematically representing the method of the present invention, Figure 2 (
A) and (B) are diagrams showing how diffracted light is generated in a line-and-space pattern and a thinning pattern, and Figure 2 (C) is a simulation of the image intensity distribution for a line-and-space pattern. A graph showing the results,
Figures 2 (D) and (E) are graphs showing simulations of image intensity distribution in the case of a thinning pattern, Figure 2 (F)
is a graph showing the simulation results obtained by superimposing the image intensities of FIGS. 2(D) and (E), FIG. 3 is a perspective view showing the configuration of a stepper suitable for carrying out the present invention, and FIG. 4 is a projection of the stepper. Diagram showing the state of image formation in the optical system, No. 5
6, 7, and 8 are diagrams each explaining the pattern decomposition method of the method of the present invention, and FIG. 9 is a flowchart diagram explaining one exposure procedure using the method of the present invention, and FIG. FIG. 10 is a diagram for explaining the pattern decomposition method according to the second embodiment, FIG. 11 is a diagram for explaining the pattern formation method according to the third embodiment, and FIG. 12 is a diagram for explaining the pattern decomposition method according to the fourth embodiment. FIG. 13 is a diagram showing the configuration of a laser light source suitable for carrying out the exposure method according to the fifth embodiment, and FIG. 14 is a diagram showing an annular filter for adjusting the numerical aperture of the imaging light beam. FIG. [Explanation of symbols of main parts] R, R,, R1, R,, R,... Reticle, W...
Wafer, CP...shot area, PA,,PB...whole pattern, PTA+, PTA! , PTAz ... PA decomposition pattern, PTB, , PTBt, PTB. ...PB decomposition pattern, 2.. Light source section, 4.. Illumination optical system, 6.. Reticle stage, 16.. Projection lens, 18.. Image formation correction mechanism.

Claims (1)

Scope of Claims: (1) A method of irradiating a mask with a geometric pattern formed by a transmitting part and a shielding part for a predetermined energy ray, and exposing the pattern to a sensitive substrate, The entire pattern to be formed on the sensitive substrate is divided into a plurality of decomposed patterns according to the local shape of the pattern or the density of the pattern, and each of the decomposed patterns is formed on a mask, and then the plurality of decomposed patterns are separated. An exposure method characterized by sequentially aligning each of the decomposed patterns on the sensitive substrate and exposing them to light in an overlapping manner. (2) In the overall pattern, when a plurality of linear patterns formed by the transparent parts and shielding parts are formed substantially parallel to each other at predetermined intervals, two adjacent linear patterns among the linear patterns formed by the transparent parts (3) In the overall pattern, the linear pattern formed by the transparent part or the shielding part is divided into different decomposition patterns. Claim 1, wherein when the linear pattern is bent at a predetermined angle, the bent linear pattern is divided into two parts at the bending part and included in different decomposition patterns. or the method described in paragraph 2. (4) When superimposing and exposing the plurality of separated patterns, an exposure apparatus equipped with a projection optical system that forms an image of each separated pattern in a predetermined image plane is used, and when exposing each separated pattern, 2. The method according to claim 1, wherein the exposure is performed by relatively displacing the pattern image formation surface of the projection optical system and the sensitive substrate in the optical axis direction of the projection optical system. (5) The exposure apparatus has an irradiation means for irradiating the mask with an energy beam having a specific wavelength, and the projection optical system has an axis of a predetermined amount for a wavelength that is shifted from the energy beam of the specific wavelength. determined to produce upper chromatic aberration,
5. The method according to claim 4, wherein the relative displacement of the pattern imaging surface and the sensitive substrate in the optical axis direction is performed by changing the wavelength of the energy beam.