JPH05304077A - Aligning method - Google Patents

Abstract

PURPOSE:To enable all shot regions arranged on a wafer to be enhanced in alignment accuracy throughout the surface of the wafer. CONSTITUTION:The coordinate positions of nine sample shot regions SA1 to SA9 are weighted corresponding to distances Lk1 to Lk9 between a shot region ESi to align and nine sample shot regions SA1 to SA9, and the weighted coordinate positions are statistically calculated to obtain parameters (a) to (f). The coordinate position of the shot region ESi is determined using the parameters (a) to (f), and a wafer is controlled to move in accordance with the determined coordinate position, whereby the shot region ESi is aligned with an exposure spot.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for sequentially aligning a plurality of processing regions (shot regions, chip patterns, etc.) regularly arranged on a substrate with respect to a predetermined reference position, The present invention relates to a positioning method suitable for an exposure apparatus for transferring a pattern of a mask used in a lithography process for manufacturing a semiconductor device or a liquid crystal display device onto a photosensitive substrate.

[0002]

2. Description of the Related Art In recent years, in a step-and-repeat type or step-and-scan type exposure apparatus, a wafer prober, a laser repair apparatus, etc., a plurality of chip patterns arranged regularly (in a matrix) on a substrate. Each of the areas (shot areas) has a predetermined reference point (for example, a processing point of various devices) in a stationary coordinate system (that is, an orthogonal coordinate system defined by two sets of laser interferometers) that defines the moving position of the substrate. ), It is necessary to perform extremely precise alignment. Particularly in an exposure apparatus, when aligning a substrate (semiconductor wafer, glass plate, etc.) with an exposure position of a pattern formed on a mask or a reticle (hereinafter referred to as a reticle), a defective product is generated in a chip at a manufacturing stage. It is desired to maintain the alignment accuracy with high accuracy and stability at all times so as to prevent a decrease in yield due to the above.

Usually, in the lithography process, 1
A circuit pattern (reticle pattern) of 0 or more layers is superposed and exposed, but if the alignment (superposition) accuracy between the layers is poor, circuit characteristics may be inconvenient.
That is, the chip does not satisfy the desired characteristics, and in the worst case, the chip becomes a defective product and the yield can be reduced.
Therefore, in the exposure step, an alignment mark is provided in advance on each of a plurality of shot areas on the wafer, and the mark position (coordinate value) is detected with the reticle pattern to be subjected to overlay exposure as a reference. Thereafter, wafer alignment for aligning (positioning) one shot area on the wafer with the reticle pattern is performed based on the mark position information.

There are roughly two types of wafer alignment, and one is a die-by-die (D / D) alignment method in which the alignment mark is detected for each shot area on the wafer to perform alignment. .. The other is a global alignment method in which the alignment of each shot area is performed by detecting alignment marks of only some shot areas on the wafer to obtain the regularity of the shot arrangement. At present, the global alignment method is mainly used in the device manufacturing line in consideration of the throughput. In particular, at present, for example, JP-A-61-144429 and JP-A-62-62.
As disclosed in Japanese Laid-Open Patent Application No. 84516, enhanced global alignment (E) for precisely specifying the regularity of shot arrangement on a wafer by a statistical method.
The GA) method is the mainstream.

The EGA system measures the coordinate positions of only a plurality of shot areas (three or more are required, usually about 10 to 15 pieces) selected as specific shot areas in advance on one wafer, After calculating the coordinate position (shot array) of all shot areas on the wafer using the statistical calculation process (such as the least squares method) from the measured values, the wafer stage is uniquely stepped according to the calculated shot array. It goes. This EGA method has an advantage that a measurement time is short and an averaging effect can be expected with respect to a random measurement error.

Here, the statistical processing method performed by the EGA method will be briefly described. Now, m (m
Designated array coordinates of specific shot areas (sample shots) of (≧ 3) are (Xn, Yn) (n = 1,
2, ..., M) and the deviation from the designed array coordinates (Δ
Xn, ΔYn) for a linear model, ie

[0007]

[Equation 1]

Suppose Furthermore, the actual array coordinates (measured values) of each of the m sample shots are (Δx _n , Δ
y _n ), the sum of squares E of the residuals when this model is applied is represented by the following equation.

[0009]

[Equation 2]

Therefore, the parameters a, b, c, d, e, and f that minimize this equation may be obtained. In the EGA method, the array coordinates of all shot areas on the wafer are calculated based on the parameters a to f calculated as described above and the design array coordinates. As described above, in the EGA method, the shot arrangement error on the wafer is treated as being linear. In other words, the EGA calculation is a linear first-order approximation. For this reason, there is a problem that it is not possible to deal with a local arrangement error variation on the wafer, that is, a non-linear factor. Therefore, for example, Japanese Patent Laid-Open No. 62-
As disclosed in Japanese Patent No. 2911133, at least three shot areas existing in a local partial area (block) on a wafer are designated as sample shots, and their coordinate positions are obtained. To EGA
By calculating (statistical calculation), the coordinate position (shot array) of all shot areas in the block is calculated.
A so-called blocked EGA (B-EGA) system has been proposed. The B-EGA method is characterized in that the sample shot used in the EGA calculation is changed for each shot area to be aligned. For example, three or more shot areas are designated as sample shots in order from the shot area to be aligned, and each measured value of the designated sample shots is used. This makes it possible to deal with a local variation (non-linearity) in array error on the wafer.

[0011]

However, in the prior art as described above, when the computer performs the process of selecting the sample shots used in the EGA calculation for each shot area to be aligned, the calculation amount for the process is increased. There is a problem that it becomes huge. Further, it is difficult to optimize the selection of the sample shot for each shot area in the block. Therefore, the B-EGA method can cope with a local variation in array error (non-linear distortion), but the alignment accuracy obtained from this cannot be said to sufficiently satisfy the required accuracy. Further, in the B-EGA method, since the sample shot is changed for each shot area, the number of sample shots on one wafer is significantly increased, and the processing time for one wafer becomes long and the throughput decreases. There is also a problem.

Further, when the wafer is placed on the wafer stage via the holder (holding member), if the wafer is largely warped due to, for example, heat treatment, the peripheral portion of the wafer is adsorbed by the holder, but the central portion. May float without being absorbed by the holder. Therefore, the shot area on the wafer in which the above phenomenon occurs, in particular, each shot area in the vicinity of the central portion is apparently distant from the center of the wafer as compared with the corresponding shot area on the wafer whose entire surface is adsorbed by the holder. This means a lateral shift (positional shift) relative to the direction.

Here, when the B-EGA method is applied to a wafer in which a non-linear distortion due to the above-mentioned phenomenon occurs, if it is clearly known which part of the wafer is lifted, the non-linear distortion occurs. Corresponding to the distortion, it is possible to prevent deterioration of the alignment accuracy to some extent. However, the same problem as described above (increase in calculation amount, sample shots, etc.) occurs here, and it is actually difficult to specify the floating (bulging) portion of the wafer. That is, the B-EGA method cannot optimize the block division of a plurality of shot areas on the wafer, and thus it is difficult to obtain the desired alignment accuracy even if the B-EGA method is applied.

If a nonlinear approximation using a higher-order function, for example, is applied to a wafer in which a nonlinear distortion has occurred, instead of the first-order approximation as in the EGA method, it is possible to prevent a decrease in alignment accuracy. However, in this case, the number of sample shots must be significantly increased as compared with the EGA method, and there is a problem that it takes time to measure the mark and throughput is reduced.

The present invention has been made in consideration of the above points, and the number of sample shots can be small and the amount of calculation can be reduced even for a wafer having a local array error (non-linear distortion). It is an object of the present invention to provide a positioning method capable of aligning all shot areas with high accuracy and at high speed with respect to a predetermined reference position while suppressing them.

[0016]

In order to solve such a problem, in the first alignment method of the present invention, the movement position of the substrate is defined for each of the plurality of processing regions regularly arranged on the substrate. The coordinate positions on the stationary coordinate system of at least three processing regions selected in advance as the specific processing regions among the plurality of processing regions on the substrate prior to alignment with respect to a predetermined reference position in the stationary coordinate system. To measure.
Further, for each processing area on the substrate, the processing area (center point)
And at least three specific processing areas (sample shots)
Each of the coordinate positions on the stationary coordinate system of at least three specific processing regions is weighted according to the distance between each of them, and the plurality of weighted coordinate positions are statistically calculated (least square method, or The coordinate position of each of the plurality of processing regions on the substrate on the stationary coordinate system is determined by performing a simple averaging process or the like).

In particular, in the first alignment method, when determining the coordinate position of the arbitrary processing area on the substrate on the stationary coordinate system, the specific processing area having a shorter distance to the processing area is
The alignment data (coordinate position) is heavily weighted. Further, the weight given to each alignment data of the specific processing region is changed according to the deformation state of the substrate, that is, the non-linear distortion amount.

Further, in the second alignment method of the present invention, each of the plurality of processing regions regularly arranged on the substrate according to the designed arrangement coordinates defines a stationary coordinate system for defining the moving position of the substrate. Prior to performing alignment with a predetermined reference position in the above, the coordinate positions on the stationary coordinate system of at least three processing regions selected in advance as the specific processing regions among the plurality of processing regions are measured. Further, for each processing area on the substrate, the distance (first information) between the processing area and a predetermined point of interest defined in advance on the substrate, and the point of interest and each of at least three specific processing areas By weighting each of the coordinate positions on the stationary coordinate system of at least three specific processing regions according to the distance (second information) between them, and statistically calculating the weighted coordinate positions, It was decided to determine the coordinate position on the stationary coordinate system of each of the plurality of processing regions on the substrate.

In particular, in the second alignment method, when determining the coordinate position of the arbitrary processing area on the substrate on the stationary coordinate system, the distance to the target point is the distance between the target point and the processing area. The closer the specific processing area is to the distance, the greater the weighting given to the alignment data (coordinate position).
The point of interest is the center of deformation of the substrate (the center of point symmetry). For example, it is thermally deformed with the substrate center as the reference, or the center part floats up (expands) from the holder.
When it is adsorbed, the center point of the substrate is set as the point of interest. Further, the weight given to each alignment data of the specific processing region is changed according to the deformation state of the substrate, that is, the non-linear distortion amount.

By the way, in the second method, the alignment data of the specific processing area is weighted and the statistical calculation is performed for each processing area. In each of the plurality of processing regions located on the same circle centered at, the weight given to each coordinate position of the specific processing region is naturally the same. Therefore, when a plurality of processing areas are located on the same circle centered on the point of interest,
If the parameters (a to f) are calculated by performing the weighting and the statistical calculation as described above in only one processing area, the previously calculated parameters (a to f) are calculated in the remaining processing areas.
f) can be used as it is to determine its coordinate position. Therefore, when there are a plurality of processing areas on the same circle, the coordinate positions of all the processing areas on the same circle may be calculated using the same parameters (a to f). In this case, there is an advantage that the calculation amount for determining the coordinate position is reduced.

The third alignment method of the present invention is
For example, it is effective when the substrate is partially lifted (bulged) from the holder and held. In the third alignment method, each of the plurality of processing regions regularly arranged on the substrate according to the designed arrangement coordinates is set with respect to a predetermined reference position in the stationary coordinate system that defines the moving position of the substrate. Prior to the position adjustment, the coordinate positions on the stationary coordinate system of at least three processing areas selected in advance as the specific processing areas among the plurality of processing areas are measured.
Further, each of the coordinate positions on the stationary coordinate system of the at least three specific processing regions is corrected based on the flatness of the substrate, and the plurality of corrected coordinate positions are statistically calculated to obtain a plurality of objects on the substrate. The coordinate position of each processing substrate on the stationary coordinate system is determined. Then, by controlling the moving position of the substrate according to the calculated coordinate position and the flatness of the substrate, each of the plurality of processing regions is sequentially aligned with the reference position.

Further, the fourth alignment method of the present invention is also effective when the substrate is partially lifted (expanded) from the holder and held. In the fourth alignment method, each of the plurality of processing regions regularly arranged on the substrate according to the designed arrangement coordinates is set with respect to a predetermined reference position in a stationary coordinate system that defines the moving position of the substrate. Of the multiple processing areas, prior to alignment
At least three processing areas are selected as specific processing areas. Furthermore, at least 3 when the surface of the selected specific processing area and the moving plane of the substrate (the plane including the stationary coordinate system, that is, the orthogonal coordinate system defined by the two sets of interferometers) are substantially parallel to each other. The coordinate position of each of the two specific processing areas on the stationary coordinate system is obtained. For example, the surface position detection system is used to detect the amount of inclination of the surface of the area with respect to the moving plane of the substrate when the coordinate position of the specific processing area is measured, and the detected value is used to detect the specific processing area. It suffices to correct the coordinate position to obtain it. Alternatively, after the moving plane of the substrate and the surface of the specific processing area are made substantially parallel by using the surface position detection system, the coordinate position of the area is measured, or the processing area is substantially parallel to the moving plane of the substrate. May be selected as the specific processing area, and the coordinate position of the selected processing area may be measured. further,
By statistically calculating the plurality of detected coordinate positions, the coordinate position of each of the plurality of processing regions on the substrate on the stationary coordinate system is calculated. Then, by controlling the moving position of the substrate according to the calculated coordinate position and the inclination amount of each processing region with respect to the moving plane of the substrate, each of the plurality of processing regions on the substrate is sequentially positioned with respect to the reference position. I decided to match.

[0023]

Now, according to the present invention, there is provided a positioning method capable of accurately positioning all the processing regions on the substrate with respect to a predetermined reference position even if the substrate has "non-linear distortion". It is intended to be provided. Then, what kind of "non-linear distortion" is a target of the present invention for improving the alignment accuracy will be briefly described with reference to FIG. FIG. 10 is a graph showing the position measurement results (circle in the figure) of a plurality of (four here) specific processing regions (sample shots) on the substrate, and the vertical axis represents the amount of positional deviation. The horizontal axis represents the position from the substrate center. In addition, in order to simplify the explanation,
It is assumed that the substrate has only scaling.

In FIG. 10A, when a first-order approximation formula is created from the alignment data (coordinate position) of the sample shot by using the least squares method, a straight line shown by a solid line in the figure is obtained. In the case of FIG. 10A, the alignment data are sufficiently approximated by a linear function (straight line), and it can be said that a linear scaling error (distortion) occurs on the substrate. The conventional EGA method employs such an approximation method. On the other hand, in FIG. 10B, since the alignment data (marked with circles) are on a smooth curve indicated by the dotted line, it can be said that “regular non-linear distortion” occurs in the substrate. In addition, in FIG. 10C, since the alignment data (marked with ◯) has no regularity, it can be said that “irregular non-linear distortion” occurs in the substrate.

Now, if the conventional EGA method is applied to FIGS. 10B and 10C as it is, and a primary approximation formula is obtained in the same manner as in FIG. It becomes a straight line as shown. As is clear from the figure, in any case, there is a shot area with poor alignment accuracy, in other words, there is a shot area that cannot be approximated by a linear function. That is, it is impossible in principle to correct the nonlinear distortion in the conventional EGA method. Therefore, in the present invention, among the non-linear distortions, in particular, “regular non-linear distortion” as shown in FIG. It is possible to accurately align all the processing areas of the above with respect to the reference position.

The first alignment method of the present invention is effective for "regular non-linear distortion". "Even if a substrate has a regular non-linear distortion, it can be generated in a local region on the substrate. The array errors of are almost equal. " Therefore, in the first alignment method, when determining the coordinate position of one processing area (shot area) on the substrate on the stationary coordinate system, the processing area and at least three specific processing areas (sample shots) are determined. The alignment data (coordinate position) of each specific processing area is weighted according to the distance from each. That is, the weight given to the alignment data is set to be larger for the specific processing area having a shorter distance to the processing area. Therefore, in the first alignment method, statistical processing (least square method, simple averaging processing, etc.) is performed after weighting each alignment data of the specific processing area according to the above distance for each processing area on the substrate. ),
The coordinate position of each processing area on the stationary coordinate system will be determined.

Therefore, even if the substrate has a local array error (regular non-linear distortion), the coordinate positions (shot arrays) of all the processing regions on the substrate can be accurately determined. Moreover, since the alignment data used for each processing area (coordinate positions of at least three specific processing areas) is the same for all processing areas, it is not necessary to select the alignment data used for each processing area, which reduces the calculation amount. It is possible to reduce. Further, by arbitrarily selecting the weighting function (or optimizing it according to the substrate), the degree of weighting the alignment data can be easily changed (or the weighting can be optimized). That is, it is possible to determine the coordinate positions of all the processing areas under the optimum processing conditions (calculation parameters) for each substrate.

The second alignment method of the present invention is effective for "regular, especially point-symmetrical non-linear distortion". , And the magnitudes of the array errors are almost equal at the positions where the distance from the center of point symmetry is equal on the substrate. ” Therefore, in the second alignment method, when the coordinate position of one processing area (shot area) on the substrate on the stationary coordinate system is determined, the processing area and the point of interest (point (Center of symmetry), the point of interest, and at least three specific processing areas (sample shots)
The alignment data (coordinate position) of each specific processing region is weighted according to the distance from each of the above. That is,
The closer the distance to the target point is to the distance between the target point and the processing area, the larger the weight given to the alignment data. Therefore, in the second alignment method, for each processing area on the substrate, each alignment data of the specific processing area is weighted in accordance with the above two distances, and then statistical calculation is performed, so that each processing area remains stationary. The coordinate position on the coordinate system will be determined. In particular, when the substrate is thermally deformed with respect to the substrate center, or when the central portion is lifted up (swelled) from the holder and is adsorbed, the center point of the substrate, which is the center of point symmetry, is the point of interest.

Therefore, the second position is set with respect to the substrate in which the local array error (regular non-linear distortion) is point-symmetrical, for example, the substrate in which the central portion is floated and held by the holder. By applying the alignment method, the coordinate positions of all the processing regions on the substrate can be accurately determined without increasing the number of sample shots, and the same effect as that of the first alignment method can be obtained. You can
Further, by arbitrarily selecting a weighting function and appropriately changing the degree of weighting to the alignment data, the coordinate positions of all the processing regions are determined under the optimum processing condition (calculation parameter) for each substrate. It becomes possible.

Furthermore, the third alignment method of the present invention is effective, for example, when the substrate is partially lifted (expanded) from the holder and held. Third
In the alignment method of (1), when any part of the substrate is lifted from the holder and is adsorbed, each of the coordinate positions on the stationary coordinate system of at least three specific processing areas is corrected (coordinate conversion) based on the flatness of the substrate. Then, the coordinate positions on the stationary coordinate system of each of the plurality of substrates to be processed on the substrate are calculated by statistically calculating the corrected plurality of coordinate positions. In aligning each of the plurality of processing regions with respect to the reference position, the previously calculated coordinate position and the flatness of the substrate are used, that is, the calculated coordinate position is corrected again based on the flatness of the substrate ( (Coordinate conversion), and the moving position of the substrate is controlled according to the corrected coordinate position. Therefore, even if there is a bulge on any part of the substrate, the coordinate positions of all the processing regions on the substrate can be accurately determined, and the alignment accuracy can be improved without increasing the number of sample shots. Becomes

The fourth position alignment method of the present invention is also effective when the substrate is partially lifted (swelled) from the holder and held, like the third method. In the fourth alignment method, for example, a surface position detection system is used to detect the coordinate position of each of at least three specific processing regions when the surface of the specific processing region and the moving plane of the substrate are substantially parallel to each other. After making the moving plane of the substrate substantially parallel to the surface of the specific processing area, the coordinate position of the area is detected. Furthermore, the coordinate positions on the stationary coordinate system of each of the plurality of processing regions on the substrate are calculated by statistically calculating the plurality of detected coordinate positions. When aligning each of the plurality of processing regions with respect to the reference position, the previously calculated coordinate position and the inclination amount for each processing region with respect to the moving plane of the substrate are used, that is, the inclination amount detected for each processing region. The coordinate position is corrected based on the above, and the moving position of the substrate is controlled in accordance with the corrected coordinate position. Therefore, even if there is a bulge on any part of the substrate, the coordinate positions of all the processing regions on the substrate can be accurately determined, and the alignment accuracy can be improved without increasing the number of sample shots. Becomes

[0032]

2 is a diagram showing a schematic structure of a projection exposure apparatus suitable for applying the alignment method according to the present invention, and FIG. 3 is a block diagram of a control system of the projection exposure apparatus shown in FIG. is there. In FIG. 2, the illumination light IL (i-line, KrF excimer laser, etc.) from the exposure illumination system (not shown) passes through the condenser lens CL and the pattern area P of the reticle R.
Illuminate A with uniform illuminance. The illumination light IL that has passed through the pattern area PA is projected on both sides of the projection optical system P.
Upon entering L, the projection optical system PL image-projects the image of the circuit pattern formed in the pattern area PA onto the wafer W having a resist layer formed on the surface thereof. The wafer W is mounted on the Z stage LS via a wafer holder (not shown), and the Z stage LS is driven by the motor 13 to project the optical system P.
It is configured to be finely moved in the optical axis AX direction (Z direction) of L and can be tilted in an arbitrary direction. Z stage LS
X and Y by step-and-repeat method by motor 12
It is mounted on a wafer stage WS that can move two-dimensionally in any direction. The position of the wafer stage WS in the X and Y directions is constantly detected by the laser interferometer 15 with a resolution of, for example, about 0.01 μm. A movable mirror 14 that reflects the laser beam from the interferometer 15 is fixed to the end of the Z stage LS. The moving mirror 14 is preferably a corner cube, for example.

Further, in FIG. 2, a TTL type laser step alignment (LSA) system 17 for detecting an alignment mark on the wafer W is provided. LS
As shown in FIG. 11A, the A system 17 irradiates the alignment mark (diffraction grating mark) Mx attached to each shot area on the wafer with the elongated strip spot light LXS via the projection optical system PL, and Is photoelectrically detected by the diffracted light (or scattered light) generated from the mark Mx during relative scanning. Since the configuration of the LSA system 17 is disclosed in, for example, Japanese Patent Laid-Open No. 60-130742, detailed description is omitted here. Further, in FIG. 2, L for detecting the position of the alignment mark in the Y direction is used.
Only the SA system is shown, but in reality, another set of LSA system for detecting the position in the X direction is also arranged. LSA system 17
The photoelectric signal from the sensor is input to the alignment signal processing circuit 16 together with the position signal from the interferometer 15, where the position of the alignment mark is detected, and this position information is output to the main controller 10.

Further, in FIG. 2, for example, Japanese Patent Laid-Open No. 2-54
Off-axis type alignment sensor as disclosed in Japanese Patent Laid-Open No. 103 (hereinafter referred to as "Field Image Alignmen").
A t (FIA) system) 20 is also provided. The FIA system 20 is an illumination light having a predetermined wavelength width (for example, white light).
12A, the image of the alignment mark (WM ₁ ) on the wafer as shown in FIG. 12A and the index mark (FM ₁ , The image of FM ₂ ) is formed and detected on the light receiving surface of the image pickup device (CCD camera or the like). FIA
The image signal from the system 20 also receives the alignment signal processing circuit 16
, The position of the alignment mark is detected, and this position information is output to main controller 10.

The main controller 10 includes a processing circuit 1 as described later.
EGA calculation is performed based on the position information from No. 6, and coordinate positions of all shot areas on the wafer W (shot array)
In addition to calculating The stage controller 11 follows a drive command from the main controller 10,
Based on various information from the interferometer 15 and the surface position detection systems 18 and 19, the wafer stage W is driven via the motors 12 and 13.
It drives and controls the S and Z stages LS. Further, FIG. 2 also shows the surface position detection systems 18 and 19 of the oblique incident light system. The surface position detection systems 18 and 19 detect the height position (the position in the Z direction) of the wafer surface and the inclination angle thereof.
Since it is disclosed in Japanese Patent No. 13706, the description is omitted here.

Next, the specific configuration of the main control unit 10 will be described with reference to FIG. In FIG. 3, an alignment signal processing circuit 16 inputs a photoelectric signal from the LSA system 17 (or an image signal from the FIA system 20) and a position signal from the interferometer 15, and attaches it to each shot area by a predetermined signal processing. The position of the alignment mark (that is, the coordinate value in the orthogonal coordinate system XY defined by the interferometer 15) is detected. Alignment data storage unit 10
The mark position information 5 can input mark position information from the alignment signal processing circuit 16. The EGA calculation unit 100
A plurality (three or more, usually 1
Shot area (sample shot) of 0 to 15)
The coordinate positions of all shot areas on the wafer W are calculated by statistical calculation based on the respective mark position information of 1) and the weighting function determined by the weight generating unit 101. The storage unit 106 can input the shot array calculated by the EGA calculation unit 100, calculation parameters, and the like.

The exposure shot position data section 102 stores design array coordinate values of shot areas to be exposed on the wafer W, and these coordinate values are the EGA calculation section 100, the weight generation section 101, and the sample shot designation section. It is output to 103. The sample shot designating unit 103 includes a data unit 10
Based on the shot position information from 2, the sample shots used for the EGA calculation are determined, and the information (number, position) regarding the arrangement of the sample shots is used as the weight generation unit 10.
1 and the sequence controller 104. The weight generation unit 101 determines a weighting function based on the shot position information from the position data unit 102 and the information regarding the arrangement of the sample shots from the designation unit 103, and gives this function to the EGA calculation unit 100. The sequence controller 104 determines a series of procedures for controlling the movement of the wafer stage WS at the time of alignment or step-and-repeat exposure, based on the above-mentioned various data, and also controls the entire apparatus.

Next, a positioning method according to the first embodiment of the present invention will be described with reference to FIG. The alignment method of the present embodiment is based on the conventional EGA method, and when determining the coordinate position of the i-th shot area ESi on the wafer W, the area ESi and m samples (m = 9 in FIG. 1) are sampled. Distance L _K1 to each of shots SA _{1 to} SA ₉
A feature is that weighting W _in is given to each of alignment data (coordinate positions) of nine sample shots in accordance with L _K9 . Therefore, in this embodiment, two sets of LSAs are used.
After detecting the alignment marks (Mx ₁ , My ₂ ) of each sample shot using the system, the residual sum of squares Ei is evaluated by the following equation (Equation 3) as in the above Equation 2, and the following equation is the minimum. The calculation parameters a to f may be determined so that
In the present embodiment, the sample shots (alignment data) used for each shot area are the same, but the distance to each sample shot is naturally different for each shot area, so the alignment data (coordinate position of the sample shot). The weight W _in given to V will change for each shot area. Therefore, by determining the parameters a to f for each shot area and calculating the coordinate positions thereof, the coordinate positions (shot array) of all the shot areas are determined.

[0039]

[Equation 3]

Here, in the present embodiment, the weighting W _in for the alignment data of each sample shot is changed for each shot area on the wafer W. Therefore, the weight W _in is expressed as a function of the distance L _kn between the i-th shot area ESi and the n-th sample shot SA _n as in the following equation. However, S is a parameter for changing the degree of weighting.

[0041]

[Equation 4]

As is clear from the equation 4, the shorter the distance L _kn to the i-th shot area ESi, the larger the weight W _in given to the alignment data (coordinate position) of the sample shot. .. Here, when the value of the parameter S in Expression 4 is sufficiently large, the result of the statistical calculation process is almost equal to the result obtained by the conventional EGA method. On the other hand, the shot areas to be exposed on the wafer are all sample shots, and the parameter S
If the value of is sufficiently close to zero, the result obtained by the D / D method is almost the same. That is, in this embodiment, by setting the parameter S to an appropriate value, the EGA method and the D
The intermediate effect of the / D method can be obtained.

For example, for a wafer having a large non-linear component, by setting the value of the parameter S small, D /
It is possible to obtain an effect (alignment accuracy) almost equal to that of the D method. That is, in the alignment method (EGA calculation) according to the present embodiment, it is possible to satisfactorily remove the alignment error due to the non-linear component. Further, when the measurement reproducibility of the alignment sensor is poor, by setting the value of the parameter S large, it is possible to obtain an effect substantially equivalent to that of the EGA method, and it is possible to reduce the alignment error by the averaging effect. Becomes

Further, the weighting function (Formula 4) as described above is prepared for each of the X-direction alignment mark (Mx ₁ etc.) and the Y-direction alignment mark (My ₁ etc.), and the X-direction and Y-direction alignment mark are prepared. It is possible to set the weighting W _in independently for each direction. Therefore, the degree of nonlinear distortion of the wafer (size), regularity, or step pitch, that is, the distance between the centers of two adjacent shot areas (although it depends on the width of the street line on the wafer, it corresponds to almost the shot size) Even if the values are different in the X direction and the Y direction, the shot arrangement error on the wafer can be accurately corrected by setting the value of the parameter S independently. Here, the value of the parameter S may be different in the X direction and the Y direction as described above, and the parameter S may be the same or different in the X and Y directions. The value of S may be appropriately changed according to the magnitude and regularity of the “regular non-linear distortion”, the step pitch, the measurement reproducibility of the alignment sensor, and the like.

From the above, the effect can be changed from the EGA system to the D / D system by appropriately changing the value of the parameter S. Therefore, alignment is performed on various layers, and further on each component (X direction and Y direction) according to, for example, the characteristics of the nonlinear component (for example, size, regularity, etc.), the step pitch, and the quality of the measurement reproducibility of the alignment sensor. It is possible to flexibly change and perform alignment under optimal conditions for each layer and each component.

By the way, although the arrangement (number, position) of the sample shots is not described in the present embodiment, the preferred arrangement in the present embodiment draws a polygonal sample shot around the wafer like the conventional EGA method. Instead of arranging in this manner, for example, sample shots may be uniformly set on the entire surface of the wafer as shown in FIG. In order to increase the density of sample shots especially in the peripheral area of the wafer,
As shown in FIG. 9B, a large number of sample shots may be set in the donut-shaped (ring-shaped) area. Of course, in FIG. 9B, a plurality of sample shots are also set inside the donut-shaped area. In the present embodiment, a shot area in a partial area in which the amount of positional deviation (that is, the amount of nonlinear distortion) on the wafer is large may be selected as a sample shot, and the number of sample shots set in the partial area may be changed. It is better to set more than in the area.

Next, a positioning method according to the second embodiment of the present invention will be described with reference to FIG. In this embodiment, the wafer W undergoes regular, especially point-symmetric, non-linear distortion. Specifically, as shown in FIG. 5, the wafer W swells with respect to its center Wc and is held by a holder. A suitable alignment method will be described. Note that FIG. 6 is a vector map showing the amount of deviation of each shot area on the wafer having the nonlinear distortion shown in FIG. 5 from the ideal lattice.

In the present embodiment as well, similar to the first embodiment, the conventional EGA method is basically used, and the deformation center point of the wafer (the center of point symmetry of the non-linear distortion) which is the point of interest on the wafer, that is, the wafer center Wc, The distance (radius) L _Ei between the i-th shot area ESi on the wafer W and the wafer center Wc and m (in FIG. 4, m = 9) sample shots S.
It is characterized in that each of the alignment data of nine sample shots is given a weight W _in 'according to the distance (radius) L _{W1 to} L _{W9 from} each of A _{1 to} SA ₉ . Therefore, in the present embodiment, the LSA system is used and the two sets of alignment marks (Mx ₁ , Mx) for each sample shot.
After detecting y ₁ ), the residual sum of squares Ei ′ is evaluated by the following equation (Equation 5) and the calculation parameters a to f are determined so that the following equation becomes the minimum, as in the above Equation 3. good. In the present embodiment as well, as in the first embodiment, the weighting W _in 'given to the alignment data changes for each shot area.
The statistical calculation is performed for each shot area to determine the parameters a to f, and the coordinate positions thereof are determined.

[0049]

[Equation 5]

Here, for each shot area on the wafer W,
'For changing the weighting W _in the following equation' weighting W _in for each sample shot it is expressed as a function of the distance (radius) L _Ei of the i-th shot area ESi and wafer center Wc of the wafer W. However, S is a parameter for changing the degree of weighting.

[0051]

[Equation 6]

As is clear from Equation 6, the distance (radius) L _Wn to the wafer center Wc is the wafer center Wc.
A sample shot closer to the distance (radius) L _Ei between c and the i-th shot area ESi on the wafer W has a larger weight W _in ′ given to the alignment data. In other words, the largest weighting W _in 'for the alignment data of the sample shots located on the circle of radius L _Ei centered on the wafer center Wc.
And the weighting W _in 'for the alignment data is reduced as the distance from the circle in the radial direction increases.

Further, the value of the parameter S in the equation 6 is the alignment accuracy required as in the first embodiment,
It may be appropriately determined according to the characteristics of the nonlinear distortion (for example, size, regularity, etc.), the step pitch, the quality of the measurement reproducibility of the alignment sensor, and the like. That is, when the non-linear component is relatively large, by setting the value of the parameter S smaller, it is possible to reduce the influence of sample shots in which the distance L _Wn from the wafer center Wc differs greatly.
On the other hand, when the non-linear component is relatively small, by setting the value of the parameter S to a larger value, it is possible to prevent the alignment accuracy of the alignment sensor (or layer) having poor measurement reproducibility from being degraded.

Further, in the present embodiment, the weighting of the alignment data of the sample shots and the statistical calculation (that is, the calculation of the parameters a to f) are performed for each shot area.
To do. However, of course, in each of a plurality of shot areas located at substantially the same distance from the point of interest (center of point symmetry) on the wafer, that is, a plurality of shot areas located on the same circle centered on the point of interest, sample shots are taken. The weighting W _in 'given to the alignment data of 1 is the same. Therefore, when a plurality of shot areas are located on the same circle centered on the point of interest, the parameters a to f are set by performing the weighting and the statistical calculation as described above in only one of the shot areas. Once calculated, the coordinate positions of the remaining shot areas can be determined using the previously calculated parameters a to f as they are. Therefore, when a plurality of shot areas exist on the same circle, the same parameters a to
Alternatively, f may be used to determine the coordinate positions of all shot areas on the same circle. In this case, there is an advantage that the calculation amount for determining the coordinate position is reduced.

By the way, in the arrangement of sample shots suitable for the alignment method according to the present embodiment, it is desirable to specify the shot area so as to be symmetrical with respect to the point symmetry center of the nonlinear distortion, that is, the wafer center Wc. It may be selected as an X-shaped reference or a cross-shaped reference. Alternatively, the same arrangement as that of the first embodiment (FIG. 9) may be used. If the point symmetry center of the non-linear distortion is other than the wafer center, the X-shaped or cruciform arrangement based on the point symmetry center may be used as a matter of course. Further, also in the present embodiment, when determining the parameters a to f, the weighting function shown in Expression 5 may be independently set in each of the X and Y directions, as in the first embodiment.
In this case, even if the magnitude and regularity of the nonlinear distortion, the step pitch, etc. are different in the X and Y directions, the parameter S
By independently setting the values of, there is an advantage that the shot array on the wafer can be accurately calculated.

Further, when S = 120 in the above formula 6 and the alignment method according to the present embodiment is applied to the wafer shown in FIG. 5, the alignment accuracy is X + 3σ = 0.
It became 09 μm. On the other hand, when the sample shots are arranged in the same manner as above and alignment is performed by the conventional EGA method, the alignment error is X + 3σ = 0.21.
It was μm, and it was confirmed that the alignment accuracy was improved as compared with the conventional method.

In the first and second embodiments, equations 4 and 6 are used.
The weights W _in and W _in 'shown in (1) are determined by the weight generation unit 101 based on the arrangement of sample shots. Further, in the alignment methods of the first and second embodiments, the degree of weighting of the sample shot alignment data can be changed by the parameter S as described above. Hereinafter, a method of determining the parameter S in the weight generation unit 101 will be described. The following formula 7 is stored in the weight generation unit 101. For example, when the operator sets the weight parameter D to a predetermined value, the parameter S,
That is, the weighting W _in or W _in 'is determined.

[0058]

[Equation 7]

Here, the physical meaning of the weight parameter D is a range of sample shots (hereinafter simply referred to as a zone) effective for calculating the coordinate position of each shot area on the wafer. Therefore, when the zone is large, the number of effective sample shots is large, and the result is close to the result obtained by the conventional EGA method. Conversely, if the zone is small, the number of valid sample shots will be small, so D
This is close to the result obtained with the / D method. However, the range (zone) referred to here is only a reference value for weighting, and even if all the sample shots exist outside the zone, the coordinate position is set in the same manner as in the above embodiment. The statistical calculation is performed by maximizing the weight for the alignment data of the sample shot closest to the shot area to be determined.

In FIG. 7, the weighting parameters D are 30, 60,
It is a visual representation of the size of the zone at 90 and 120 [mm]. However, the weight parameter, that is, the diameter D of the zone, is as shown in FIG. The diameter of the area (sampling zone) (unit is mm) ”. It has been confirmed that the optimum value of the diameter D is generally in the range of 30 to 150 [mm].

Therefore, in the above embodiment, the operator inputs the optimum zone diameter D determined based on the experience of the operator, or by experiment or simulation, to the main controller 10 via the input device (keyboard or the like). Then, the degree of weighting of the alignment data from Equation 7, that is, the weights W _in and W of Equations 4 and 6
_in 'will be decided. Therefore, it is possible to flexibly change the alignment for various layers according to the magnitude of the non-linear component, the quality of the measurement reproducibility of the alignment sensor, etc., and it is possible to perform the alignment for each layer under optimum conditions. .

By the way, in addition to the operator directly inputting the optimum zone diameter D to the main controller 10, for example, the first (first) wafer among a plurality of wafers stored in a lot is selected. On the other hand, mark detection in almost all shot areas is performed. Then, the main controller 10 calculates the regularity, degree (magnitude), etc. of the nonlinear distortion of the wafer based on the detection result, and then the optimum zone diameter D (in the second embodiment, the distortion center of the nonlinear distortion is further increased). May also be set). As a result, the weighting functions W _in and W _in 'of the equations 4 and 6 are automatically determined without any operator intervention, and the second and subsequent wafers are positioned under the weighting functions as described above. A matching operation will be performed.

The position of the first wafer may be adjusted by using the above mark detection result, or the shot arrangement is calculated by using the weighting function determined as described above to perform the position adjustment. You may do so. In addition, here, the mark detection (coordinate position measurement) of almost all shot areas is performed only for the first wafer, but the mark detection of almost all shot areas is performed for the first to several wafers. Alternatively, the weighting W _in and W _in ′ may be determined by calculating the regularity and magnitude of the non-linear distortion using averaging processing or the like. Further, the regularity, magnitude, etc. of the non-linear distortion calculated by the main control device 10 are displayed on a display device (CRT, etc.) as a vector map as shown in FIG. 6, and the operator is based on the map displayed here. Alternatively, the optimum zone diameter D may be determined and input to the main controller 10.

Here, in the above description, the operator inputs the optimum value of the diameter D of the zone to the main control unit 10. However, for example, a wafer or a lot (loader cassette) containing a plurality of wafers is stored. , The above value is written in the form of an identification code (bar code or the like), and the code is read by a reading device (bar code reader or the like).
May be determined. In addition, the weight generation unit 101
The determination formula of the parameter S stored in is not limited to the formula 7, and the following formula 8 may be used.
Here, A is the area of the wafer (unit is mm ² ), m is the number of sample shots, and C is the correction coefficient (positive real number).

[0065]

[Equation 8]

In the equation (8), changes in the wafer size (area) and the number of sample shots are reflected in the determination of the parameter S so that the optimum value of the correction coefficient C to be used in the determination does not change much. Is. Here, when the correction coefficient C is small, the value of the parameter S is large, so that the result is close to the result obtained by the conventional EGA method, exactly as in the case of the expression 7. On the contrary, when the correction coefficient C is large, the value of the parameter S is small, so that the result is close to the result obtained by the D / D method as in the case of the formula 7. Therefore, only by inputting the correction coefficient C previously determined by an experiment, a simulation, or the like to the main controller 10 through the operator or the reading device of the identification code, the degree of weighting of the alignment data from Equation 8 can be calculated.
That is, the weights W _in and W _in 'of the equations 4 and 6 are automatically determined. For this reason, alignment is performed for various layers, and further for each component (X direction and Y direction) according to, for example, characteristics of non-linear component (for example, size, regularity, etc.), step pitch, and measurement reproducibility of the alignment sensor. Can be changed flexibly, and alignment can be performed under optimum conditions for each layer and each component. In particular, when Equation 8 is used, even if the wafer size, the step pitch (shot size), the number of sample shots, etc. change, the coordinate positions of all shot areas on the wafer can be accurately determined regardless of these changes, and it is always stable. There is also an advantage that it is possible to perform the alignment with the required accuracy.

In each of the above embodiments, m sample shots are selected from a plurality of shot areas on the wafer, the alignment data of the selected sample shots are weighted, and then statistical calculation is performed. To do. At this time, if there are sample shots in which the two sets of alignment marks cannot be measured or the measured values are suspicious (unreliable), the shot area in the vicinity of the shot is designated as a substitute shot and this designation is performed. Alternatively, the alignment data of the alternative shot may be used. Alternatively, sample shots that cannot be measured or have low reliability may be rejected, or the alignment data may be weighted to zero, and only the alignment data of the remaining sample shots may be used. Furthermore, in the case of a sample shot in which only one of the two sets of alignment marks (for example, the X mark) cannot be measured or the reliability of the measured value is low,
Only the coordinate position of the other alignment mark (Y mark) is used. Alternatively, the X mark in the shot area near the shot may be detected and the coordinate position thereof may be used.

By the way, although the alignment accuracy may not be improved even if the alignment method of each of the above embodiments is applied, this is due to the fact that in the above embodiments, the non-linear distortion is particularly regular. This is because it is a correction target. Therefore, if the alignment accuracy does not improve as described above, it is considered that the wafer has many irregular nonlinear components. Generally, it is difficult to improve the alignment accuracy for a wafer having irregular nonlinear distortion, whichever alignment method is applied, but here it has irregular nonlinear distortion, that is, alignment accuracy does not improve. For the wafer, consider separately the case where the measurement reproducibility of the alignment sensor is good and the case where it is bad.

If the measurement reproducibility of the alignment sensor is poor, a result as if the wafer had an irregular nonlinear distortion can be obtained even though the irregular nonlinear distortion does not occur in the wafer itself. Sometimes.
In such a case, an alignment sensor and / or a signal processing condition that can obtain good measurement reproducibility for the wafer may be selected and used. Specifically, when the alignment mark on the wafer has a low step, for example, JP-A-2-272305 and JP-A-3272305.
As disclosed in Japanese Patent Publication No. 272406, a one-dimensional interference fringe is formed by irradiating a one-dimensional diffraction grating mark on a wafer with coherent parallel beams from two directions.
An alignment sensor that photoelectrically detects diffracted light interference light generated in the same direction from the mark (hereinafter referred to as Laser
r Interferometric Alignment (LIA) system) may be used. If a metal layer is formed on the wafer surface, the FIA system 20 shown in FIG. 2 may be used.

Also, the alignment sensor is not changed,
You may make it correspond by changing only the signal processing conditions of the said sensor. The signal processing conditions in the LSA system refer to a waveform analysis algorithm, an algorithm slice level, a processing gate width, and the like. The processing gate width is defined around the designed mark position. Also,
As the waveform analysis algorithm, for example, the following 3
There are two algorithms.

In the first algorithm, after smoothing the signal waveform in the section determined by the predetermined processing gate width, the signal waveform is sliced at the level set by the algorithm slice level, and shown in FIG. As shown, if there are intersections on the left and right of the signal waveform, the center point of the two intersections is detected as the mark position. The second algorithm smoothes the signal waveform in a section equal to or higher than a predetermined level L ₁ (voltage value), and then sets a plurality of slice levels with a level L ₂ close to the peak value at regular intervals, Find the intersection and its length at each slice level. Then, based on the length at each slice level, select a slice level that maximizes the slope of the signal waveform in the portion below the level set by the algorithm slice level, and set the center point of the intersection at the level as the mark position. It is something to detect. The third algorithm slices the signal waveform at the level set by the algorithm slice level and obtains the center point as the reference position. Next, after smoothing the signal waveform in a section equal to or higher than a predetermined level L ₁ (voltage value), a plurality of slice levels are set at a constant interval with the level L ₂ close to the peak value, and each slice level is set. Then, the center point of the two intersection points in (3) and the midpoint difference (that is, the difference between the center points of adjacent slice levels) are obtained. Then, the center point at each slice level is not far apart from the previously determined reference position, and each center point is stable (that is, the midpoint difference is small,
The area where the slice level is longest and continuous is selected), and the center point in the area is detected as the mark position.

Further, referring to FIGS. 12 and 13, the FIA
Signal processing conditions in each of the LIA system and the LIA system will be briefly described. FIG. 12A shows the state of the wafer mark WM ₁ detected by the FIA system, and FIG. 12B shows the waveform of the image signal obtained at that time. As shown in FIG. 12A, the FIA system 20 (imaging device) is the wafer mark WM.
Electrically scanned along the bar mark and index mark FM ₁ of three _1, the image of the FM ₂ to the scanning line VL. At this time, 1
Since only one scanning line is disadvantageous in terms of the S / N ratio, the image signal levels obtained by a plurality of horizontal scanning lines that enter the video sampling area VSA (dashed-dotted line) are averaged for each pixel in the horizontal direction. good. As shown in FIG. 12B, the image signal includes index marks FM ₁ and FM _{2 on} both sides.
There is a waveform portion corresponding to each of the marks, and the alignment signal processing section 16 obtains the center position (position on the pixel) of each mark by processing this waveform portion with the slice level SL ₂ and obtains the center position x ₀ . ing. Instead of obtaining the respective center positions of the index mark FM _1, FM _2, by obtaining the respective positions of the right edge and the left edge of the index mark FM ₂ of the index mark FM _1, so as to obtain the center position x ₀ I don't care. On the other hand, here, as shown in FIG. 12B, the waveform on the image signal is the bottom at the positions corresponding to the left edge and the right edge of each bar mark, and the signal processing unit 16 uses the slice level SL ₁ to After performing the waveform processing to obtain the center position of each bar mark, the respective positions are arithmetically averaged to calculate the center position x _C of the wafer mark WM ₁ . Further, the difference Δ between the previously determined position x ₀ and the mark measurement position x _C
x (= x ₀ −x _C ) is calculated, and a value obtained by adding the position of the wafer stage WS when the wafer mark WM ₁ is positioned in the observation area of the FIA system 20 and the difference Δx between the positions is used as mark position information. Output as.

Therefore, the signal processing conditions that can be changed in the FIA system 20 as described above include a waveform analysis algorithm, a slice level SL ₁ (voltage value), a contrast limit value, and a processing gate width Gx (width Gx on a pixel). Center position and its width) etc. Further, as a waveform analysis algorithm, for example, as disclosed in Japanese Patent Laid-Open No. 4-65603, when the center position of each bar mark is obtained, the waveform portions BS _1L and BS corresponding to the left edge and the right edge of the bar mark are obtained. _{Of 1R} , BS _2L and BS _2R ,
There are a mode using only the outer slopes BS _1L and BS _2R , a mode using only the inner slopes BS _1R and BS _2L , a mode using the outer slopes BS _1L and BS _2R , and a mode using the inner slopes BS _1R and BS _2L .

Next, the signal processing conditions in the LIA system (particularly the heterodyne system) will be described with reference to FIG.
As shown in FIG. 13, when two coherent beams (parallel light beams) BM ₁ and BM ₂ having a frequency difference Δf are incident on the one-dimensional diffraction grating mark WM ₂ on the wafer at a crossing angle (2ψ ₀ ). , The pitch P on the mark WM ₂ (however,
A one-dimensional interference fringe IF having a grating pitch 2P) is formed. This interference fringe IF moves in the pitch direction of the diffraction grating mark WM ₂ corresponding to the frequency difference Δf, and its velocity V is V = Δf.
-It is expressed by the relational expression P. As a result, from the diffraction grating mark WM ₂ as shown in FIG. 13 the diffracted light B ₁ (-1), B ₂
(+1), ... occurs. The subscripts 1 and ₂ represent the correspondence with the incident beams BM ₁ and BM _2, and the numbers in parentheses represent the diffraction orders. Usually, in the LIA system, a reference signal created separately from the photoelectric signals of the ± 1st-order diffracted lights B ₁ (-1) and B ₂ (+1) interfering light traveling along the optical axis AX and _two light-transmitting beams. The positional deviation is detected by obtaining the phase difference between the interference light for use and the photoelectric signal. Alternatively, the 0th-order diffracted light B ₂ (0) and −
The positional deviation amount detected from the phase difference between the photoelectric signal of the interference light with the second-order diffracted light B ₁ (-2) and the reference photoelectric signal, and the 0th-order diffracted light B ₁ (0) and the −2nd-order diffracted light B ₂ ( The position shift amount may be obtained by averaging the position shift amounts detected from the phase difference between the photoelectric signal of the interference light with +2) and the reference photoelectric signal.

Therefore, the only signal processing condition that can be changed in the LIA system as described above is the selection of the interference light (the order of the diffracted light) to be photoelectrically detected. That is, in the LIA system, the first mode using ± 1st-order diffracted lights B ₁ (-1) and B ₂ (+1), the 0th-order diffracted light B ₂ (0) and the −2nd-order diffracted light B ₁ (-2), and 0 A second mode using the second-order diffracted light B ₁ (0) and the −second-order diffracted light B ₂ (+2),
Further, there is a third mode in which the intensities of the interference light in the first mode and the second mode are compared, and the one with the larger intensity value is selected and used.
The simulation will be performed by changing the two modes.

As described above, when the measurement reproducibility of the alignment sensor is poor, the alignment sensor and the signal processing conditions are changed so that the mark detection is performed under the optimum condition for the layer. Just do it. Also,
When the alignment sensor and the signal processing condition are not changed, one alignment mark may be measured a plurality of times.

On the other hand, when the measurement reproducibility of the alignment sensor is good, it can be considered that the measurement value is reliable or unreliable. If the sample shot has high reliability with respect to the alignment data and the alignment data has irregular nonlinear distortion, it is considered that the wafer actually has irregular nonlinear distortion. In such a case, the number of sample shots may be increased, the diameter D of the zone in Expression 7 may be decreased, or the correction coefficient C in Expression 8 may be increased. Further, if the alignment accuracy does not improve even if the parameter S or the arrangement (number, position) of the sample shots is optimized, the alignment sensor cannot accurately measure the mark position due to the influence of the coverage, that is, the reliability of the alignment data is low. It is considered low. In such a case, FIA that is not easily affected by coverage etc.
The system 20 may be used for alignment.

Although the case where the wafer is considered to have irregular nonlinear distortion has been described above, the parameter S (that is, the diameter D of the zone and the correction coefficient C) is also applied to the wafer having regular nonlinear distortion. By changing at least one of the number of sample shots, the arrangement (position), the alignment sensor, and the signal processing condition of the alignment sensor, it is possible to perform alignment under optimum conditions for each layer. good. In other words,
It is preferable to optimize the parameter S, the number of sample shots, and the arrangement condition after selecting the optimum alignment sensor and signal processing condition for the layer.

In the EGA method, regarding the regularity of the arrangement of shot areas on one wafer, the wafer scaling amounts Rx, Ry in the X and Y directions, and the wafer offset amounts Ox, Oy in the X and Y directions are set. The residual rotation error θ of the array coordinate system of the shot area and the inclination amount (orthogonality) ω of the array coordinate system are introduced as variable elements. That is, these six elements are represented by the following equations by the operation parameters af.

Rx = a Ry = d Ox = e Oy = f θ = c / d ω = − (b / a + c / d) Then, applying the first and second embodiments, one wafer When determining the coordinate position of each shot area, two types of alignment sensors, for example, LSA system and FIA system are used. That is, the coordinate positions of all sample shots are measured using each of the LSA system and the FIA system,
Further, after calculating the calculation parameters a to f using the least squares method, the coordinate position of one shot area is determined using the two sets of calculation parameters a to f. Specifically, L
The above six variable elements are determined from the calculation parameters a to f calculated from the SA system measurement result, and the above six variable elements are determined from the calculation parameter a to f calculated from the FIA system measurement result, particularly the scaling parameters Rx and Ry. To decide. Then, after replacing the scaling parameter of the LSA system with the scaling parameter of the FIA system, the scaling parameter Rx, Ry and the remaining four variable elements (Ox, Oy, θ, ω) of the LSA system are used as calculation parameters. a to f are determined, and the coordinate position of one shot area is calculated under the parameter. As described above, by using the two types of alignment sensors and determining the calculation parameters a to f, it is possible to improve the calculation accuracy of the coordinate position of the shot area as compared with the previous embodiments. However, it is assumed that the LSA system and the FIA system have exactly the same arrangement (number and position) of sample shots. Further, the weighting degree (value of the parameter S) is also set to be the same. Although the LSA system and the FIA system are used here, the number of alignment sensors to be used,
The combination and the like may be arbitrary.

In the above description, L is set for each shot area.
The mark detection of the sample shot was performed by each of the SA system and the FIA system, and the calculation parameter was calculated by the least square method. Here, when the difference between the calculation parameters calculated by the LSA system and the FIA system when determining the coordinate position of one shot area on the wafer is large,
It is considered that the measurement error of either the system or the FIA system is large, and it may be difficult to accurately calculate the coordinate position of the shot. Therefore, the main controller 10 notifies the operator of the state by a warning, screen display, or the like, or automatically performs remeasurement when the difference exceeds a predetermined allowable value. When re-measurement is performed, the type and combination of alignment sensors may be changed, and a new alignment sensor may be added. Also,
If the difference is less than or equal to the allowable value, one of the parameters a to f determined after replacing some variable elements as described above, or one of the two sets of calculation parameters, or the average value thereof is used. It suffices to determine the coordinate position of the shot area.

Further, after measuring the coordinate positions of all the sample shots by using one type of alignment sensor, for example, the FIA system, the degree of weighting (value of parameter S)
Each of the two sets of different weighting functions (Equation 4 or 6) is used to determine the operation parameter. At this time, the values of the two parameters S are set to, for example, a value at which a result substantially equal to that of the EGA method is obtained and a value at which a result substantially equal to that of the D / D method is obtained. When determining the coordinate position of the shot area using the two sets of calculation parameters calculated under the above two sets of weighting functions, the calculation parameters determined under the EGA-like weighting function, that is, six Of the variable elements, the offset amounts Ox and Oy are used, and for the remaining four variable elements, variable elements determined from the operation parameters determined under the D / D weighting function are used. Further, the calculation parameter may be determined from these six variable elements and the coordinate position may be determined using the parameter. The two parameters S may be arbitrary values, and may be set appropriately according to the type of layer, the characteristics of nonlinear distortion, and the like.

Next, a positioning method according to the third embodiment of the present invention will be described. Here again, a preferable alignment method for the wafer W having the regular nonlinear distortion shown in FIG. 5 will be described. However, in the present embodiment, weighting is not performed for each sample shot as in the second embodiment. , The flatness of the wafer surface is used, for example, the bulge of the wafer is obtained by the surface position detection systems 18 and 19, and the alignment is performed using this.

Now, the pitches Px and P in the X and Y directions, respectively.
When a measurement point (for example, an alignment mark) is provided at y and the height h (i, j) at an arbitrary position (i, j) on the wafer is measured, the height h (i, j) at the position (i, j) is measured in the X and Y directions. Incline, IncY for each inclination (for example, the movement plane of the wafer, that is, the inclination with respect to the orthogonal coordinate system XY defined by the interferometer)
Is expressed by the following equation.

[0085]

[Equation 9]

FIG. 14 is an enlarged view of a part of the cross section of a wafer having a partial bulge (a part of the bulge), where the thickness of the wafer is t and the upper and lower surfaces of the wafer are shown. Assuming that no displacement occurs in the central portion, the displacement amounts (lateral displacement amounts) ΔSx and ΔSy in the X and Y directions at the position (i, j) are expressed by the following equations.

[0087]

[Equation 10]

Therefore, the wafer (that is, the shot area)
It is possible to obtain the non-linear distortion due to the bulge (warp) of the wafer based on the flatness (flatness) and the thickness t of the wafer. In addition, using the surface position detection systems 18 and 19, the inclinations IncX in the X and Y directions at the position (i, j),
IncY is directly measured and the result is used for ΔSx, ΔS
You may make it calculate | require y from Numerical formula 10.

Next, the positioning operation of this embodiment will be described. In the present embodiment, the wafer center is assumed to be the highest and the flatness of the wafer is measured in advance by using the surface position detection systems 18 and 19, for example, the difference in height between each shot area on the wafer and the wafer center is measured. Are stored in the storage unit 106. Also,
The storage unit 106 also stores information about the thickness t of the wafer. Here, the surface position detection systems 18 and 19 are, for example, the moving coordinate system (orthogonal coordinate system XY) of the wafer stage WS.
It is assumed that the calibration has been performed in advance so that the plane including is the zero reference. The apparatus shown in FIGS. 2 and 3 can be used as it is for the alignment method according to the present embodiment. However, in the present embodiment, since it is not necessary to perform weighting in the EGA calculation, the weight generation unit 101 is unnecessary.

In this embodiment, in the apparatus shown in FIGS. 2 and 3, the coordinate positions of a plurality of preselected sample shots are measured by the LSA system 17, and the measured values are used to calculate the equation 9 The lateral deviation amounts ΔSx and ΔSy in each sample shot are calculated from 10, and the lateral deviation amount ΔS is calculated.
The coordinate position of each sample shot is corrected using x and ΔSy. That is, the coordinate position of each sample shot in a state where the wafer is apparently held substantially flat by the holder is obtained. After that, the EGA calculation (formula 1 and 2) as in the related art is performed using the corrected plurality of coordinate positions to calculate the coordinate positions (first shot array) of all the shot areas on the wafer. Next, using the calculated coordinate positions, the lateral shift amounts in the X and Y directions of each shot area are calculated (back-calculated) from Equations 9 and 10, and the lateral shift amount of each shot area previously calculated using this lateral shift amount is calculated. The coordinate position (first shot array) is corrected. As a result, the coordinate positions (second shot array) of all shot areas on the wafer shown in FIG. 5 are obtained. Therefore, by controlling the moving position of the wafer W (wafer stage WS) based on the second shot arrangement, each shot area is sequentially and accurately aligned (positioned) with respect to the reference position (exposure position). Is possible.

As described above, in the present embodiment, even if the wafer is partially swelled, it is possible to perform high precision and high speed operation without increasing the number of sample shots and without performing special EGA calculation such as weighting. The coordinate position of the shot area can be calculated, and the alignment accuracy can be improved on the entire surface of the wafer. In this embodiment, the case where the substantially central portion of the wafer is swollen (FIG. 5) has been described. However, even when an arbitrary portion of the wafer is swollen, the method according to this embodiment is applied as it is, and the same result is obtained. The effect can be obtained.

Next explained is a positioning method according to the fourth embodiment of the invention. This embodiment also describes the case of aligning the wafer shown in FIG. 5, but the difference from the third embodiment is that the flatness of the wafer is not measured in advance. In the present embodiment, in the apparatus shown in FIGS. 2 and 3, first, the LSA system 17 measures each coordinate position of a plurality of preselected sample shots. At this time, the surface position detection systems 18 and 19 are used to tilt the Z stage LS so that the surface of each sample shot is substantially parallel to the plane including the orthogonal coordinate system XY, and then the coordinate position of each sample shot is determined. It shall be measured. The coordinate positions obtained as a result are substantially equal to the coordinate positions of the sample shots in the state where the wafer is apparently held substantially flat by the holder. After that, the EGA calculation as usual (Formulas 1 and 2) is performed using the previously measured coordinate positions.
Then, coordinate positions of all shot areas on the wafer are calculated. Next, the movement position of the wafer stage WS is controlled according to the calculated coordinate position to position each shot area with respect to the exposure position. At this time, the shot area,
In particular, the shot area located at the center of the wafer is positioned away from the exposure position due to the bulge of the wafer. Therefore, after positioning an arbitrary one shot area to be aligned according to the previously calculated coordinate position, the surface position detection systems 18 and 19 are used to detect the surface inclination of the area (inclination with respect to the orthogonal coordinate system XY). Then, using the detected values, the lateral shift amounts in the X and Y directions of the shot area to be aligned are calculated (backwards calculated) from Equations 9 and 10. Then, the shot area to be aligned is accurately positioned with respect to the exposure position by positioning the wafer stage WS by offsetting it from the previously calculated coordinate position as an offset corresponding to this lateral displacement amount. Hereinafter, the lateral shift amount is detected for each shot area, and the coordinate position previously calculated by using the detected lateral shift amount as an offset is corrected.
By positioning S, it is possible to sequentially and accurately position (position) each shot area with respect to the exposure position.

As described above, also in the present embodiment, even if the wafer to be aligned is partially swollen, high precision is achieved without increasing the number of sample shots and performing special EGA calculation such as weighting. The coordinate positions of all shot areas can be calculated at high speed, and since it is not necessary to measure the flatness of the wafer in advance, it is possible to improve the alignment accuracy on the entire surface of the wafer without lowering the throughput. It becomes possible.
In this embodiment, the case where the substantially central portion of the wafer is swollen (FIG. 5) has been described. However, even when an arbitrary portion of the wafer is swollen, the method according to this embodiment is applied as it is, and the same result is obtained. The effect can be obtained.

Further, when the reticle pattern is superposed and exposed on each shot area, it is necessary to exactly match the best image plane of the projection optical system PL with the surface of the shot area. The inclination of the shot area (inclination with respect to the best image plane) is detected using the detection systems 18 and 19, and the Z stage LS is inclined based on the detected value. At this time, the shot area is also shifted relative to the exposure position in the orthogonal coordinate system XY as the Z stage LS is tilted, and the coordinate position of the shot area is calculated accurately, but the final position is determined. The alignment accuracy will decrease. Therefore, in the present embodiment, the surface position detection system 18, 1
9, the tilt of the shot area is detected, and thus the detected values are used to predict the shift amounts in the X and Y directions when the Z stage LS is tilted, and the predicted shift amount together with the lateral shift amount. It is desirable to determine the coordinate position of the final shot area in consideration of the above.

Since the surface position detection systems 18 and 19 are calibrated so that the plane including the Cartesian coordinate system XY serves as a zero point reference, the detected values can be used as they are for the calculation of the predicted shift amount. Can not. For this reason,
Projection optical system PL for a plane including an orthogonal coordinate system XY in advance
The inclination of the best image forming plane is obtained in advance, and this inclination amount is given as an offset to the detection values of the surface position detection systems 18 and 19,
It is desirable to predict the shift amount in the X and Y directions when the Z stage LS is tilted using this corrected value. Further, in the present embodiment, the coordinate position of the sample shot is measured in a state where the shot surface and the plane including the orthogonal coordinate system XY are substantially parallel to each other, but between the shot surface and the plane including the orthogonal coordinate system XY. The coordinate position is measured with the tilt still occurring, and at the time of this measurement, the surface position detection system 18, 1
The coordinate position may be corrected based on the tilt amount detected in 9, and this correction value may be used for EGA calculation.

Here, for example, when the wafer shape (the bulge thereof) can be clearly specified as shown in FIG. 5, from the inside (flat peripheral portion of the wafer in FIG. 5) where there is no influence of the bulge (the outer peripheral portion of the wafer). By selecting all the sample shots, it becomes unnecessary to perform coordinate conversion (correction of the coordinate position by the amount of lateral deviation) using the inclination amount of the shot surface caused by the bulge, and various errors that may occur due to the coordinate conversion are eliminated. It will be possible to eliminate them and shorten the measurement / calculation time.

As described above, when the alignment method according to the third and fourth embodiments is applied, since the height position and the inclination of the shot surface are measured, in order to realize highly accurate alignment, Ideally, there is no unevenness in the shot area. However, in reality, irregularities are generated in the shot area due to uneven coating of the resist and the like, but the reduction in alignment accuracy due to the irregularities is negligible. In order to reduce this effect, a collimator type sensor is used as a sensor for detecting the inclination of the shot surface, and a sensor for detecting the height position of the shot surface is used at multiple points in the shot area. It is desirable to use one that can detect the height position and use the average value of each height position in the above calculation. The collimator type is effective in that it can detect the average inclination of the entire surface of the shot. still,
A sensor capable of detecting height positions at a plurality of points in the shot area may be used as the surface position detection system. In this case, it is not necessary to separately provide a sensor for detecting the inclination of the shot surface.

Further, in the above embodiments, the non-linear distortion due to the bulge that occurs when the wafer is attracted to the holder has been a problem. Can be obtained. Furthermore, among the sample shots selected in advance, for sample shots where mark measurement is impossible or the measured value is suspicious (error is large), a new shot area near the shot is newly sampled. The shot position may be designated as a shot and the coordinate position of the designated sample shot may be used for EGA calculation.

In all of the above embodiments, the case where the LSA system is used as the alignment sensor has been described, but any type of alignment sensor may be used. That is, whether the method is the TTR method, the TTL method, or the off-axis method, the detection method thereof is the LSA method as described above, the image processing method as the FIA system 20, or the two-beam interference like the LIA system. Any method may be used. Further, the alignment method of the present invention may be realized by either software or hardware in the exposure apparatus. In addition to the exposure apparatus of various methods including the exposure apparatus of the step-and-repeat method, the step-and-scan method, or the proximity method (projection type exposure apparatus, X-ray exposure apparatus, etc.), the present invention also includes a repair apparatus, The same can be applied to a wafer prober or the like.

[0100]

As described above, according to the present invention, even if a local array error (non-linear distortion) occurs on a substrate in which a plurality of processing regions are regularly arrayed, the point of interest ( That is, the processing area to be aligned, or the center of regular nonlinear distortion (for example, partial bulge), such as the center point of the substrate), and at least three preselected specific processing areas. Weighting each coordinate position (alignment data) of a specific processing area according to the distance to each (sample shot), that is, changing the weighting of each alignment data for each processing area to be aligned. And For this reason, it becomes possible to perform highly accurate alignment on the entire surface of the substrate in response to local distortion (particularly, non-linear component) in the substrate. Moreover, since the alignment data used when determining the coordinate position of each processing area on the substrate is always the same, it is not necessary to select the alignment data to be used for each processing area, and the amount of calculation can be reduced. In addition, it is not necessary to increase the number of sample shots, and it is possible to prevent a decrease in throughput.
Especially in the semiconductor manufacturing process, which will become more and more diverse in the future,
As semiconductor substrates grow in size and undergo various heat treatments, the expansion and contraction of the substrate itself can also increase non-linear components.
Also for these, the present invention can perform the alignment sufficiently satisfying the required accuracy.

[Brief description of drawings]

FIG. 1 is a diagram for explaining an alignment method according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a schematic configuration of a projection exposure apparatus suitable for applying the alignment method according to the embodiment of the present invention.

3 is a block diagram of a control system of the projection exposure apparatus shown in FIG.

FIG. 4 is a diagram for explaining an alignment method according to a second embodiment of the present invention.

FIG. 5 is a diagram showing a state of a wafer to which an alignment method according to an embodiment of the present invention is applied.

6 is a vector showing the amount of deviation from the ideal lattice in each shot area on the wafer having the nonlinear distortion shown in FIG.
map.

FIG. 7 is a diagram for explaining a method of determining a parameter S in the alignment methods of the first and second embodiments.

FIG. 8 is a diagram for explaining a method of determining a parameter S in the alignment methods of the first and second embodiments.

FIG. 9 is a diagram showing an example of the arrangement of sample shots suitable for the alignment method of the first embodiment.

FIG. 10 is a diagram for explaining non-linear distortion to which the alignment methods of the first and second embodiments should be applied.

FIG. 11 is a diagram for explaining how mark position measurement is performed by the LSA system.

FIG. 12 is a diagram for explaining how mark position measurement is performed by the FIA system.

FIG. 13 is a diagram for explaining how mark position measurement is performed by the LIA system.

FIG. 14 is a diagram for explaining the principle of the alignment method according to the third embodiment of the present invention.

[Explanation of symbols]

 10 Main Controllers 18 and 19 Surface Position Detection System W Wafer WS Wafer Stage 100 EGA Calculation Unit 101 Weight Generation Unit

Claims (12)

[Claims] 1. A plurality of processing regions regularly arranged on a substrate in accordance with a designed arrangement coordinate are aligned with a predetermined reference position in a stationary coordinate system that defines a moving position of the substrate. In the method, the coordinate position on the stationary coordinate system of at least three processing areas selected in advance as specific processing areas among the plurality of processing areas is measured, and the processing area is defined as the processing area for each processing area on the substrate. Each of the coordinate positions of the at least three specific processing regions on the stationary coordinate system is weighted based on the information regarding the distance between each of the at least three specific processing regions, and the plurality of weighted plurality of coordinate positions are weighted. Positioning characterized by determining the coordinate position on the stationary coordinate system of each of the plurality of processing regions on the substrate by statistically calculating the coordinate position. The method allowed. 2. When determining the coordinate position of the arbitrary processing area on the substrate on the stationary coordinate system, the specific processing area having a shorter distance to the processing area is given to the coordinate position of the specific processing area. The weight is increased, and the weight is increased.
Positioning method described in. 3. The alignment method according to claim 2, wherein the weight given to the coordinate position of the specific processing area is changed according to the deformation state of the substrate. 4. The alignment method according to claim 3, wherein the deformation of the substrate is a non-linear distortion. 5. A plurality of processing areas regularly arrayed on a substrate according to a design array coordinate are aligned with a predetermined reference position in a stationary coordinate system that defines a moving position of the substrate. In the method, the coordinate position on the stationary coordinate system of at least three processing areas selected in advance as specific processing areas from among the plurality of processing areas is measured, and Based on the first information regarding the distance between the predetermined point of interest defined in advance on the substrate and the second information regarding the distance between the point of interest and each of the at least three specific processing regions, On the substrate by weighting each of the coordinate positions on the stationary coordinate system of the at least three specific processing regions and statistically calculating the weighted plurality of coordinate positions. Alignment method characterized by determining a coordinate position on the plurality of processing areas each of said stationary coordinate system. 6. When determining the coordinate position of the arbitrary processing area on the substrate on the stationary coordinate system, the distance to the target point is close to the distance between the target point and the processing area. The alignment method according to claim 5, wherein the weight given to the coordinate position of the specific processing area is increased as the processing area is increased. 7. The alignment method according to claim 5, wherein the point of interest is a center point of deformation of the substrate. 8. The alignment method according to claim 7, wherein the weight given to the coordinate position of the specific processing area is changed according to the deformation state of the substrate. 9. The alignment method according to claim 7, wherein the deformation of the substrate is a non-linear distortion. 10. The alignment method according to claim 7, wherein the deformation center point is substantially the center point of the substrate. 11. A plurality of processing regions, which are regularly arranged on a substrate according to a design arrangement coordinate, are aligned with a predetermined reference position in a stationary coordinate system which defines a moving position of the substrate. Method, at least 3 selected in advance from the plurality of processing areas as the specific processing area.
A measuring step of measuring coordinate positions of the two processing areas on the stationary coordinate system; a correcting step of correcting each of the coordinate positions of the at least three specific processing areas on the stationary coordinate system based on the flatness of the substrate. And; calculating a coordinate position on the stationary coordinate system of each of the plurality of substrates to be processed by statistically calculating the corrected plurality of coordinate positions. A positioning method, wherein each of the plurality of processing regions is sequentially positioned with respect to the reference position by controlling a moving position of the substrate according to a coordinate position and a flatness of the substrate. 12. A plurality of processing regions regularly arranged on a substrate in accordance with a designed arrangement coordinate are aligned with a predetermined reference position in a stationary coordinate system that defines a moving position of the substrate. In the method, when at least three processing regions are selected as the specific processing regions from the plurality of processing regions, and the surface of the selected specific processing regions and the moving plane of the substrate are substantially parallel to each other. A detecting step of detecting a coordinate position on the stationary coordinate system of each of the at least three specific processing regions; and a plurality of processing regions on the substrate by statistically calculating the detected plurality of coordinate positions. And a calculation step of calculating a coordinate position on the stationary coordinate system, according to the calculated coordinate position and an inclination amount for each processing region with respect to the moving plane of the substrate. By controlling the movement position of the plate, the alignment method characterized by registering successively position each of said plurality of processing regions relative to the reference position.