JPH05190423A - Projection aligner - Google Patents

Abstract

PURPOSE:To perform a calibration operation in each measuring point by using a best image formation face as a reference when a focal-point detection system according to an obliquely incident light system or the like is formed as a multipoint system. CONSTITUTION:A light-emitting mark (a slit) is formed on a Z-stage 20; the image of a beam of light from it is formed on the side of a reticle R by using a projection lens; a beam of reflected light is returned again to the side of the light-emitting mark; a change in the light quantity of a beam of returned light is detected by moving the Z-stage 20 up and down; and the offset value of the signal output of each measuring point in a multipoint AF system is stored.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for projecting and exposing an image of a circuit pattern or the like on a photosensitive substrate, and more particularly to a projection exposure apparatus equipped with a focus detection device for focusing the photosensitive substrate.

[0002]

2. Description of the Related Art In a conventional projection exposure apparatus, a reticle (mask) pattern is exposed to a photosensitive substrate (a wafer or glass plate coated with a resist layer) through a projection optical system having a high resolution.
When projecting and exposing the light onto the upper surface, the work of accurately matching the surface of the photosensitive substrate with the image forming surface of the reticle pattern, that is, focusing is essential. In recent years, the depth of focus of the projection optical system has been narrowed, while the wavelength of the exposure illumination light has been reduced to 36
Even with the 5 nm i-line, only a depth of about ± 0.7 μm can be obtained at present. Further, the projection field of view of the projection optical system tends to increase year by year, and it is becoming difficult to design and manufacture a projection optical system that secures a maximum depth of focus over the entire wide exposure field (for example, 22 mm square). .. Therefore, several methods have been proposed for increasing the depth of focus.

However, in order to achieve good focusing over the entire wide exposure field, the flatness of the partial area on the photosensitive substrate which falls within the exposure field and the flatness of the image plane, that is, the so-called image It is required that both the surface curvature and the image plane inclination are good. Of these, the curvature of field and the inclination of the field often depend on the optical performance of the projection optical system itself, but the flatness and parallelism of the reticle may be another factor. On the other hand, the flatness for each partial area on the photosensitive substrate, that is, for each projection exposure area (shot area) is
Although there is a difference in the degree depending on the photosensitive substrate, it is possible to set the surface of the shot area on the photosensitive substrate and the image plane in parallel by inclining the holder holding the photosensitive substrate by a minute amount.

As a method for focusing in consideration of the inclination of the surface of one shot area on the photosensitive substrate as described above, Japanese Patent Laid-Open Nos. 58-113706 and 55-1 are available.
The technique disclosed in Japanese Patent Publication No. 34812 is known.
Particularly, in JP-A-55-134812, spots of a light beam are projected onto four points on a photosensitive substrate through a projection optical system, and a spot image resulting from the reflected light is photoelectrically detected to focus the photosensitive substrate and correct inclination ( Leveling) is disclosed.

However, in the above-mentioned two conventional techniques, it is determined how much the surface of the photosensitive substrate deviates in the direction of the projection optical axis from the reference plane (which coincides with the image plane as much as possible) which is virtually set in advance. Only the detection is performed, and the method of directly detecting the deviation between the imaging surface and the surface of the photosensitive substrate is not used. Therefore, if a virtual reference plane that serves as a reference for measuring the positional deviation of the photosensitive substrate in the optical axis direction deviates from the image plane of the projection optical system due to drift or the like, the deviation is caused when the pattern is projected and exposed on the photosensitive substrate. Was the residual focus offset.

Therefore, a method for reducing such residual focus offset is disclosed in Japanese Patent Application Laid-Open No. 60-168112 or Japanese Patent Application Laid-Open No. 59-94032. Among them, in Japanese Patent Laid-Open No. 60-168112,
A reference pattern is provided on the stage (holder) on which the photosensitive substrate is placed, and the reference pattern is back-projected onto a specific pattern on the reticle via a projection optical system to form an image of the reference pattern formed on the specific pattern. After adjusting the height position of the stage to maximize the contrast, the surface on which the reference pattern is formed is the best focus surface (the best focus surface) due to the focus detection system (oblique incidence light type) for focusing the photosensitive substrate. The focus detection system is calibrated so that it is detected as the image plane). Further, in Japanese Patent Laid-Open No. 59-94032, a best focus plane is obtained by using a slit-shaped light receiving sensor as a reference pattern on a stage and detecting a contrast of a pattern image when a slit pattern on a reticle is projected by a projection optical system. Has been identified.

Another method is disclosed in Japanese Patent Laid-Open No. 1-2626.
As disclosed in Japanese Patent No. 24, a slit-shaped light emitting mark is provided on the stage, and the image of the light emitting mark is back-projected onto a specific mark on the reticle, and the stage is moved to X-axis.
There is also known a method of detecting the best focus surface by photoelectrically detecting the light transmitted above the reticle when the reticle mark is moved in the Y direction and scanned with the light emission mark image.

The above-mentioned Japanese Patent Laid-Open No. 58-11370.
As a development of the oblique incidence focus detection system disclosed in Japanese Patent No. 6, a pinhole image is formed on each of a plurality of points (for example, 5 points) in a shot area on a photosensitive substrate without passing through a projection optical system. A multi-point grazing incidence optical focus detection system of a type in which a two-dimensional position detection element (CCD) receives the reflected images in a grazing incidence manner in a batch is also known from Japanese Patent Laid-Open No. 102518/1990. There is. The method disclosed in this conventional example is generally called an oblique incidence type multipoint AF system, and can perform focus detection and tilt detection with high accuracy. However, in this conventional example,
There is no disclosure or suggestion of calibration of focus offset with respect to the best focus plane during projection exposure.

[0009]

In each of the above-mentioned conventional techniques, the system for detecting the positional deviation (focal point deviation) of the surface of the photosensitive substrate in the projection optical axis direction during the actual pattern exposure is not limited to the reticle pattern and the photosensitive substrate. Instead of directly detecting the in-focus state, the positional deviation in the optical axis direction of only the photosensitive substrate is detected. For this reason, it is ideal to calibrate from time to time in consideration of the drift on the device. Further, considering the flatness and parallelism of the photosensitive substrate, the multi-point AF system method, in which a plurality of points in the shot area are individually and substantially simultaneously focus-detected, is superior.

Therefore, when a multi-point AF system is adopted and it is necessary to calibrate the AF system, it has been found difficult to apply the conventional calibration method as it is. That is, in the above-described conventional calibration method, since it was necessary to detect a specific pattern on the reticle by some detection system, the specific pattern was always engraved in the peripheral portion of the circuit pattern area of the reticle or in the street line area. It was necessary to do so. Therefore, the positions of a plurality of measurement points within the projection field determined by the multipoint AF system and the position of the specific pattern on the reticle to be detected during calibration are completely different. Even if it is calibrated as it is, there is a problem in that accurate calibration cannot be performed due to the influence of the aberration of the projection optical system, the deflection of the reticle, and the like.

Further, even if it is not a multi-point AF system, it is a fixed-point AF system in which the focus measurement point on the oblique incident light type photosensitive substrate is set only in the center of the projection visual field, that is, in the center of the shot area. Had the same problem. The present invention has been made in view of these problems, and even if a measurement point of the AF system is set at an arbitrary position within the projection visual field, accurate calibration at that measurement point (focus calibration)
It is possible to improve the accuracy, reproducibility, stability and the like of the multipoint AF system.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a TT for detecting the best focus plane of a projection optical system in a projection exposure apparatus according to an embodiment of the present invention.
It is a figure which shows the focus detection system of L type. In FIG. 1, a reticle R having a circuit pattern area PA for manufacturing an actual device formed on its lower surface is held by a reticle holder (not shown). The optical axis AX of the projection lens PL, which is schematically represented by dividing the diaphragm surface (pupil surface) EP into a front group and a rear group, has the center of the reticle R, that is, the center of the pattern area PA, with respect to the reticle pattern surface. Pass vertically. Below the projection lens PL, the wafer stage W is placed and the Z stage 20 is moved by a small amount (for example, within ± 100 μm) in the optical axis AX direction.
Are provided on the XY stage 21. The XY stage 21 moves the wafer W two-dimensionally in an XY plane perpendicular to the optical axis AX. Furthermore, Z stage 20
A reference plate FM is fixed to a part of the reference plate FM at a height position substantially equal to the surface of the wafer W. This reference plate FM is shown in FIG.
As shown in FIG. 7, a mark ISy in which a plurality of transmission slits extending in the X direction are arranged at a constant pitch in the Y direction, and a plurality of transmission slits extending in the Y direction are arranged in a constant pitch in the X direction. A mark ISx and an oblique slit ISa which is 45 ° with respect to each of the X and Y directions are formed. These slit marks ISx, ISy, IS
In a, a chromium layer (light-shielding layer) is vapor-deposited on the entire surface of the quartz reference plate FM, and is engraved as a transparent portion there.

Now, in FIG. 1, below the reference plate FM (Z
Inside the stage 20), a mirror M1, an illumination objective lens 40, and an optical fiber 41 are provided, and the illumination light from the exit end of the optical fiber 41 is condensed by the objective lens 40 to form a slit mark ISx of the reference plate FM. , I
Both Sy and ISa are irradiated from the back side. A beam splitter 42 is provided on the incident end side of the optical fiber 41,
The exposure illumination light IE is introduced into the optical fiber 41 via the lens system 43. The illumination light IE is preferably obtained from a light source for reticle R illumination (a mercury lamp, an excimer laser, etc.), but a dedicated light source may be prepared separately. However, when using a different light source, it is necessary to use illumination light having the same wavelength as the exposure illumination light or a wavelength very close to it.

The illumination condition of the reference plate FM by the objective lens 40 is matched as much as possible with the illumination condition of the projection lens PL at the time of pattern projection. That is, the numerical aperture (NA) of the illumination light on the image side of the projection lens PL and the numerical aperture (NA) of the illumination light from the objective lens 40 to the reference plate FM are substantially matched. Now, with such a configuration, the illumination light I
When E is introduced into the optical fiber 41, the image light fluxes propagating to the projection lens PL are generated from the marks ISx, ISy, ISa on the reference plate FM. In FIG. 1, the Z stage 2
0 is the best image plane (reticle conjugate plane) F of the projection lens PL
It is assumed that the reference plate FM is set slightly downward from o. At this time, the image light flux L1 generated from one point on the reference plate FM passes through the center of the pupil plane EP of the projection lens PL, is condensed in a plane Fr slightly shifted downward from the pattern surface of the reticle R, and then diverges, After being reflected by the pattern surface of the reticle R, the original optical path is returned. Where surface Fr
Is in a position common to the reference plate FM with respect to the projection lens PL. When the projection lens PL is a telecentric system on both sides, the emission marks ISx, ISy, ISa on the reference plate FM are displayed.
The image light flux from is normally reflected on the lower surface (pattern surface) of the reticle R and returns so as to be superimposed again on the marks ISx, ISy, ISa. However, as shown in FIG. 1, the reference plate FM
Is deviated from the image plane Fo, a blurred reflection image of each mark ISx, ISy, ISa is formed on the reference plane FM, and when the reference plate FM coincides with the image plane Fo, Fr also matches the pattern surface of the reticle R,
On the reference plate FM, sharp reflection images of the marks ISx, ISy, ISa are formed so as to overlap the respective marks. FIG. 3 schematically shows the relationship between the light emission mark ISx and the reflected image IMx when the reference plate FM is defocused. In the projection lens PL which is telecentric on both sides, the reflected image IMx is projected on the emission mark ISx which is the source of the reflected image IMx. When the reference plate FM is defocused, the reflected image IMx is
The size is larger than the shape size of the mark ISx, and the illuminance per unit area is also reduced.

Therefore, of the reflected image formed on the reference plate FM, the light flux of the image portion which is not shielded by the original marks ISx, ISy, ISa is received by the optical fiber 41 via the mirror M ₁ and the objective lens 40, and the beam Splitter 42,
The photoelectric sensor 45 receives light through the lens system 44. The light receiving surface of the photoelectric sensor 45 is arranged substantially conjugate with the pupil surface (Fourier transform surface) EP of the projection lens PL.

In the various known examples described above in the prior art,
A light emitting type mark, a light receiving type mark, or a simple reflection type reference mark is arranged on the image plane side of the projection lens, and an electrical signal for contrast detection is generated by optically or electrically scanning the reticle mark in the XY plane. I was getting. However,
In the configuration of FIG. 1, it is not necessary to scan the reference plate FM in the XY plane to obtain an electric signal representing a contrast change as in the conventional case, and the contrast signal can be obtained by simply moving the Z stage 20 in the vertical direction (Z direction). Can be obtained.

FIG. 4 shows the signal level characteristic of the output signal KS of the photoelectric sensor 45, and the horizontal axis represents the position of the Z stage 20 in the Z direction, that is, the height position of the reference plate FM in the optical axis AX direction. In FIG. 4, FIG. 4A shows the emission mark ISx,
FIG. 4B shows the signal levels when ISy and ISa are back-projected on the chrome portion in the pattern surface of the reticle R.
Indicates the signal level when back projected onto the glass portion (transparent portion) in the pattern surface. Usually, the chrome portion of the reticle is vapor-deposited on a glass (quartz) plate with a thickness of about 0.3 to 0.5 μm, and the reflectance of the chrome portion is, of course, much higher than that of the glass portion. However, since the reflectance at the glass portion is not completely zero, the level is considerably small as shown in FIG. 4B, but detection is possible. Generally, since the reticle for manufacturing an actual device has a high pattern density, it is considered that the probability that all the back projected images of the emission marks ISx, ISy, ISa are simultaneously applied to the glass part (transparent part) in the reticle pattern is extremely small. Be done.

In any case, when the surface of the reference plate FM is moved in the optical axis direction so as to cross the best image plane Fo, the signal level becomes a maximum value at the position Zo in the Z direction. Therefore, by simultaneously measuring the Z direction position of the Z stage 20 and the output signal KS and detecting the Z direction position when the signal level becomes maximum, the position of the best imaging plane Fo can be obtained, and this detection is performed. According to the method, the image plane Fo can be detected at any position in the reticle R. That is, it is not necessary to use a pattern (mark) formed at a specific position in the reticle R as in the prior art, and as long as the reticle R is set on the object side of the projection lens PL, it is always in the projection visual field. The absolute focus position (best image plane Fo) can be measured at the position. Also, as mentioned above, the reticle R
Has a thickness of 0.3 to 0.5 μm, and the detection error of the best imaging plane Fo caused by this thickness is caused by the projection lens PL.
If the projection magnification of is reduced to 1/5, (0.3 to 0.5) ×
(1/5) ² = 0.012 to 0.02 μm, which is almost negligible.

Next, an oblique incident light type AF system will be described with reference to FIG. 5, but here, multipoint AF is adopted. The multipoint AF is one in which measurement points for measuring the positional deviation of the wafer W in the optical axis direction (so-called focal point deviation) are provided at a plurality of positions within the projection visual field of the projection lens PL. In FIG. 5, the illumination light I, which is non-photosensitive to the resist on the wafer W, is used.
L illuminates the slit plate 1. The light that has passed through the slits of the slit plate 1 has a lens system 2, a mirror 3, a diaphragm 4,
The wafer W is obliquely irradiated via the projection objective lens 6 and the mirror 6. At this time, when the surface of the wafer W is on the best imaging plane Fo, the image of the slit of the slit plate 1 is imaged on the surface of the wafer W by the lens system 2 and the objective lens 5. The angle between the optical axis of the objective lens 5 and the wafer surface is 5
The center of the slit image of the slit plate 1 is located at a point where the optical axis AX of the projection lens PL intersects the wafer W.

The slit image light flux reflected by the wafer W is reflected by the mirror 7, the objective lens 8 for receiving light, the lens system 9, the vibrating mirror 10, and the plane parallel plate (plane parallel) 12.
An image is re-formed on the light-receiving slit plate 14 via. The vibrating mirror 10 microvibrates a slit image formed on the light-receiving slit plate 14 in a direction orthogonal to its longitudinal direction, and the plane parallel 12 forms a slit on the slit plate 14 and a reflected slit image from the wafer W. The relative relationship with the vibration center is shifted in the direction orthogonal to the slit longitudinal direction. The vibrating mirror 10 is vibrated by a mirror driving unit (M-DRV) 11 driven by a driving signal from an oscillator (OSC.) 16.

When the slit image oscillates on the light-receiving slit plate 14 in this manner, the light flux transmitted through the slit of the slit plate 14 is received by the array sensor 15. This array sensor 15 is one in which the longitudinal direction of the slit of the slit plate 14 is divided into a plurality of minute regions, and individual photoelectric cells are arranged in each minute region. Here, an array sensor of a silicon photodiode or a phototransistor is used. Shall be used. Then, the signals from the respective light receiving cells of the array sensor 15 are selected via the selector circuit 13,
Alternatively, they are grouped and input to the synchronous detection circuit (PSD) 17. This PSD 17 has an OSC. An AC signal having the same phase as the drive signal from 16 is input, and synchronous rectification is performed based on the phase of this AC signal. At this time PSD
Reference numeral 17 includes a plurality of detection circuits for individually synchronously detecting respective output signals of a plurality of light receiving cells selected from the array sensor 15, and the respective detection output signals FS are provided to the main control unit (MCU) 30. Is output to. Detection output signal FS
Is a so-called S-curve signal, and becomes zero level when the slit center of the light-receiving slit plate 14 and the vibration center of the reflection slit image from the wafer W coincide with each other.
Is a positive level when is displaced upwards from that state,
It is at a negative level when the wafer W is displaced downward. Therefore, the wafer W whose output signal FS becomes zero level
The height position of is detected as the focal point. However, in such an oblique incident light system, there is no guarantee that the height position of the wafer W at the in-focus point (the output signal FS is zero level) always matches the best image plane Fo. That is, in the oblique incident light system, the system has a virtual reference plane determined by the system itself, and when the wafer surface coincides with the reference plane, PS
Since the D output signal FS becomes zero level, the reference surface and the best image forming surface Fo are set so as to coincide with each other as much as possible at the time of manufacturing the device, but they coincide with each other for a long period of time. There is no guarantee. Therefore, by inclining the plane parallel 12 in FIG. 5 to displace the virtual reference plane in the optical axis AX direction, the reference plane and the best imaging plane Fo can be matched (or the positional relationship can be defined). it can.

In FIG. 5, the MCU 30 inputs the output signal KS from the photoelectric sensor 45 to calibrate the oblique incident light type multipoint AF system, the function to set the inclination of the plane parallel 12, and the multipoint AF. Based on each output signal FS of the system, the motor 19 for driving the Z stage 20
Signal DS to the circuit (Z-DRV) 18 that drives the
And a function of outputting a command to a drive unit (including a motor and its control circuit) 22 that drives the XY stage 21.

FIG. 6 shows the projection visual field If of the projection lens PL,
It is the figure which looked at the positional relationship with the projection slit image ST of AF system on the wafer surface. The projection visual field If is generally circular, and the pattern area PA of the reticle R is rectangular included in the circle. The slit image ST is formed on the wafer with an inclination of about 45 ° with respect to each of the moving coordinate axes X and Y of the XY stage 21. Therefore, both optical axes AFx of the light projecting objective lens 5 and the light receiving objective lens 8 extend in the direction orthogonal to the slit image ST on the wafer surface. Further, the center of the slit image ST is set so as to substantially coincide with the optical axis AX. With such a configuration, the slit image ST is set to extend as long as possible within the projection area of the pattern area PA. Generally, on the surface of the wafer onto which the pattern area PA is projected, a shot area to be superposed thereon is already formed. Within the shot region, the variation of the uneven portion increases every time the device manufacturing process progresses, and a large uneven change may exist in the longitudinal direction of the slit image ST. Particularly when a plurality of chips are arranged in one shot area, a scribe line for separating each chip is formed in the shot area so as to extend in the X direction or the Y direction. In the extreme case, a step difference of 2 μm or more may occur between the point on the chip and the point on the chip. Since which part in the slit image ST the scribe line is located in is known in advance from the designed shot arrangement, the chip size in the shot, etc., the slit image S
It is naturally understood from which part of the longitudinal direction of T the reflected light should be selected on the array sensor 15.

FIG. 7 shows an example of a concrete structure of the light-receiving slit plate 14 and the array sensor 15.
4 is a chrome layer (light shielding) deposited on the entire surface of the glass substrate,
A transparent slit is formed in a part of it by etching. The slit plate 14 is fixed to a holding metal piece 14A, and the metal piece 14A is fixed to a printed circuit board 15A such as ceramics holding the array sensor 15. As a result, the slits of the slit plate 14 are in parallel with the one-dimensional array of light receiving cells of the array sensor 15 and are in close contact therewith. Thus, the slit plate 14 and the array sensor 15
It is better to contact with or as close as possible to the slit plate 1
An imaging lens system is provided between 4 and the array sensor 15,
The slit plate 14 and the array sensor 15 may be optically conjugated. Although the length of the slit image ST on the wafer shown in FIG. 6 varies depending on the diameter of the projection visual field If, the diameter of the visual field If is 32 with the 1/5 reduction projection lens.
When it is around mm, it is desirable to make it about 1 to 1/3 times the diameter.

FIG. 8 shows an example of a concrete processing circuit of the array sensor 15, the selector 13, the PSD 17, and the MCU 30. Here, the light receiving cells in the array sensor 15 are divided into three groups Ga, Gb, Gc, It is assumed that the light receiving cells are selected and integrated in each group. The group Gb is for detecting the central portion of the slit image ST, and the groups Ga and Gc are for detecting both ends of the slit image ST. First, since a plurality of light receiving cells in the array sensor 15 are included in the group Ga, at least one of the light receiving cells is selected by the selector 13A and its output signal is output to the PSD 17A.
Output to. The selector 13A selects an arbitrary one of the light receiving cells in the group Ga and outputs its output signal to P
In addition to the function of sending to the SD17A, it also has a function of arbitrarily selecting two or three adjacent light receiving cells in the group Ga and sending a signal obtained by adding their output signals to the PSD17A. The output signals from the light receiving cells in the groups Gb and Gc are similarly processed by the selectors 13B and 13C, and the selected signals are sent to the PSDs 17B and 17C. PSD 17A, 17B, 17C are OS respectively
C. Detection output FS by inputting the fundamental wave AC signal from 16
It outputs a, FSb, and FSc. These detection outputs FS
Each of a, FSb, and FSc is converted into digital by an analog-digital converter (ADC) 30A in the MCU 30, and sent to the correction calculation unit 30B. The correction calculation unit 30B calculates a value corresponding to the target position in the Z direction of the Z stage 20 based on the values of the three detection outputs, that is, the focus shift amounts at the three points, and calculates the value as the deviation. Output to the detection circuit 30C. This detection circuit 30C is an ADC 30A
Z-DR as the command signal DS with the difference from the detection output value from
Output to V18. At this time, the correction calculation unit 30B causes the detection outputs FSa, F stored in advance in the storage unit 30D.
Various calculations are performed by introducing offset values for Sb and FSc, respectively. This offset value is measured and calculated by the calibration value determination unit 30E, and the determination unit 3
0E inputs the three detection outputs FSa, FSb, FSc and the output signal KS of the photoelectric sensor 45, and calculates the deviation between the reference surface of the multipoint AF system itself and the best focus surface Fo from the zero level on the detection output. The deviation voltage is calculated as

Since the deviation detection circuit 30C selects any one of the three detection output values from the ADC 30A and compares it with the target value from the arithmetic unit 30B, the difference becomes zero or a predetermined value. Such a servo system is assembled. The calibration value determination unit 30E also includes an AD converter, a waveform memory, etc. for simultaneously digitally sampling the respective levels of the three detection outputs and the level of the signal KS (FIG. 4).

Here, referring to FIG. 9, a calibration value determination unit 30E
A specific example of the configuration will be described. First, the output signal KS from the photoelectric sensor 45 of the absolute focus detection system of the TTL (through the lens) system is an analog-digital converter (AD).
C) Input to 300, converted into a digital value corresponding to the signal level, and stored in the memory (RAM) 301. This RAM 301 is addressed by a counter 304
The increment of the counter 304 and the conversion timing of the ADC 300 are both in response to the clock pulse from the clock generator (CLK) 303. Similarly, three detection output signals FSa,
One of FSb and FSc is input to the ADC 305 via the selection switch, and the converted digital value is addressed by the counter 307 RAM 306.
Memorized in. Therefore, the RAMs 301 and 306 respectively capture the temporal changes in the waveforms of the output signal KS and one detection output. The waveforms in the RAMs 301 and 306 are used by the arithmetic processing unit 310 as processing data such as smoothing and maximum value detection. In the arithmetic processing unit 310, a signal for instructing the Z stage 20 to move at a constant speed is output to the Z-DRV 18 in order to capture the signal waveforms in the RAMs 301 and 306, and each measurement of the multipoint AF system is performed. A signal for moving the center of the emission mark to the position of the point is output to the XY-DRV 22.

FIG. 10A shows the time change characteristic of one detection output signal, and RAM 306 when the Z stage 20 is moved at a constant speed within a certain range including the best focus plane.
10B shows the waveform inside, and FIG.
01 represents the waveform of the signal KS obtained in 01. Since the synchronous detection signal has a waveform that is substantially point-symmetric with respect to the zero point, negative level data smaller than the zero point is AD converted in consideration of the negative level.

Since the maximum value waveform of the signal KS is stored in the RAM 301 with the time axis as an address,
The arithmetic processing unit 310 analyzes the waveform of FIG. 10B to obtain the time T ₁ when the maximum point is obtained. This time T ₁ is RA
It uniquely corresponds to a specific address point in M301. Next, the arithmetic processing unit 310 obtains the corresponding address point in the RAM 306, and obtains the level value ΔFS of the detection output signal stored at that address point. This level value ΔFS is an offset voltage from the zero point on the detection signal, and the detection output is + ΔF at the measurement point of the multipoint AF system that generates the detection output as shown in FIG. 10 (A).
When the wafer surface at the measurement point is moved in the Z direction so as to be S, the wafer surface and the best focus surface Fo coincide with each other.

By the way, when using the circuit of FIG.
The stage 21 is moved so that the center of the light emitting mark on the reference plate FM is positioned at any one of the measurement points of the multipoint AF system. The positioning does not have to be so strict, and the measurement points of the multipoint AF system and the center of the light emitting mark group may deviate by about 100 μm in the X and Y directions. Therefore, when the measurement point of the multi-point AF system, that is, the point in the slit image ST shown in FIG. 6 is determined, the position of the light emission mark group is set to X, within a range of about ± 100 μm around the point.
A coordinate position at which the peak of the signal KS becomes large to some extent may be obtained by shifting in the Y direction and shaking in the Z direction. Although this is extremely small in probability, the inconvenience that the entire emission mark group coincides with the transparent portion of the reticle R (signal K
This is to prevent the S / N ratio of S from decreasing as much as possible. However, when the calibration operation is performed at high speed, it is possible to obtain the offset value ΔFS with substantially the same accuracy without searching for the coordinate position where the peak of the signal is particularly large.

FIG. 11 shows three measurement points MP of the multipoint AF system.
a, MPb, MPc and the existence range CAa, CAb, CAc of the center point of the light emitting mark group are schematically shown, and the optical axis AX of the projection lens PL is at the intersection of the X axis and the Y axis.
Shall pass. Each measurement point MPa to MPc corresponds to the position of the light receiving cell on the array sensor 15.
Further, PA indicates the range when the pattern area of the reticle R is projected, that is, the range of one shot area on the wafer. In the present embodiment, three points on the slit image ST are set as measurement points, the central measurement point MPb is set near the shot center point, and the other measurement points MPa and MPc are set near both ends of the slit image ST. And Further, in order to simplify the following description, it is assumed that there are no scribe lines on the X axis and the Y axis in FIG. 11 and the range PA is one chip.

Therefore, three measuring points MPa to M shown in FIG.
Aligning the centers of the emission mark groups with Pc
The stage 20 (reference plate FM) is moved in the Z direction from the top to the bottom (or from the bottom to the top) only once at a substantially constant speed, and as described in FIG. 10, the absolute focus detection system of the TTL system and the oblique incident light system are used. The offset amounts ΔFSa, ΔFSb, and ΔFSc at each measurement point are individually obtained by simultaneously using the multipoint AF system of (1) and stored in the storage unit 30D in FIG. During the operation for this calibration, the surface of the reference plate FM is detected by the oblique incident light type multi-point AF system, so that the emission mark groups ISx, ISy, ISa, etc. are exactly on the slit image ST. When it is positioned and the detection output FS at each measurement point is affected, the positions of the light emission marks are shifted within the existing range CAa to CAc in FIG. 11, and the slit image ST is slightly displaced from the light emission marks. The light may be projected onto the reflection surface portion of the reference plate FM.

The offset amounts ΔFSa, ΔFSb, and ΔFSc (unit: voltage V) at the respective measurement points thus obtained are all in the optical axis AX direction in which each of the three points of the multipoint AF system is the in-focus point (zero point). The height position of is the best focus plane F
a constant distance offset ΔZ with respect to the corresponding point in o
It means that it has a, ΔZb, and ΔZc.
If the slope k (unit V / μm) of the straight line portion including the zero point on the waveform of the detection output signal shown in (A) is known, the offsets ΔZa to ΔZ as the distance in the Z direction are calculated by the following equation.
c is obtained.

ΔZa = ΔFSa / k ₁ , ΔZb = ΔFS
b / k ₂ , ΔZc = ΔFSc / k ₃ Here, k ₁ , k ₂ , and k ₃ originally have the same value, but strictly speaking, they are individually different, so it is desirable to obtain them in advance for each measurement point. .. FIG. 12 is an example schematically showing the appearance of the offset amount ΔFS, and FIG. 12 (A) shows 3 points MPa,
Offset amounts ΔFSa and ΔFS of MPb and MPc, respectively
It shows a case where both b and ΔFSc have positive values and there is a substantially constant inclination in the extending direction of the slit image ST. Various factors can be considered as factors that represent such an offset tendency, but they can be divided into two problems: a problem on the multipoint AF system side and a problem on the reticle R side. First, as a problem on the side of the multipoint AF system, when the slit image ST from the light-projecting slit plate 1 is re-imaged on the light-receiving slit plate 14, the parallelism with the slit of the slit plate 14 is slightly increased. It may be crazy. Further, the problem on the reticle R side is that the reticle pattern surface is inclined at least in the longitudinal direction of the slit image ST, or the image surface inclination of the projection lens PL itself is considered.

In any case, since the problems of both of them are compounded, it is difficult to specify them independently of each other. Further, FIG. 12B shows the offset amount ΔF of the central measurement point MPb.
A case where Sb is a negative value and offset amounts ΔFSa and ΔFSc at the measurement points MPa and MPc at both ends are positive values and have substantially the same magnitude is shown. The main cause of such characteristics is a problem on the reticle R side, and the curvature of the reticle pattern surface and the curvature of field of the projection lens PL itself are considered. However, the curvature of field of the projection lens PL has been suppressed sufficiently small by inspection and adjustment during manufacturing of the apparatus, and even if it remains slightly, the degree of curvature of field is quantitatively obtained beforehand. Figure 12
If you consider it by subtracting it from the characteristics of (B), the reticle R will be used exclusively.
It can be seen whether or not the curvature is the main factor.

Further, FIG. 12C shows a case where the inclination tendency and the bending tendency of FIG. 12A are combined. Therefore, a specific example of a series of calibration sequences in consideration of the tendency of offset at each measurement point as shown in FIG. 12 will be described. However, in principle, each measurement point MPa to MPc
Since the offset amounts ΔFSa to ΔFSc obtained in step 6 are calibration values at the measurement points, for example, when the surface of the wafer W is detected at the measurement points MPa, the detection output signal F
If the Z stage 20 is driven so that Sa becomes ΔFSa, the wafer surface and the best focus surface Fo can be matched at the measurement point MPa.

However, the actual focusing is not always performed with only one point, and it may be necessary to decide it in consideration of the surfaces. In addition to this, it is also necessary to consider the depth of focus of the projection lens PL. FIG. 13 shows a flow chart of the measurement operation in the calibration sequence, in which multipoint A
The F-system measurement points are generalized to n points, the measurement points shown in FIG. 11 are MPi (where i = 1 to n), and the area where the center of the emission mark group is to be positioned is CAi (where i = 1 to 1).
n), the synchronous detection output signal at each measurement point is FSi (i = 1
Up to n) and the offset amount is ΔFSi (i = 1 to n).

First, in step 100, i = 1 is set,
In step 101, the XY stage 21 is moved to move to the area C.
The center of the light emitting mark group is positioned in Ai. Next, in step 102, the Z stage 20 is positioned in the vicinity of the start point of the signal waveform to be stored in the RAMs 301 and 306 of FIG. 9 based on the detection output signal FSi at the measurement point MPi. In this case, since it is necessary to capture the waveforms shown in FIG. 10 in the RAMs 301 and 306, the detection output signal FS
The Z stage 20 is set at a position where i is outside the maximum value or the minimum value as shown in FIG. In order to do this easily, the focus servo system is operated in the straight line portion including the zero point of the detection output FSi, and the servo enable signal that prohibits the feedback to the servo system is made outside the straight line portion, and the servo enable signal is set to the servo enable signal. The Z stage 20 may be positioned based on this.

Next, in step 103, the light emitting mark group is lit with light having the same wavelength as the exposure light, and the Z stage 20 is moved in one direction at substantially the same speed, while the RAM 30 in the circuit of FIG.
Waveforms of the detection output signal FSi and the signal KS are digitally sampled for each of 1 and 306. When the waveform acquisition is completed, the offset amount ΔFSi is calculated in step 104 and stored in the storage unit 30D. Then step 1
In 05, it is determined whether or not i = n. When i has not reached the n-th, i is incremented by 1 in step 106 and the process returns to step 101 to repeat the same sequence for other measurement points. If i = n is determined to be true in step 105, the sequence of FIG. 14 is executed. Note that in the sequence of FIG. 13, step 102 does not necessarily have to be executed for all measurement points, and for example, just before step 101, only measurement points near the center in the visual field If of the multipoint AF system are measured. Alternatively, the waveform acquisition start point (the height position of the Z stage 20) stored at that time may be applied at the time of measurement at another measurement point.

Now, FIG. 14 shows a unique calibration method in the multipoint AF system, and shows an example of performing the calibration in consideration of various factors explained in FIG. First, in step 110, it is determined whether or not the averaging calibration considering all the plurality of measurement points of the multipoint AF system is performed. In this case, the offset amount Δ at each measurement point
When only FSi needs to be obtained (for example, when it is premised on pinpoint AF that individually focuses at each measurement point), the series of calibration processes ends. When performing averaging calibration, the following step 11
At 1, the range of each offset amount ΔFSi is checked.
The algorithm for the check is whether the difference between the offset amounts at two adjacent measurement points is extremely large, or the absolute value of the offset amount at each measurement point is the upper limit or the lower limit of the linear range of the detection signal FS. It is possible to adopt whether or not you are close to.

In the next step 112, as a result of the check in step 111, it is judged whether or not there is an offset amount that has become an abnormal value. If an abnormal value exists, the process proceeds to step 113. At this time, the number of measurement points that become abnormal values and how much the position is allowed are determined by the reticle R mounted on the reticle R.
Pattern is determined due to the surface structure in the shot area on the wafer W on which the pattern of FIG.

Now, in step 113, it is determined whether or not the measurement operation of FIG. 13 is to be performed again. Whether or not to perform the remeasurement is set in advance by the operator when setting the parameters of the device operating conditions. And step 11
When it is determined in 3 that the re-measurement is performed (retry), it is determined in the next step 114 whether or not the re-measurement has been performed the preset number of times. If it is determined in step 114 that the number of times is not exceeded, the sequence from step 100 in FIG. 13 is repeatedly executed. If it is set in step 113 that the remeasurement is not executed even once, the process proceeds to step 118, an error display and warning indicating that an abnormal value exists is generated, and the series of sequences is ended. ..

When the number of times is further exceeded in step 114, the process proceeds to the next step 115, and it is determined whether or not the setting for realigning the reticle is set.
As described above, when there is a large variation for each offset amount ΔFSi, there is a problem in the reticle R as one factor. For example, when the reticle R is sucked onto the holder, a relatively large foreign substance or the like may be sandwiched between the suction surface and the reticle R. Alternatively, the adsorption may cause the reticle R to bend. Therefore, by removing the reticle R from the holder once and re-aligning it, the inclination of the pattern surface of the reticle R,
The deflection is changed to the desired one.

When it is determined in step 115 that reticle realignment is to be performed, the next step 116
Then, it is judged whether or not the number of times exceeds the number designated in advance. Here, when the reticle realignment number exceeds the set number (step 116) or when the reticle realignment execution is not set (step 115), the process proceeds to step 118 and an error display or warning is generated. To do. And step 11
If it is determined in 6 that the number of times has not exceeded, in step 117, the reticle R is realigned (remove the reticle R once from the holder and again attracted to the holder for alignment), and then the process proceeds to step 100 in FIG. At this time, the counter for the number of remeasurements determined in step 114 is reset to zero.

On the other hand, when it is determined in the previous step 112 that there is no abnormal value, each offset amount ΔF is determined in step 120.
Calculate the average value ΔFA by adding and averaging Si, and
Each offset amount ΔFSi centered on the average value ΔFA at 21
(The deviation amount of ΔFSi from ΔFA) is checked. If the variance is larger than a predetermined value, the process proceeds to step 122, and if the variance is small, the process proceeds to step 130.

FIG. 15 shows the average value ΔFA and the variance.
Will be described. FIG. 15A shows a case where the offset amount ΔFSi at each measurement point (here, 5 points) has a positive value and has a tendency to be uniformly inclined.
The average value ΔFA at this time appears at approximately the midpoint of the uniform slope characteristic (broken line). Further, the dispersion increases or decreases on the peripheral side depending on the inclination of the inclination characteristic. Therefore, small dispersion means that multi-point AF
It means that the focusing reference plane of the system itself and the best focus surface Fo have good parallelism, and it can be approximated to the best focus plane Fo by simply shifting the focusing reference plane of the multipoint AF system itself in parallel. become.

FIG. 15B shows the characteristics when the field curvature, reticle deflection, etc. are combined with the relative inclination. The average value ΔFA may be small, but the dispersion (ΔFSi
The amount of deviation from ΔFA) may increase as a whole. Of course, the smaller the degree of bending and the degree of relative inclination, the smaller the dispersion. Therefore, if it is determined in step 121 that the variance is small to some extent, in step 130, the average value Δ
Judge whether FA is large to some extent. When the average value ΔFA is relatively large, it is determined in step 131 whether or not the offset correction by the plane parallel (harbing) 12 shown in FIG. 5 is performed. Harving 1
In No. 2, the focusing reference plane (the plane defined by the zero point of the detection output FS) of the multipoint AF system itself can be translated in the Z direction as a whole by adjusting the inclination. Therefore, in step 131, it is selected to perform the harving correction, and in the next step 132, the harving 12 is selected.
When the correction is performed by, the offset amount ΔFSi is corrected so that the average value ΔFA shown in FIG. That is, after the harving 12 is tilted by the interval ΔZf in the Z direction corresponding to the average value ΔFA in step 132, the true offset amount ΔFSi at each measurement point is treated as corrected to ΔFSi = ΔFSi−ΔFA. You can

However, the correction amount by the harving 12 is
It is not always necessary to exactly match the average value ΔFA, and the point is that the correction amount by the harving 12 should be accurately obtained. Next, at step 133, it is judged whether or not the offset amount at each measurement point is to be measured again, and when re-measured, the routine proceeds to step 100 in FIG. This step 13
The selection in 3 is performed according to the parameters preset by the operator.

When it is determined in step 133 that re-measurement is unnecessary (when the correction amount due to harving is accurately obtained, etc.), each offset amount ΔFSi is calculated by the amount corresponding to the harving correction amount (ΔFA ′) in the next step 134. Correct and memorize. That is, ΔFSi = ΔFSi−ΔF
Calculate A '. As described above, the process when the average value ΔFA is large is completed in step 130, and the process proceeds to the next step 122. This step 122 is the same as the previous step 1.
When it is determined that the variance is large in Step 21, or Step 1
It is also a jump destination when it is determined that the average value ΔFA is small in 30.

In step 122, an approximate plane (or an approximate straight line) is specified by the least square method using each offset amount ΔFSi thus obtained. This approximate plane (or straight line)
15A, under the offset amount ΔFSi having the uniform inclination characteristic as shown in FIG. 15A, the inclination characteristic (broken line) becomes an approximate plane, and the curved characteristic (broken line) as shown in FIG. 15B. With an offset amount ΔFSi having
It becomes an approximate plane (or a straight line) like the inclined line Q in the inside.

Next, in step 123, the amount of inclination of the approximate plane, that is, the amount of inclination of the line Q in FIG. 15B with respect to the zero level is obtained, and in the next step 124, it is judged whether or not the amount of inclination is greater than or equal to the allowable value. To be done. The tilt amount is a relative tilt between the best focus surface Fo and the approximate plane determined so that the deviation is smallest everywhere with respect to the offset amount ΔFSi at each measurement point of the multipoint AF system. If this inclination is large, there is a possibility that a part of the shot area may be out of focus due to the relationship between the depth of focus of the projection lens PL and the size of the shot area, that is, an intra-shot focus error may occur. Therefore, when the inclination of the approximate plane is extremely large, it is determined that there is some problem at the time of measurement or on the reticle R side, and step 113 described above is used.
~ 118 are executed.

When it is determined in step 124 that the inclination of the approximate plane is within the allowable value, the final step 125 is executed. In this step 125, the offset amount ΔFSi at each measurement point is set to ΔF using the approximation plane as a new reference.
Calculate the correction to Ji. Specifically, for example, FIG. 15 (B)
At the measurement point where the offset amount is ΔFS ₃ ,
The offset amount ΔFJ ₃ of the corresponding point on the approximate plane is replaced. By doing so, the Z stage 20 is moved up and down and tilted so that the detection output signal FSi obtained at each measurement point will have the offset amount ΔFJi, thereby achieving good focusing over the entire shot area. Will be done. For that purpose, it is desirable that the Z stage 20 is provided with a leveling mechanism and the entire wafer W is tilted in the same direction by the tilt of the approximate plane.

Although a series of calibration operations are completed as described above, it is sufficient in principle to execute the sequence of FIGS. 13 and 14 once when reticle conversion is performed and reticle alignment is completed. In the exposure process using the same reticle, the initial offset amount ΔFSi or ΔFJi may be changed due to a slight drift of the multipoint AF system, a temperature change in the apparatus, an error at the focus position correction due to the pressure control in the projection lens PL, and the like. It can be crazy. Therefore, the sequence of FIGS. 13 and 14 may be executed for each lot (usually 25) during the wafer exposure process.

Further, in the present embodiment, when a tendency such as image plane inclination appears, realignment of the reticle is executed in steps 115, 116 and 117. Alternatively, the holder of the reticle R may be inclined. It is possible. In this case, the best focus plane Fo and the focusing reference plane (or approximate plane) of the multi-point AF system can be set parallel to each other, but the image of the pattern area projected on the wafer W is trapezoidal distortion or rhombus distortion. Etc. may occur. Therefore, the distortion can be reduced by inclining some of the optical lenses near the reticle side in the projection lens PL from the plane perpendicular to the optical axis AX by a small amount.

Further, a tendency such as image plane inclination also appears due to the slit parallelism between the reflection slit image ST of the multipoint AF system and the light-receiving slit plate 14, so that, for example, FIG.
When the inclination tendency as in (A) is strongly generated and does not change even by realignment of the reticle, the slit plate 1 for light projection or the slit plate 14 for light reception is relatively automatically moved by a slight amount according to the inclination tendency. The slit parallelism may be adjusted by rotating the slit.

By the way, in the projection method of the slit image ST shown in FIGS. 6 and 11, the angle is 45 ° with respect to the XY coordinate axes.
However, the multi-point AF system lacks the ability to specify a surface. Therefore, another set of multipoint AF systems may be arranged in the 135 ° direction with respect to the X and Y axis directions, and the slit image ST ′ as shown in FIG. 16 may be projected at the same time. In this way, nine measurement points (see FIG. 1) are included in the vicinity of the four corners in the pattern area PA (shot area on the wafer), the center, and the like.
6) can be set. For that purpose, 5 points are selected on each of the two sets of array sensors, and a total of 10 (or 9) synchronous detection circuits are required.

In addition, the multipoint AF system projects a minute slit image or a spot image obliquely onto a plurality of predetermined measurement points within the projection visual field If, and reflects slits (spots) from the respective measurement points. The same effect as that of the present invention can be obtained by a method in which images are collectively received by a two-dimensional image pickup device, for example, a CCD, and a focus shift is obtained by a shift amount of a plurality of received images from a reference pixel position on each CCD. can get. Further, as described in the prior art, it can be similarly applied to a multi-point AF system in which spot light is projected onto four points on a wafer through only a projection lens.

Although the sequence for calibration has been mainly described with reference to FIGS. 13 and 14, some methods of operating the multipoint AF system after further calibration will be described. First, when the offset amounts ΔFSi at the n measurement points of the multipoint AF system are obtained, the offset amounts ΔFSi are used to execute steps 122 and 123 in FIG. The relative tilt amount and tilt direction between the average reference focus surface and the best focus surface of the AF system are calculated. As described above, the relative tilt obtained here is a relative value with respect to the best focus plane, so the fact that the relative tilt becomes zero means that the measurement points of the multipoint AF system are different. The virtual plane (that is, the reference focusing plane) formed by connecting the reference focusing positions (Z position where the zero point of the detection signal is obtained) at is the best with almost constant offset amount at any point in the projection visual field. It means that it became parallel to the focus plane. However, since it is practically difficult to readjust the device so that the offset amount ΔFSi at each measurement point is substantially constant, when the average surface determined by the n offset amounts ΔFSi is inclined, When operating the multi-point AF system, it is preferable to tilt the wafer W as a whole so as to match the tilt, that is, deal with it at the wafer global level.

For that purpose, for example, when the offset characteristic is as shown in FIG. 15A, the wafer holder is preliminarily inclined by the leveling mechanism in the same amount as the inclination (broken line) of the characteristic. .. With this arrangement, in the case of the offset characteristic as shown in FIG. 15A, when the wafer W is positioned in the Z direction so that the offset amount is ΔFS ₃ at the central measurement point, the offset amounts at other measurement points are set. Is the previously stored values ΔFS ₁ , ΔFS ₂ , ΔF
It is close to S ₄ and ΔFS ₅ , respectively. Therefore, when leveling the wafer holder immediately (or at the same time), the leveling drive amount is extremely small, so the throughput when exposing the reticle pattern on the wafer by the step-and-repeat method or the step-and-scan method is improved. There are uses such as doing.

Further, even in the case of the offset characteristic accompanied by the curvature as shown in FIG. 15B, it is preferable to find the approximate plane and tilt the wafer holder in advance by the same amount as the inclination (Q) of the approximate plane. With this configuration, when the shot area on the wafer is detected, the offset amount at each measured value is close to the offset amount ΔFJi stored in advance, and the drive amount of the leveling mechanism is extremely small. Will be enough.

As described above, when the entire wafer holder is pre-leveled based on the inclination of the offset characteristic obtained at the time of calibration, each measurement of the multipoint AF system at the time of exposure for each shot area on the wafer W is performed. The amount to be leveled based on the detected value at the point can be extremely small, and the throughput can be improved. As a second operation method, after performing the leveling correction for each shot area, the best focus plane within the shot area is set based on the focus detection signal (detection output) only at a specific measurement point in the shot area. How to match local parts,
There is a so-called pinpoint AF method. FIG. 17 shows a wafer W
An example of the cross-sectional structure of the surface of one shot area above is shown, and the arrows represent five measurement points MPa to MPe of the multipoint AF system. Generally, relatively large unevenness is generated in the shot area on the wafer where the process is advanced. In FIG. 17, regions AS ₁ and AS ₃ are convex regions,
Reference numeral ₂ is a recessed area. Here, assuming that the best focus surface is focused on the area AS ₂ for exposure, after leveling correction, focusing (Z stage fine movement) is performed based only on the detection output signal FS ₃ obtained at the measurement point MPc. To do. When the offset characteristic of the multipoint AF system is as shown in FIG. 15 (B), the surface of the shot area is parallel to the line Q in FIG. 15 (B) due to the leveling correction. In the previous embodiment, focusing was performed by the offset amounts ΔFJ _{1 to} ΔFJ ₅ representing the straight line Q, but here, the Z stage is finely moved so that the detection output value at the measurement point MPc becomes the offset amount ΔFS _3. Become. When the harving 12 is corrected by an amount corresponding to the offset amount ΔFJ ₃ calculated at the time of calibration of the measurement point MPc, the detection output value at the measurement point MPc should be equal to (ΔFS ₃ −ΔFJ ₃ ). To Z
The stage 20 may be slightly moved.

In this way, the area AS _{2 in} FIG.
Therefore, the best focus can be obtained, and the resolution of the selected local portion in the shot area can be maximized. By the way, in each of the above examples, the pair of light emission marks ISx, ISy, etc. were sequentially moved to the vicinity of each measurement point of the multipoint AF system, and then each measurement point was calibrated. As described above, when the large reference plate FM that substantially covers the projection visual field If of the projection lens PL can be arranged on the Z stage 20, the reference plate F
On the M, n sets of light emitting marks may be provided corresponding to the positions of the respective measurement points of the multipoint AF system. However, in this case, the light emitted from the projection lens PL and returning to each slit defining the n sets of light emission marks must be guided to each of the n photoelectric sensors by an optical fiber or the like provided individually for each set. There is.

As described above, when a plurality of sets of light emitting marks are provided on the reference plate FM in correspondence with the positions of the respective measurement points, the calibration operation is extremely simplified, and the reference plate FM.
The up and down movement of can be done only once. Instead, A shown in FIG.
It is necessary to prepare n sets of DC 305, RAM 306, counter 307, and the like. The small circle shown by the broken line in FIG. 18 represents the area where the slits of the emission marks ISx and ISy are illuminated from the back side. Further, when the light emitting mark groups are arranged at a plurality of locations (9 locations in FIG. 18) on one large reference plate FM in this way,
It seems that the degree of orthogonality between the surface of the reference plate FM and the optical axis AX of the projection lens PL influences the mounting error of the reference plate FM.
Even if the surface of M and the best focus surface are relatively inclined, it can be said that there is no substantial effect.

[0064]

As described above, according to the present invention, by the second focus detecting means, it is the same as or near the exposure light through the device mask used for the actual exposure and the projection optical system. Measurement of the first focus detection system for detecting the height position of the surface of the photosensitive substrate, since the absolute image plane position can be detected at arbitrary plural points within the projection visual field by using light of wavelength Focus calibration can be performed at this point, and improvement in calibration accuracy can be expected. When the focus detection system is a multi-point AF system having multiple measurement points, high-precision calibration can be performed at each measurement point, so there is a step structure on the photosensitive substrate, and focus is applied to a specific step part among them. Even if you want to perform the measurement, select a measurement point from multiple measurement points according to the position of the specific step, and use the detection signal obtained from that measurement point to accurately mask only the specific step. The projected image of the pattern can be focused. Further, according to the second focus detection system, it is possible to specify the best image forming plane at a plurality of points in the projection field of view, so that the curvature of field, the inclination of the image surface, etc. of the projection optical system can be accurately obtained. Become. Therefore, the above-mentioned field curvature and image plane tilt are caused by vertical movement of the mask, leveling (tilt) of the mask, movement of a part of the optical element in the projection optical system, leveling of a holder (stage) holding the photosensitive substrate, and the like. Will also be able to support.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a projection exposure apparatus having a second focus detection system according to an embodiment of the present invention.

FIG. 2 is a plan view showing a pattern arrangement on a reference plate FM in FIG.

FIG. 3 is a diagram showing a state of a light emitting mark and a reflection image superimposed on it.

FIG. 4 is a diagram showing a waveform of a photoelectric signal obtained by a second focus detection system.

FIG. 5 is a diagram showing a configuration of a first focus detection system suitable for an embodiment of the present invention.

FIG. 6 is a diagram showing an arrangement relationship between a projection visual field and a slit projection image by the first focus detection system.

FIG. 7 is a perspective view showing a configuration of a light receiving system of a first focus detection system.

FIG. 8 is a block diagram showing a configuration of a signal processing circuit of a first focus detection system.

FIG. 9 is a block diagram showing a configuration of a signal processing circuit of a second focus detection system.

FIG. 10 is a waveform diagram showing how the detection signal from the first focus detection system and the detection signal from the second focus detection system are processed.

FIG. 11 is a diagram showing an arrangement example of measurement points in a shot area on a wafer by the first focus detection system.

FIG. 12 is a graph schematically showing the tendency of the offset amount at each measurement point.

FIG. 13 is a flowchart showing a sequence of measurement operation when calibrating each measurement point of the multipoint AF system.

FIG. 14 is a flowchart showing a sequence of calculation processing for calibration of a multipoint AF system.

FIG. 15 is a graph showing a tendency of an offset amount at each measurement point assumed during calibration.

FIG. 16 is a diagram showing a modification of the arrangement of measurement points of a multipoint AF system.

FIG. 17 is a diagram showing a cross-sectional structure of the surface of a shot area on a wafer.

FIG. 18 is a diagram showing a modification of the arrangement of the light emitting marks on the reference plate.

[Explanation of symbols for main parts]

 R Reticle RL Projection lens W Wafer FM Reference plate 20 Z stage 21 XY stage 30E Calibration value determination unit 30D Storage unit 41 Optical fiber 45 Photoelectric sensor

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] February 26, 1992

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0012

[Correction method] Change

[Correction content]

[0012]

According to the present invention, a part of a Z stage which holds a photosensitive substrate and is movable in the optical axis direction of a projection optical system is provided with a predetermined shape in a plane parallel to a projection image plane. A reference plate (FM) having a light emitting mark is provided, and further, of the light that reaches the reticle from the light emitting mark through the projection optical system, it is reflected by the reticle and passes through the projection optical system to become the reference plate. Second focus detecting means (TT) for detecting the positional deviation of the returning plate in the optical axis direction with respect to the best projection image forming plane by photoelectrically detecting the returning light by moving the Z stage.
L type focus detection system) is provided. Then, the means for designating the position of the XY stage so that the light emitting mark of the reference plate is positioned in the vicinity of the measurement point of the first focus detection system of the oblique incident light type, and the first focus detection means. A calibration means is provided to calibrate the error between the position detected as ## EQU1 ## and the absolute best image plane on average or singly.

In the second focus detecting means according to the present invention, no matter where the light emitting mark is arranged in the visual field of the projection optical system,
The reflected light from the reticle (mask) can be received to accurately detect the best image plane at that position. Therefore, compared with the conventional method of detecting the best image plane only at a specific position within the projection visual field, the degree of freedom is dramatically increased. Therefore, when a multi-point AF system having multiple measurement points in the projection field of the projection optical system is used as the first focus detection means, calibration is performed using the light emission mark at each measurement point with the best imaging plane as a reference ( Calibration) is possible. When such a calibration is performed, if the offset (error) during calibration is corrected for the position deviation amount obtained from each measurement point of the multipoint AF system, the surface of the photosensitive substrate will always be the best image plane. Focusing that matches

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a TT for detecting the best focus plane of a projection optical system in a projection exposure apparatus according to an embodiment of the present invention.
It is a figure which shows the focus detection system of L type. In FIG. 1, a reticle R having a circuit pattern area PA for manufacturing an actual device formed on its lower surface is held by a reticle holder (not shown). The optical axis AX of the projection lens PL, which is schematically represented by dividing the diaphragm surface (pupil surface) EP into a front group and a rear group, has the center of the reticle R, that is, the center of the pattern area PA, with respect to the reticle pattern surface. Pass vertically. Below the projection lens PL, the wafer stage W is placed and the Z stage 20 is moved by a small amount (for example, within ± 100 μm) in the optical axis AX direction.
Are provided on the XY stage 21. The XY stage 21 moves the wafer W two-dimensionally in an XY plane perpendicular to the optical axis AX. Furthermore, Z stage 20
A reference plate FM is fixed to a part of the reference plate FM at a height position substantially equal to the surface of the wafer W. This reference plate FM is shown in FIG.
As shown in FIG. 7, a mark ISy in which a plurality of transmission slits extending in the X direction are arranged at a constant pitch in the Y direction, and a plurality of transmission slits extending in the Y direction are arranged in a constant pitch in the X direction. A mark ISx and an oblique slit ISa which is 45 ° with respect to each of the X and Y directions are formed. These slit marks ISx, ISy, IS
In a, a chromium layer (light-shielding layer) is vapor-deposited on the entire surface of the quartz reference plate FM, and is engraved as a transparent portion there.

Claims (1)

[Claims] 1. A projection optical system for image-forming and projecting a pattern of a mask on a predetermined image forming plane, and a projection optical system which holds a photosensitive substrate substantially parallel to the image forming plane and is arranged in a plane parallel to the image forming plane. An XY stage that moves dimensionally, a Z stage that moves the photosensitive substrate in the optical axis direction of the projection optical system, and measurement points at a plurality of predetermined positions within the projection field of the projection optical system,
A projection exposure apparatus comprising: first focus detection means for detecting a positional deviation of the surface of the photosensitive substrate in the optical axis direction at each of the plurality of measurement points; A reference plate formed with a light emitting portion that emits light with a predetermined shape in a plane parallel to the image plane; and when the light emitting portion is located at an arbitrary point in the projection field of view, the light emitting portion causes the projection optical system to move. Of the light reaching the mask through the light reflected by the mask and returning to the reference plate through the projection optical system,
Second focus detecting means for detecting a positional relationship in the optical axis direction of the reference plate with the image forming surface as a reference by moving the stage and photoelectrically detecting the plurality; Designating means for designating the coordinate position of the XY stage so that the light emitting section is sequentially positioned at the position of the measurement point or at a position very close thereto; and the second at each of the designated coordinate positions. Calibration means for calibrating the detection error at each of the plurality of measurement points by the first focus detection means individually or on average based on each positional relationship detected by operating the focus detection means. A projection exposure apparatus characterized in that