JPS6224624A - Exposure method and exposure system - Google Patents

Abstract

PURPOSE:To alleviate drop of aligning accuracy by distortion between different exposure apparatuses by comparing bidimensional distortions respectively of the region to be exposed by the one exposure apparatus and exposed image by the other exposure apparatus and then relative shifting the aligning position from the design position so that error may be minimized. CONSTITUTION:When a wafer W is placed on a stage ST1 of stepper A and the first layer is exposed on the basis of shot arrangement data for design, the main controller CNT1 stores the position data of centers C1, C2,...of each shot as the shot arrangement data. The wafer W exposed with distortion is placed on the stage ST2 of stepper B after the process E and it is stored in the main controller CNT2. A pattern image for second layer is aligned on the first layer and it is exposed by stepping the stage ST2 in accordance with the shot arrangement data. Therefore, the main controller CNT2 inputs the data DSD of projection lens PL1 from the main controller CNT1 and obtains optimum value of displacement so that a matching error between the projected exposed image by the stepper B and region to be exposed on the wafer W can be minimized.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field of the Invention) The present invention provides a method for overlapping exposure of different glabella areas on a photolithography substrate (semiconductor wafer, etc.) using a plurality of different exposure apparatuses (one of which -Relates to an exposure method and an exposure system when an and-repeat type exposure device (so-called stepper) is used in combination.

(Background of the invention) In recent years, reduction projection exposure equipment has been used in the production of VLSIs.
It has achieved great success in improving productivity.

This type of projection exposure apparatus has improved overlay accuracy during overlay exposure between different eyebrows on a semiconductor wafer as the line width of the pattern on the reticle to be projected exposed becomes finer and the pattern itself becomes more highly integrated. is required to improve.

This is especially true between different devices. Recently, various types of projection exposure equipment with different numerical apertures, magnifications, or exposure areas (projection field of view) have appeared, and in VLSI manufacturing factories, the required resolution and throughput are taken into consideration during the process of manufacturing one VLSI. Therefore, it has become common to use separate devices for different exposures between the eyebrows. The so-called mix and match is the mixing of exposure devices with different projection magnifications and projection fields, or devices with different exposure methods such as refractive systems or reflective systems in the photolithography process of semiconductor manufacturing. Matc
In h), overlay accuracy becomes particularly important. Distortion (image distortion) of the projection optical system is cited as a factor that affects overlay accuracy. In general, even if projection optical systems have the same structure, the distortion characteristics (aberration curves) differ slightly from one projection optical system to another, and even more so when projection optical systems with different magnifications and fields of view have different distortion characteristics. There are different things. Therefore, even if mix-and-match is performed using such devices, sufficient overlay accuracy may not necessarily be obtained. This causes serious problems such as decreased productivity due to poor overlay.

(Objective of the Invention) An object of the present invention is to provide an exposure method and an exposure system that solve the above problems and reduce the reduction in overlay accuracy due to distortion between different exposure apparatuses.

(Summary of the Invention) The present invention provides a method for mixing and exposing at least two different exposure devices (one is a stepper) when stacking and exposing a large number of layers on a photolithography substrate (for example, a wafer).
A method and system for performing and-matching, the method and system comprising: two-dimensional distortion of a region formed on a wafer by one exposure device and which becomes an exposed portion during overlapping exposure (the optical system of one exposure device); (including distortion) and the two-dimensional distortion of the exposed image by the other exposure device during overlay exposure, and The technical requirement is to perform step-and-repeat exposure using a stepper so that the overlapping position of the exposed area and the exposed image is relatively shifted from the designed position.

(Example) Figure 1 shows two reduction projection exposure apparatuses (hereinafter referred to as steppers) used in a part of a specific production line in a photolithography process suitable for carrying out the exposure method of the present invention. FIG. 2 is a perspective view showing a schematic configuration. In this example, a semiconductor wafer (hereinafter referred to simply as a wafer) W exposed by stepper A is processed in process E including development treatment, and then a new photoresist is applied and overlapping exposure is performed by stepper B. A case in which this is done will be explained. The stepper A is provided with a projection lens PL having a reduction magnification of, for example, 1/2.5 and a maximum projection field of view on the wafer W of 30×30n squares (circular field with a diameter of about 42 fl). In this embodiment, the reticle RI is mounted so that the optical axis AX of the projection lens P L + passes through the center of the pattern area PA. The pattern area PA of the reticle R3 is divided into four parts with its center as the origin, and the same circuit pattern is formed in each quadrant. The size of the projected image on the wafer W in each quadrant is determined to match the size of the exposed image on the wafer W of the stepper B. Now, the wafer W mounted on the stepper A is vacuum-adsorbed onto a two-dimensionally moving stage ST, and exposed by a step-and-repeat method.

Stage ST.

is moved straight in the X direction by a drive motor MX, and is moved straight in an X direction perpendicular to the X direction by a drive motor MY. These motors MX, , MY + are controlled by a main controller CNT, and move the stage ST to a desired position or stop it. Although not shown, the motors MX, , MY, feedback control. In addition, the projection lens PL+ has
A pressure regulator BCI is provided for adjusting (correcting) the projection magnification by a minute amount. Pressure regulator BC,
The pressure regulator BC adjusts the pressure in the air chamber between the plurality of lenses that make up the projection lens PLl, and finely adjusts the projection magnification of the projection lens PLt itself from changes in the refractive index depending on changes in gas pressure. , are disclosed in detail in, for example, JP-A-60-28613, JP-A-60-78454, etc., and therefore their explanation will be omitted here. This pressure regulator BC is also centrally controlled by the main controller CNT.

On the other hand, the basic configuration of stepper B is also similar to that of stepper A.
, and is provided with a projection lens PL, a stage ST2, motors MX2 and MYt, a pressure regulator BC2, and a main controller CNT.

However, it is assumed that the projection lens PLz has a reduction magnification of 1/10 and a maximum projection field of view of 1010×10 angles (circular field with a diameter of about 14 cranes). At that point, reticle R2
To explain in detail the dimensional relationship between reticle R3 and reticle R3, stepper B
1 shot by stepper A is 4 shots by stepper A.
Since it will be overlapped with one of the divided pieces, the dimensions of one shot by stepper B will be the maximum of 1010 x 10.
If it is a corner, then the dimension of one shot made by stepper A is 2
It becomes 20×20+n angle, which is smaller than the maximum projection area of the projection lens PL in this embodiment.

Therefore, the dimensions of the pattern area PAI of the reticle R are 50×50 angles based on the magnification of the projection lens PL+.

There are four pattern areas of about 25 x 25 squares within this 50 x 50 mm square.

Now, between the main controller CNT and CN Tz,
Data DSD regarding distortion (image distortion) characteristics and the like of the projection lenses PL+ and PL2 are exchanged. This may be performed online between the main controllers CNT and CNTZ, or may be performed online via a magnetic disk or the like. In addition, each main controller CNTl
, CNTZ also stores step-and-repeat positioning information of the stages ST, , ST, ie, design data for the shot array on the wafer W. The main controllers CNT+ and CNTz include a computer, etc., and control the actual stepping position by a minute amount with respect to the designed shot arrangement in order to maximize the matching accuracy of the two exposure devices. Calculation is performed to correct the projection magnification or to correct the projection magnification by a minute amount.

In this embodiment, in order to simplify the following explanation, it is assumed that the stepper A exposes the first layer on the wafer W, and the stepper B performs overlapping exposure of the second layer, and the projection lens PL of the stepper B.

It is assumed that the distortion of is ideally zero.

First, wafer W is transferred to stage ST of stepper A.

The first layer is exposed using a step-and-repeat method based on designed shot array data. This situation is shown in FIG.

FIG. 2 is a plan view showing an arrangement of shots formed on a wafer W based on shot arrangement data stored in the main controller CNT. In FIG. 2, one shot by stepper A is represented by a solid square area,
St, St, 33--1-1S b----
The code is S ts. The center of each shot is Ct
, Cz----. Each shot is divided into four identical circuit pattern regions, which are indicated by broken lines. Stage ST steps in the direction exactly opposite to the arrow shown in FIG. 2 to expose each shot. The main controller CNT is the center of each shot '7 C, , c2 --
- One position information is stored as shot array data.

The maximum projection field of the projection lens PL is 3 on the wafer W.
Since the size of the reticle R is 0 x 30 m, the blind is opened to illuminate only the pattern area PA of the reticle R using a reticle blind (illumination field diaphragm) not shown. FIG. 3 is a projection chart of ideal lattice points exaggerating an example of the distortion characteristics of one shot formed on the wafer W in this manner. In the same figure, the orthogonal coordinate system XyOX
The intersections of equally spaced straight lines shown by broken lines parallel to the axis and the y-axis are ideal grid points, and the intersections of curved lines shown by solid lines represent projection points of the ideal grid points. This chart represents the entire area of the maximum projection field of the projection lens PL. Here, the origin P0 of the coordinate system xy is the projection lens PL,
The optical axis AX of is determined to pass through, and is also the center CI, C, of each shot. For example, the first coordinate system xy
Focusing on the quadrant, ideal lattice points Pb+, Pbz, Pb:+, aligned in the radial direction at an angle of 45 degrees from the origin P0,
For pb, the projection points P at , P ag of each point,
P as and P a4 are substantially shifted in the radial direction. Here, when the actual projection point is located closer to the origin P0 than the ideal grid point, the distortion amount at that position (image high point) is negative, and in the opposite case, it is positive. In the case of a general projection lens, the amount of distortion ideally occurs in a point-symmetrical manner with the optical axis (origin P,) as the center.

Now, the wafer W exposed with the above-described distortion is placed on the stage ST2 of the stepper B after the process E. Main controller CN of stepper B
Shot array data is stored in Tz such that one square area indicated by a broken line in FIG. 2 is one shot. Therefore, in terms of design, by stepping the stage ST2 according to the shot array data, a pattern image for the second layer can be superimposed and exposed on the first layer. However, if the step-and-repeat method is used to simply superimpose the parts based on the design values, the best matching accuracy cannot be obtained due to the distortion of stepper A, even if the distortion of stepper B is zero. There is also.

Figure 4 shows stepper B performing steps according to the design values.
FIG. 3 is a chart showing an exaggerated state of overlapping in the case of AND-repeat exposure. The projection points of ideal grid points by stepper A are represented by solid lines, and the projection points of ideal grid points by stepper B are represented by broken lines. The method of determining the coordinate system xy is the same as in Fig. 3, and l0 by stepper B
The four shots of XIO Tsurugaku are C1, C1□, and C.
28, shows how it is driven centering on C2□.

Since the distortion of stepper B is assumed to be zero,
The four dashed shots by stepper B all represent ideal grid points. Therefore, in FIG. 4, focusing on the shot in the first quadrant of the coordinate system xy, for the pattern area exposed by stepper A,
It can be seen that the projected exposure image by the stepper B is generally shifted by a minute amount in the direction of diagonal radiation 456 away from the origin P0. Although this shift is relative, in this example, stepper A was used first to expose a large area, so the projected exposure image by stepper B was shifted based on the pattern already formed on wafer W. Think of it as something that exists. This amount of deviation is due to the difference in distortion characteristics between the two projection lenses, and is actually an extremely small value (for example, 0.2 μm or less).

However, at the point where the difference in distortion characteristics is the greatest, the deviation will inevitably become larger, and there is a possibility that a matching failure will occur in a part of the exposed area.

Therefore, the projected exposure image and the wafer W are
In order to minimize the matching error due to distortion with the exposed area above, the main controller CNT inputs data DSD regarding the distortion of the projection lens PL from the main controller CNT, and moves the projection lens PL from the designed stepping position. Find the optimal solution such as the amount of shift or the amount of magnification correction. The distortion characteristic DSI of the projection lens PL is stored in the main controller CNT as a curve obtained by plotting the distortion 11Ds against the image height position for a large number of image high points, as shown in FIG. 5, for example. This characteristic D S + corresponds to the chart shown in FIG. 3, and is checked together with the distortion inspection of the projection lens PL during device manufacture. or,
As disclosed in Japanese Patent Application Laid-Open No. 59-94032, an image of a reference reticle provided with a plurality of cross-shaped marks was projected for distortion inspection, and the position of the image of the cross-shaped marks was set on a stage. Distortion characteristics can also be determined by detecting with a photoelectric element with a slit. In this case, since the distortion characteristics can be measured at any point during the operation of the stepper, there is an advantage that it is possible to respond to changes in the distortion characteristics over time.

In FIG. 5, the horizontal axis represents the image height (unit: fl), and the vertical axis represents the amount of distortion (the amount of deviation from the ideal grid point). Such characteristic curves tend to be almost the same for general projection lenses.

However, the maximum positive and negative values of the distortion amount differ depending on the projection lens, and in some cases, the origin P0 is located above the image high axis.
, center C8, and projection point Pbs so that the image high axis is a diagonal 456 straight line, the amount of distortion becomes negative at each point within the range up to projection point Pb3. Here, if the image height is r1, the distortion curve is f(r), and the magnification adjustment coefficient is C, then the deviation Δ(r) of the projection point from the ideal grid point (ideal image high point) is as shown in equation (1). expressed.

Δ (r)=f (r)+C−r −−−−−−(
1) Now assume that the projection exposure area by stepper B has a relationship as shown in FIG. In Figure 6, the coordinate system x
The origin P of y (0,0) is the optical axis position of stepper A, the square area indicated by the broken line is the exposure area when stepper B steps according to the design value, and its center C0 is the coordinate on the shot array data. It is set to the value (Xc, Yc). Also, the square area indicated by the dashed line is x, y from the design value.
This is the exposure area when the wafer W is shifted by ΔXc and ΔY in the directions, and its center CI+' is the coordinate value (Xc
+ΔXc, Yc+ΔYc). Shift amount ΔX
c and ΔYc are wafer position correction amounts for obtaining the best matching accuracy.

Corresponding to the above equation (1), the deviation from the ideal position of the exposed image by each of steppers A and B is expressed as a vector (ΔAz,
Expressed as ΔA, ), (ΔBX, ΔB,), stepper A is expressed by equation (2), and stepper B is expressed by (3).
) is expressed by the formula.

Here, Xc'=Xc+ΔXc, Yc'=Yc+ΔYc
shall be.

(ΔAX, ΔAy) = (f, (heT]ko2) + C1 iris Fko]7) (ΔB
X, ΔB, )= However, the function f8 is the distortion curve of stepper A, the function fb is the distortion curve of stepper B,
C-1Cb are magnification adjustment coefficients of stepper A1B, respectively. Furthermore, if we consider the distortion of stepper B to be zero as assumed earlier, then the function fb in (3)
The term becomes zero.

Now, from equations (2) and (3), the difference in distortion (ΔX, Ay) between the final superimposed images is expressed as equation (4).

(Δx1Δy) = (ΔAx, ΔAy) − (ΔBx, ΔB )') ----(4) Furthermore, (4
), if the maximum absolute value of the difference ΔX in the overlapping exposure area is Dx, the maximum absolute value of the difference Δy is Dy, and the maximum value of the scalar amount of the vector (ΔX, Ay) is 6, then ( 5), (6), and (7).

Dx=MAX(lΔxl) -----(5)D)
r=MAX (lΔy l ) -----(6) 100
= M A X <RτTp] and Ay)")
(7) Therefore, the computer in the main controller CNTz,
Regarding multiple points within the projection exposure area by stepper B,
By changing the shift amount ΔXC1ΔY0 little by little, the difference ΔX,
Repeat the calculation of Ay to find the maximum values Dx, Dy, and 100,000, and then minimize any one of Dx, Dy, and 100, or minimize the larger of Dx and l)y, Alternatively, the amount of shift ΔXC%ΔY is determined so that both Dx and Dy are as small as possible. At this time, it is possible to find the optimal solution for the magnification adjustment coefficients C1 and Cb at the same time, but in this example, the above calculation is performed with both c, and Cb fixed at their initial values (zero), and only the shift amount is calculated. After asking,
Suppose that C-1Cb is determined.

By this calculation, in this embodiment, the characteristic D shown in FIG.
From S, it is best to shift the four shots exposed by stepper B within one shot exposed by stepper A by a minute amount toward the origin P0, as is clear from Fig. 4. A solution like this is obtained. FIG. 7 shows the situation when overlapping exposure is performed by this shift. FIG. 7 is the same chart as FIG. 4, and here only the exposure image of stepper B in the first quadrant of the coordinate system xy is shown. In the same figure, C8 and C1'' are the same as each center shown in FIG. 6.Compared with the state in FIG. It can be seen that the matching accuracy has improved overall.It should be noted that the exposure area by stepper B is
, the third (scribe) line is formed, and the amount of shift is 1 μm or less, so the actual pattern area does not protrude into the pattern area in the adjacent quadrant.

FIG. 8 is a diagram showing the state shown in FIG. 7 in terms of distortion characteristics, and the vertical and horizontal axes are the same as those in FIG. 5. Distortion characteristic DSt is stepper B
The projection lens PL is expressed as follows, and in this example, since it is assumed that the distortion is ideally zero, the characteristic D S
z is represented by a straight line parallel to the image high axis. In this figure, the deviation between the characteristic D S z and the image high axis corresponds to the amount of deviation described above. As is clear from FIG. 8, within the exposure range from the origin P0 to the point pba, at approximately the center CII'', the discotic characteristic DS is obtained.

The difference in the amount of distortion between C1+ and DS2 becomes zero, and the relative difference in the amount of distortion before and after the center C1+'' is almost equally distributed above and below the center C1+'' with respect to the distortion DS.
In the case shown in the figure, the difference ΔD in both characteristics on the left side of the center C1l' is maximum at the origin P0, and the difference ΔD on the right side of the center C1'; . By calculating the equations (2) to (7) above, in this example, the absolute value of the difference ΔD and the difference ΔD
The amount of shift of the characteristic DS is determined so that the absolute value of b is equal to the absolute value of b. For example, in order to make the characteristic D1 times 1ΔDh l, so that the absolute value of the difference between the two characteristics becomes the maximum value, it may be set as 1ΔD, l=lΔDb1.

Now, up to FIG. 8, the values can be obtained by calculating equations (2) to (7) while fixing the magnification adjustment coefficients C, . . . Cb to the initial values (zero).

Next, two characteristics DS are determined as shown in FIG.

and DSz, projection lens PL of stepper B.

By adjusting the projection magnification of , a solution that provides the best matching accuracy is found.

Here, by substituting the shift amounts ΔXc and ΔYc determined in equations (2) to (7) above, and further setting C1-〇, C5
By changing little by little, the maximum absolute value DX% Dy% D
and calculate any one of them, or D x s
Either D and y are minimized, or with the distortion characteristic DS as a reference, the characteristic DS becomes a characteristic D S zo tilted around C11' as shown in FIG. The amount of inclination corresponds to the magnification adjustment coefficient C5. In FIG. 9, the horizontal axis represents the image height (n), and the vertical axis represents the amount of distortion. Further, in the figure, a straight line 1 represents the characteristic D S t when the magnification adjustment coefficient Cb is zero. Therefore, the main controller CNT changes the characteristic DS, while changing the coefficient Cb.
and DSL' at a minute interval (for example, 0
．． 5 mm), and find the maximum value among the absolute values of each difference.

Then, a coefficient Cb is determined such that the maximum value becomes the minimum at the high points of multiple images up to the exposure range pb3. Thereafter, the main controller CNT2 controls the pressure regulator B based on the determined coefficient Cb.
C, to obtain a characteristic DSz' as shown in FIG. Here, since the characteristic D321 is adjusted to the right, the reduction ratio of the projection lens PL is slightly larger than the designed value, and the projected image of the pattern area PA2 is smaller than the designed dimension. It will be slightly shrunken.

When the magnification of stepper B is adjusted in this way and overlapping exposure is performed, the matching accuracy is greatly improved as shown in FIG. 10. Figure 10 is a projection chart of ideal lattice points similar to Figure 7, and corresponds to Figure 7.
Only the quadrants are shown. In general, the distortion characteristics of this type of projection lens are ideally point symmetrical with respect to the optical axis. Second
, the magnification of stepper B may be adjusted by the same amount for the third and fourth quadrants. However, since the shifting direction during stepping by stepper B differs for each quadrant, the main controller CN Tz calculates and stores the shifting direction and shift amount for each quadrant. Given that the shift direction and amount of each quadrant are expressed by vectors Δ1, Δ2, Δ1, Δ4, the shot position determined by the data is expressed as vectors Δ3, Δz1Δ1,
Correct by either Δ4. The situation is shown in FIG. In Figure 11, stepper A is S, S
There are four: o, Sl, 5, and 1.

Now, stepper B is in the shot area S, ,S, ,S, ,
Assuming that the upper row of S4 is exposed sequentially from left to right, the vector Δ2 for shot S1m, the vector ΔI for shot Slb, the vector Δ2 for shot 84&, and the vector Δ for shot S4b. , the stepping is performed with each position correction. Of course, the projection magnification is adjusted based on the determined coefficient C5 before starting the exposure operation.

In this embodiment, the distortion of stepper B is ideally zero. However, in an actual projection lens, it is impossible for the distortion to be zero, and the distortion curve will change more or less in some way. exists. For this reason, as shown step-by-step in Figures 4, 7, and 10, it is not possible to expect significant effects such as improvement in matching accuracy. Improvements in accuracy can be expected. Further, in this embodiment, magnification adjustment is also performed, but this is not necessarily necessary in the present invention. Further, contrary to the present embodiment, it is also conceivable that exposure is performed by stepper B and then exposure is performed by stepper A. In that case, the same effect can be obtained by simply adjusting the magnification of the stepper B in advance and correcting the shot position from the designed position as shown in FIG. 10 before exposing. In addition, as a method of adjusting the projection magnification, if the object (reticle) side of the projection lens is a non-telecentric optical system, the optical distance between the reticle and the projection lens (
(optical path length) may be made variable.

Next, FIG. 12 shows an exposure method according to a second embodiment of the present invention.
This will be explained with reference to FIG. FIG. 12 is a diagram showing a schematic configuration of a stepper suitable for implementing the present method. This stepper is used to perform overlapping exposure on a pattern formed in advance on the wafer W, and unlike the previous embodiment, the stepper controls the relative shift between the exposed area on the wafer W and the projected image. , is achieved by slightly moving the reticle R. In FIG. 12, the reticle R is held on a reticle stage R3T that can be moved slightly in at least the X direction and the y direction. Pattern area P of reticle R
Around A is alignment with wafer W.
For this purpose, marks RM, , RM z are formed.

On the other hand, the wafer W held on the stage ST has marks RM, RM! Marks WM, WM2 are formed so that they can be superimposed on each other. And Mark WM.

The back projection image and mark RM by the projection lens PL.

The degree of overlap with mirror 10a, objective lens 11a,
and a positional deviation detection section 12a including a photoelectric converter. Similarly, the degree of overlap between mark WM2 and mark RM is determined by mirror 10b.
, the objective lens 11b1 is detected by an alignment sensor consisting of a positional deviation detection section 12a. Main controller CN
T is mark R detected by both alignment sensors
M.

Input information according to the amount of relative positional deviation with WM,
The step-and-repeat drive motor M of the stage ST is controlled so that the positional deviation becomes zero. Of course, in order to improve the matching accuracy, the main controller CNT is necessary to perform the matching process before exposure. Multiple formations. This shot S
In each of the four regions S1, Sb, SC, and Sd, marks WM+ and WMz are formed at predetermined positions for alignment with the marks RM, , RM2 of the reticle R.

Now, during actual overlay exposure, the main controller CNT
Stage S according to the shot arrangement data stored in the
By stepping T, the projection lens PL is positioned so that its optical axis AX substantially coincides with, for example, the center CCa of the area Sa. Next, two alignment sensors detect the two-dimensional deviations of marks RM, , WMl, and mark R.
Next, we move on to an alignment operation (so-called Gui-Pai-Gui alignment) for detecting a two-dimensional deviation between M z and W M t.

In normal Gui-Pai-Gui alignment, the wafer W is aligned so that the marks RM and WM and the marks RMz and WM2 have a predetermined positional relationship based on the design.
All you have to do is slightly move it two-dimensionally. However, in this embodiment, in order to obtain the same effect as the shift shown in FIG. The drive unit 13 is controlled so as to shift the position from the correct position. When you move the reticle R in this way,
There is a possibility that the positioning accuracy will be improved compared to moving the wafer W as in the previous embodiment. For example, if the reduction ratio of the projection lens PL is 1/10, when the projected image of the pattern area PA and the area Sa on the wafer W are relatively shifted by 0.1 μm, the reticle R is moved by 1 μm, which is 10 times that amount. This is because it is fine. To perform such an operation,
For example, the mark RM on the reticle detected by the alignment sensor when the Gui-Pai-Dai alignment is completed,
, RM, can be retched by 1 μm very easily. Of course other areas sb, Sc, S
The same applies when exposing d, but the center CCb of each area
The directions in which the center point of the pattern area PA is shifted with respect to 1CCc and CCa are different as shown in FIG.

As described above, in this embodiment, when performing Gui-Pie-Die alignment, the reticle is slightly moved and the amount of shift is taken into account to reduce the difference in distortion, so matching accuracy is improved without reducing throughput. can be done. Note that when the reticle is slightly moved as in this embodiment, it is not necessarily necessary to use an alignment sensor to perform goo-pai-goo alignment. For example, when measuring the position of the stage ST with a laser interferometric length measuring device, the optical axis AX of the projection lens PL and the center CC of each area S, , Sl, 5cSSd
This means that positioning is possible within a range sufficiently smaller than the amount of shift. Therefore, an encoder, a laser interferometric length measuring device, etc. are installed to precisely read the amount of movement of reticle stage R3T from a predetermined initial position (designed position), and the movement of reticle stage R3T is servo controlled based only on these measured values. You may also do so. Of course, in this embodiment as well, it goes without saying that correction for finely adjusting the projection magnification may be added at the same time.

When the reticle is shifted as in this example, the projected pattern image itself may have an asymmetrically distorted shape. By repeating the calculation, it is possible to determine the direction and amount of shift that will provide the best matching accuracy, including its asymmetry. Furthermore, the 12th
As shown in this embodiment shown in FIGS. , which corresponds to the wafer shifting described in the first embodiment.

Next, a third embodiment of the present invention will be described. In this embodiment, a case will be described in which the dimensions of one side (or diagonal line) of projection exposure areas (shots) of two steppers used for mix-and-match are not integral multiples. Further, in this embodiment, either or both of the wafer shift and the reticle shift can be performed, but for the sake of simplicity, only the wafer shift will be described here. Now, the projection lens PL of stepper A
, the maximum projection area is 30X3 with a reduction ratio of 1/2.5
0n angle, and the projection lens of stepper B is
Projection pattern area of reticle R, 24 x 24
Reticle R2 is used with a focus on m square, and is used with stepper B.
It is assumed that the projected pattern image is narrowed down to 16 x 16 crane angles. For example, as shown in FIG. 14, a realistic usage situation in which mixed use is performed with such shot sizes (screen sizes) is that nine identical chip patterns are provided on reticle R3, and those chip patterns are used on reticle R2. This is a case where four identical chip patterns are provided to be overlapped. FIG. 14(a) shows reticle R.

FIG. 3 is a plan view showing a chip arrangement, in which nine square chip patterns 702 to CP with a - side dimension of do are arranged in a matrix. This trial R, is the projection lens PL +
is positioned so that the optical axis AX of Therefore, the size of one shot exposure image on the wafer is d+/2.5
(= 24mm) square. On the other hand, Fig. 14 (
b) is a plan view showing the chip arrangement of reticle Rt;
- 4 square chip patterns cp with side dimensions d2
,,cP, ,CPc,CP, are arranged in each quadrant of the coordinate system xy. Therefore, the pattern area on the reticle R2 is a square area with a negative side dimension d. Therefore, the size of one shot exposure image on the wafer is d.

It becomes a square of 15 (-16 m). Further, on the wafer, the dimensions of the exposed images of the chip patterns CP l'' CP g are determined to match the dimensions of the exposed images of the chip patterns CP to CP4.

Now, among these two types of reticles R,, R1,
First, stepper A moves reticle R.

The pattern is transferred onto the wafer, and then the shot arrangement of the steppers is determined as shown in FIG. The shot centers (projected points of the optical axis) by stepper A are 5C11sc2, sc3, and SC4, and their shot areas are represented by solid lines, and the shot centers (projected points of the optical axis) by stepper B are SC--SCb 1SCc1SC.
a, 5C-1SCr, SC,, SCh, SC4, and the shot area is represented by a broken line. In this way, if the size of one shot between steppers A and B is not an integral multiple, the overlapping state will differ depending on the shot between steppers A and B, as is clear from FIG. 15. In the case of this embodiment, s by stepper B
c, ,5ccSsc, ,sc.

Each of the four shots centered on SC+1SCz, SCs, and SC4 by stepper A includes each of the four shots centered on SC+1SCz, SCs, and SC4.

Also, S Cb -S Ca -S by stepper B
Each of the four shots centered on Cr5Ch is arranged equally across two shots by stepper A. Furthermore, one shot centered on SC8 by stepper B is arranged equally across four shots by stepper A. Note that the square area indicated by diagonal lines in FIG. 15 corresponds to one chip pattern (here, a projected image of CPI or CP). In this way, the exposed images to be overlapped are
When it spans multiple shots in the previous process, it is necessary to take into account how it spans and add some kind of restriction to the direction in which the wafer is shifted.

The situation will be explained with reference to FIG. The chart indicated by a broken line in FIG. 16 is an exaggerated representation of the distortion of the projected image (24fl angle) of the ideal grid point by stepper A, and the chart indicated by a solid line is the projection of the ideal grid point by stepper B. This is an exaggerated representation of the distortion of the image (16 mm square). This overlapping state is when steppers A and B are stepped as designed, and the shot center S of stepper A is
In this case, the shot center sc of stepper B is a point sc that should correspond to the center sc in the exposed image by stepper A.
, ' are significantly shifted, and each grid point is also shifted overall.

Therefore, regarding the area where the shot of stepper B and the shot of stepper A overlap, the above-mentioned (2) to (7)
Based on the formula, the direction and amount of shift that will give the best matching accuracy are determined. As a result, as shown in FIG. 17, it is found that it is advisable to slightly shift the shot center sc by stepper B in a direction closer to the shot center S01 by stepper A. FIG. 17 is a chart similar to FIG. 16, and the coordinate system is also the same. As is clear from FIG. 17, when the actual shot center sca is shifted from the designed shot position s01'' by stepper B so that it almost coincides with the point sc,',
Matching accuracy is the best.

Next, for the two shots by stepper A, the stepper
Consider the situation where one shot of B spans. For example, if we focus on the shot centered on SCb in Fig. 15, this shot equally spans the two shots in the previous process lined up in the X direction on the wafer, so the distortion characteristics of the two steppers are Since both shots are point symmetrical within the wafer, the shot center SCb can be moved from the design position (on the center line of the boundary between the two shots on the wafer) to
When shifted in this direction, matching accuracy inevitably deteriorates compared to before shifting. Therefore, in such a case, the only option is to shift the shot center SCb in the y direction, and the amount of shift to obtain the best matching accuracy is also the same as in (2) to
Based on equation (7), by changing the shift amount only in the y direction little by little, find the maximum value of the absolute value of the deviation of the two distortion characteristics, and when the deviation is the minimum among these maximum values, All you have to do is find the amount of shift.

Next, consider a situation in which one shot from stepper B evenly spans four shots from stepper A.

In this embodiment, as shown in FIG.
, occurs for one shot centered on . In this case, the shot center SC.

If shifted from the designed position, for example, the center SC.

Even if the matching accuracy improves for overlapping areas in the shifted direction, the matching accuracy decreases for overlapping areas in the opposite direction to the shifted direction. Therefore, there is no need to shift the shot center SC from its designed position.

Therefore, by summarizing each of the above-mentioned states, it is possible to create a map of the shifting direction as shown in FIG. In FIG. 18, the same parts as in FIG. 15 are given the same reference numerals. As shown in FIG. 17, each of the shot centers SC, , scc, sc, , SC, by the stepper B is shifted approximately diagonally by 4 with respect to the xy axis, and the shot center SCb is
, SC4, SCf, and SCh are shifted by ΔB1%ΔB2, ΔBl, and ΔB4 in the x1y direction. However, ΔAl, ΔAz, ΔA3, and ΔA4 are not necessarily in the 456 direction, but as a general tendency, they are often in the 45'' direction.The shot center SC, is not shifted. As explained in the example, each of shot centers SC3, SC2, SC3, and S04 by stepper A is
As shown by the arrow in Figure 8, matching accuracy can be further improved by arranging the four shots by stepper A so that they are shifted by a minute amount from the designed shot position in the direction of the intersection (coinciding with the shot center SC0). may be improved.

It goes without saying that in this embodiment as well, by finely adjusting the magnification of stepper B1 or stepper A, the matching accuracy can be further improved.

Note that each shift indicated by an arrow in FIG. 18 includes shifting the reticle by a minute amount, as in the second embodiment. Therefore, FIG. 19 shows how the reticles are superimposed when the reticles are further shifted from the state shown in FIG. 17. In this case, since the center of reticle R2 is slightly shifted from the projection optical axis A x z of stepper B, the projected image by stepper B will be asymmetrically distorted, but the area to be overlappingly exposed on the wafer will also be distorted in the same way. Since this is accompanied by asymmetric distortion, matching accuracy improves as a result. In Fig. 19, sc,'' is the designed shot center position by stepper B (when the displacement of the wafer and reticle is zero); CC is the position of the center of the shot when the reticle is shifted (the point where the smoothness of the reticle is zero and coincides with the optical axis A A point S through which z passes
C,<or SC,') is shifted diagonally upward to the left by a minute amount. Comparing FIG. 19 with FIG. 17, it can be seen that the matching accuracy is further improved in the peripheral area of the overlapping exposure area by stepper B.

Next, a fourth embodiment of the present invention will be described based on the flowchart shown in FIG. In this example, we will simulate different combinations of the methods for improving matching accuracy by shifting the wafer, shifting the reticle, and adjusting the magnification that have been described so far, and select the method that provides the best matching accuracy. I made it.

First, the distortion data of the exposed area on the wafer to be overlaid exposed is read into the main controller CNT or the upper central management computer (step 1).
00). At this time, data on linear deformation of the wafer due to heat treatment or chemical treatment applied to the wafer is also read as distortion data. This linear deformation can be actually measured using a known wafer alignment sensor provided in the stepper. Specifically, a wafer alignment sensor is used to measure the positions of alignment marks formed at multiple design positions on the wafer for wafer alignment, and the deviation between the measured value and the design value is calculated. By finding the linear deformation (run
out). The amount of deformation varies slightly depending on the shot position on the wafer, and generally the amount of deformation is zero at the center of the wafer, and the amount of deformation in the radial direction increases toward the periphery. Distortion data resulting from this linear deformation is used for step and
It is used as the amount of shift in the stepping position of stage ST during AND repeat exposure.

Next, using the distortion data of the exposed area on the wafer (the projection field of the projection optical system of the exposure device used in the previous process) and the distortion data of the projection lens of the stepper that performs overlapping exposure, the first As in the example, the wafer is shifted by minute amounts in various directions from the designed stepping position, and the direction and amount of shift of the wafer that will provide the best matching accuracy is determined by simulation (step 101). .

Then, the evaluation result (A) at this time, for example, the σ value (average deviation) of matching accuracy is stored.

Next, similarly to the second embodiment, move the reticle to the designed position (
While shifting the reticle by minute amounts in various directions from the position where the optical axis passes through the center of the pattern area, the direction and amount of shift of the reticle that will provide the best matching accuracy are determined by simulation (step 102).

Then, the evaluation result (B) is stored.

Next, when only the magnification is finely adjusted to various values without shifting the reticle and wafer, the magnification adjustment coefficient that will provide the best matching accuracy is determined by simulation, and the evaluation result (C) is stored. (Step 103).

The direction and amount of shift, as well as the magnification adjustment coefficient, that will give the best matching accuracy are determined by simulation, and the evaluation results (D) are stored (step 1).
04).

Similarly, when reticle shift and magnification adjustment are used together, the direction and amount of shift and the magnification adjustment coefficient that provide the best matching accuracy are determined by simulation, and the evaluation results (E) are stored. (Step 105).

Next, in the case where both the reticle and wafer are shifted relative to each other without adjusting the magnification, the direction and amount of shift of the reticle and wafer that will give the best matching accuracy are determined by simulation, and the evaluation results ( F)
is stored (step 106).

More precisely, it is also necessary to perform a similar simulation in the case where the three methods of shifting the reticle and wafer and adjusting the magnification are used together.

However, in practice, after the optimal simulation result is obtained by shifting the wafer and adjusting the magnification, it is necessary to try to shift the reticle further in that state, or to obtain the optimal simulation result by shifting the reticle and adjusting the magnification. After that, it is almost sufficient to either try to shift the wafer further in that state. For this reason, in this embodiment, evaluation by simulation especially when the three are used together is omitted from illustration.

Next, select the one with the smallest σ value from among the evaluation results (A) to (F) above (step 107), and change the direction, amount, and magnification adjustment coefficient of the reticle or wafer shift for which the evaluation result is obtained. (step 108). In this case, for example, if the result is to only shift the reticle, a predetermined value is stored in the memory that stores the reticle shift vector (direction and amount), and the wafer shift vector and magnification adjustment coefficient are stored. Zero is stored in each memory.

Now, when only wafer smoothing is required, the designed shot array data is corrected to correspond to the wafer shift vector (step 109). If it is also necessary to shift the reticle in different directions for each shot,
A flag representing the reticle shift vector is set in the array data corresponding to each shot.

Although each of the arithmetic operations described above is performed by a computer that processes a large amount of data at high speed, it still takes a considerable amount of time, so it is preferable to use a computer different from the main controller CNT of the exposure apparatus.

Next, when magnification adjustment is required prior to exposure of the wafer, the pressure regulator BC is controlled according to the adjustment coefficient to correct the projection magnification to a predetermined projection magnification (step 110).

Then, the stage ST is stepped based on the corrected shot array data, and if a reticle shift flag is set in the data, the reticle is made to correspond to a predetermined shift vector according to the type of the flag. The pattern of the reticle is made to move slightly and the pattern of the reticle is overlapped with the exposed area on the wafer at that position, and the exposure is repeated (step 111).

In the manner described above, all exposed areas on the wafer are overlaid and exposed with the best matching accuracy.

As described above, by selecting one of reticle shifting, wafer shifting, or magnification adjustment, or an appropriate combination of multiple methods, distortion of the exposed area on the wafer (previous process exposure device (including the distortion of the overlay) and the distortion of the stepper for overlaying can be minimized, and the degree of freedom in selecting the most efficient method based on the configuration of the exposure apparatus is increased.

In each of the embodiments of the present invention described above, two steppers are used together, but similar effects can be obtained when any other exposure apparatus and stepper are used together. In particular, when an aligner and a stepper are used together, such as in mirror projection, when a wafer that has been exposed using an aligner is overlaid and exposed using a stepper, distortion unique to the aligner that occurs on the entire wafer surface is removed for each shot using the stepper. Since it is possible to add and overlap images evenly, matching accuracy is dramatically improved.

In any case, as long as the distortion characteristics of at least two exposure devices (one of which is a stepper) used in mix-and-match are known, the same effect can be obtained no matter what type of exposure device is combined.

(Effects of the Invention) As described above, according to the present invention, the best matching accuracy can always be obtained no matter what field size (exposure area) the overlapping exposure is performed when two exposure devices are used together. Therefore, even when aligners and steppers are used together as in the past, the matching accuracy for each shot by the stepper is significantly improved, and the productivity (yield) of semiconductor devices and the rate of non-defective products are increased. .

Furthermore, when steppers with different field sizes are used together, it is necessary to carefully control the distortion characteristics of each projection optical system from the stepper manufacturing stage. However, according to the present invention, the matching accuracy may be improved even if the steppers are manufactured by controlling the distortion characteristics of the projection optical system or the like so severely.

[Brief explanation of the drawing]

FIG. 1 is a perspective view showing the schematic configuration of two steppers used in mix-and-match according to an embodiment of the present invention, and FIG. 2 is a plan view showing shot arrangement using the step-and-repeat method, and the stepper Fig. 7 is a chart showing how the wafers are superimposed after the wafers are shifted, and Figure 8 is a diagram showing how two steppers are aligned when the wafers are shifted. A graph comparing the distortion characteristics of
Fig. 9 is a graph comparing two distortion characteristics when magnification adjustment is performed, Fig. 10 is a chart showing the overlapping state when wafer shifting and magnification adjustment are performed, and Fig. 11 is a graph comparing two distortion characteristics when magnification adjustment is performed. Diagram showing shift vectors for each shot during step-and-repeat exposure, 1st
2 is a diagram showing the configuration of a stepper suitable for the second embodiment of the present invention, and FIG. 13 is a plane showing the shot arrangement and alignment marks on the wafer placed on the stepper of FIG. 12. 14 is a plan view showing the chip arrangement of two reticles used as a third embodiment of the present invention,
Fig. 15 is a plan view showing the shot arrangement on the wafer when the two reticles shown in Fig. 14 are used, and Fig. 16 is the matching when the shots are overlapped according to the designed shot positions in the third embodiment. Chart diagram showing the situation, 1st
Fig. 7 is a chart showing the state of matching when the wafer is shifted from the state shown in Fig. 16, Fig. 18 is a map schematically showing the direction of shift according to the shot position, and Fig. 19 is the state shown in Fig. 17. FIG. 20 is a chart showing the state of matching when the reticle is shifted from
FIG. 2 is a flowchart showing a processing procedure of a system that combines various types of shifts according to the embodiment. [Explanation of symbols of main parts] A, B---Stepper (reduction projection type exposure device) Rh
, Rz --- reticle P L+ , P L2 --- one projection lens CNTr,
CNT2---Lifetime control device B C+ -B Cz -
---Pressure regulator D S D, D S+, D
Sz, D S! '---Distortion curve

Claims (5)

[Claims] (1) When using two exposure devices, at least one of which is a step-and-repeat method, during overlapping exposure between different layers on a substrate for photolithography, the image formed on the substrate by one of the exposure devices is The two-dimensional distortion of the area to be exposed during overlapping exposure is compared with the two-dimensional distortion of the exposed image by the other exposure device during overlapping exposure, and the In the step-and-repeat method, the overlapping position of the exposed area and the exposed image is relatively shifted from the designed position so that the overlapping error caused by the difference in stator is minimized. An exposure method characterized by performing exposure. (2) When the step-and-repeat method is used to perform exposure with a relative shift, a two-dimensional moving stage on which the substrate is placed is shifted from the designed position. The method described in Section 1. (3) When performing exposure using the step-and-repeat method with a relative shift, the mask attached to the step-and-repeat method exposure device is shifted from the designed position. A method according to claim 1. (4) The step-and-repeat exposure apparatus has an imaging optical system that optically projects the mask pattern onto the substrate, and performs the relative shift and also changes the projection magnification by a minute amount. 4. The method according to any one of claims 1 to 3, characterized in that the adjustment is performed by: (5) After forming a predetermined layer on a photolithography substrate using a first exposure device, a second exposure device different from the first exposure device is used to form a new layer on the predetermined layer. When performing overlapping exposure to form a layer, at least one of the two exposure devices is equipped with a step and
In a system using a repeat type exposure device, when overlapping exposure is performed with a two-dimensional distortion of an area formed on the substrate by one of the exposure devices and which becomes an exposed portion during overlapping exposure. an accuracy estimating means for comparing the two-dimensional distortion of the exposure image by the other exposure device and estimating the overlay accuracy due to the difference between the two distortions; using the accuracy estimating means, When the overlapping position of the exposed area and the exposed image is shifted relative to the designed position, and the overlay accuracy is the best, the direction and amount of the shift is determined by the step-and-repeat method. 1. An exposure system comprising: correction means for providing a correction value for a designed overlapping exposure position according to the invention.