JPH0552660B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field of the Invention) The present invention relates to a device for detecting the position of a pattern (mark) provided on a substrate to be aligned, and in particular to detecting a pattern position suitable for a wafer alignment device in a semiconductor manufacturing device. This invention relates to a detection device.

(Background of the Invention) In semiconductor manufacturing equipment, an exposure device (aligner or stepper) that transfers a pattern on a mask (reticle) onto a semiconductor wafer (hereinafter simply referred to as wafer) uses an exposure device (aligner or stepper) that accurately aligns the mask and wafer immediately before exposure. alignment is required. Generally, for this type of alignment, the wafer is provided with a wafer mark for minute alignment, and the mask is also provided with a mask mark that matches the wafer mark. The exposure device is configured to observe the mask mark and wafer mark from the mask side using an alignment microscope, photoelectrically detect the images of both marks, and automatically correct the relative positional deviation. It's summery. In this configuration, the wafer mark in particular is a single straight pattern, and the edges on both sides of this mark are photoelectrically scanned in a direction across, and the midpoint between the edges on both sides is detected using an image signal corresponding to the mark. After that, the midpoint is set as the mark position.

As an example, a method is known in which symmetrical pattern matching (correlation calculation) is performed using image signals, as disclosed in Japanese Patent Publication No. 56-2284. This is based on the principle that matching is best (left-right symmetry) when the image signal is folded back at the midpoint between the two edges of the mark. Normally, a photosensitive agent (photoresist) is coated on the wafer surface to a thickness of about 0.5 to 2 μm for exposure, but unevenness occurs in the thickness of the resist near the marks formed as minute irregularities. For this reason, when observing a mark with a microscope, when a mark is illuminated with monochromatic light, a plurality of interference fringes are generated near the edge of the mark and approximately parallel to the edge. If the interference fringes occur symmetrically on both sides of the mark, the position of the mark can be detected with sufficient accuracy by the conventional symmetrical pattern matching method. However, when applying resist to the wafer surface using a spin coater, etc., the thickness of the resist may not change symmetrically at both edges of the mark (steps of unevenness) due to centrifugal force, resulting in asymmetrical interference fringes. may occur. As a result, the left-right symmetry of the entire image signal deteriorates, and processing that relies only on symmetry has the disadvantage that mark position detection accuracy decreases.

(Objective of the Invention) An object of the present invention is to solve the above-mentioned drawbacks and to obtain a pattern position detection device that can detect the position of a mark etc. with high precision even if the left and right symmetry of a pattern signal such as an image signal is poor. .

(Summary of the invention) As shown in FIG. pattern detection means 200 that generates a pattern signal V′), an extraction means 201 that extracts all scanning positions at which the waveform of the pattern signal has a predetermined shape according to the edges of the pattern from a predetermined scanning range including the pattern; Create all combinations of any two scanning positions from among the extracted plurality of scanning positions, and determine to what extent the pattern signal satisfies predetermined waveform conditions at least between the two combined scanning positions. and a combination detection means 202 that detects one combination that most satisfies the waveform condition, and a predetermined position that divides the two scanning positions in that one combination into two at a predetermined ratio (for example, 1:1). The technical point is to provide position detection means 203, 203a for detecting the position of the pattern.

Furthermore, in the embodiment of the present invention, the pattern detection means includes an image detector (image pickup tube) 21 that outputs an image signal V having a characteristic waveform such as a bottom or a slope at the edge position of the pattern, and a differential signal
A differentiating means 200a that outputs V' is provided, and the extracting means 201 includes a first extracting means 201a and a second extracting means 201b that extract scanning positions according to characteristic waveforms for each of the image signal V and the differential signal V'. is provided, and the combination detection means 202 includes a first extraction means 2
01a and the second extraction means 201b, a first combination detection means 202a and a second combination detection means 202 detect one combination (P), (P').
b, and the fixed point position detection means combines (P), (P) based on the magnitude relationship between the characteristic waveform of the image signal V detected by the extraction means 201 and the characteristic waveform of the differential signal V The key point of the technical point is to provide a selection means 203b for selecting either one of the above.

In another embodiment of the present invention, the combination detection means 202 includes a storage means (for example, a random access memory) 202c that stores standard waveforms of pattern signals as templates in advance, and when determining the waveform conditions described above, The waveform of the pattern signal and the template are superimposed and correlated so that the center point of the two scanned positions coincides with the center point of the two edge positions of the template, and one combination with the best correlation is detected. Correlation detection means 20
2d, and a pattern width determining means 202e for selecting a combination in which the interval between the two combined scanning positions is approximately equal to the interval between the two edges of the pattern during correlation calculation.

(Embodiment) FIG. 2 is a schematic optical layout diagram of a reduction projection type exposure apparatus suitable for an embodiment of the present invention, and FIG. 3 is a circuit block diagram of a control system for controlling the exposure apparatus.

In FIG. 2, an exposure light beam 1 is supplied from a light source section (not shown) and illuminates a reticle (mask) 3 as a projection original through a main condenser lens 2. The main condenser lens 2 forms an image of the light source within the entrance pupil plane 6 of the projection lens 5. The surface of the reticle 3 on the projection lens 5 side,
That is, a predetermined pattern is formed on the pattern surface 3a, and this pattern is formed on the projection lens 5.
The image is reduced and projected onto the pattern surface 7a of the wafer 7 at a predetermined magnification. The exposure light flux 1 is effective light for exposing the photoresist coated on the pattern surface 7a of the wafer 7, and includes, for example, light with a wavelength of 435.8 nm (g-line) generated from an ultra-high pressure mercury lamp, or light with a wavelength of 365 nm.
Single wavelength light such as (i-line) light is used.

For such an exposure optical system, there is an alignment optical system for observing the positional relationship between the alignment mark A provided on the wafer 7 and the mark B provided on the periphery of the pattern surface 3b of the reticle. is as follows, and light having the same wavelength as the exposure light beam 1 is used as illumination light. At the upper part of the periphery of the reticle 3, a reflecting mirror 14, a first objective lens 15,
A second objective lens 16 is provided. The optical axis 13 of the first objective lens 15 is bent at a right angle by the reflecting mirror 14 and is perpendicular to the pattern surface 3a of the reticle 3. Therefore, it is substantially parallel to the optical axis 9 of the projection lens 5, and the front focal point of the first objective lens 15 is located on the pattern surface 3a of the reticle 3. The first objective lens 15 and the reflecting mirror 14 can move together along the optical axis 13 while maintaining this positional relationship. Furthermore, the first objective lens 15 and the second objective lens 16 are always in a parallel system, and the aperture 18 provided at the rear focal position of the second objective lens 16 and the pattern surface 3a of the reticle 3 are It is conjugate with respect to the first and second objective lenses, and is also conjugate with the pattern surface 7a of the wafer 7 with respect to the projection lens 5. A diaphragm 1 provided to limit the illumination area on the wafer 7
A condenser lens 19 is provided behind the aperture 8, and its focal point matches the position of the aperture 18.
The condenser lens 19 is arranged coaxially with the optical axis 13 of the first and second objective lenses. and,
The exit surface 20a of the light guide 20 for supplying illumination light having the same wavelength as the exposure light beam 1 is provided with its center offset from the optical axis 13. On the other hand, aperture 1
A beam splitter 17 is arranged between the second objective lens 16 and the second objective lens 16 . An imaging device 21 such as a television camera is provided in the optical path divided by the beam splitter 17, and the imaging surface 21a of the imaging device 21 is a pattern surface with respect to the first and second objective lenses 15 and 16, as well as the aperture 18. 3a, and also the pattern surface 7a of the wafer 7.

With this configuration, the light beam from the light guide 20 is supplied to the mark A provided on the pattern surface 7a of the wafer 7 and the mark B provided on the pattern surface 3a of the reticle 3, and the marks A and B are
Both images are observed by the imaging device 21.

Reflector 14 and first objective lens 1 as described above
5, second objective lens 16, beam splitter 1
7. An alignment optical system composed of an aperture 18, a condenser lens 19, a light guide 20, and an imaging device 21 is provided with another set in the direction perpendicular to the plane of the paper in FIG. The system detects two-dimensional positional deviation between the reticle 3 and the wafer 7.

By the way, in the configuration shown in Figure 2, reticle 3
A non-telecentric optical system exists between the projection lens 5 and the projection lens 5, and a telecentric optical system exists between the projection lens 5 and the wafer 7. In this case, if the mark on the reticle 3 is placed at a position where the optical axis 9 of the projection lens 5 does not pass, for example around the reticle 3, the principal ray 11 of the illumination light beam (or alignment optical system) of the mark will be aligned with the mark. Projection lens 5
The principal ray 1 coincides with the line connecting the center of the pupil 6 of the projection lens and is inclined with respect to the optical axis 9 of the projection lens.
1 is not perpendicular to the putter surface 3a of the reticle 3,
It will be tilted at a certain angle. Therefore, even if the mark B on the reticle 3 is formed of a light-reflective metal thin film, the reflected light from the mark does not return to the alignment optical system or the imaging device 21, but instead travels in another direction. From, (due to the angular characteristics of the illumination light)
The mark is transmitted and illuminated by the light reflected by the wafer 7, and the light image of the mark B is captured as a dark shadow. In this way, reticle 3
If the side is a non-telecentric optical system,
The mark B formed on the reticle 3 may be reflective and light-shielding, or light-absorbing.

In addition, if the optical system is telecentric on both the reticle 3 side and the wafer 7 side, the chief ray 11 is perpendicular to the reticle 3 (parallel to the optical axis 9 of the projection lens 5), so the mark on the reticle 3 becomes light absorbing. Alternatively, you can illuminate the reticle 3 with polarized light and insert a member (for example, a 1/4 wavelength plate) that changes the polarization state in the optical path between the reticle 3 and the wafer 7 to It is sufficient to provide an optical system having polarization separation characteristics that cuts reflected light and transmits reflected light from the wafer 7.

Now, in FIG. 2, the wafer 7 is vacuum-adsorbed onto the wafer holder 8. The stage 10 is movable two-dimensionally (in the x and y directions perpendicular to each other) on the base, and the wafer holder 8 is movable on the base.
0 and move together in two dimensions. Further, the wafer holder 8 is movable in the direction of the optical axis 9 of the projection lens 5 (Z direction) with respect to the stage 10 for focus adjustment. The reference mark plate 12 is provided with a reference mark that can be aligned with the mark B of the reticle 3, and is used to define the position of the reticle 3 and the stage 10. This reference mark plate 12
Surfaces other than the mark above have a predetermined light reflectance,
It is fixed to the wafer holder 8.

Now, in the circuit block diagram of Fig. 3, the second
The brightness of the image of mark A or B formed by the objective lens 16 is photoelectrically converted by the imaging device 21, and the image signal is input to the signal processing circuit 30. The signal processing circuit 30 processes the image signal, detects the contrast of the image formed on the imaging surface 21a, and outputs detection information corresponding to the amount of relative positional deviation between the marks A and B to the control circuit 31. Further, a motor 35 driven by a drive circuit 34 is connected to a wafer holder 8.
is moved in the Z direction, and the amount of movement is controlled by drive information output from the control circuit 31. Further, motors 37 and 38 driven by a drive circuit 36 move the stage 10 in the x direction and y direction, respectively, and the amount of movement is controlled by drive information output based on detection information from the control circuit 31. be done. The two-dimensional position of the stage 10, that is, the coordinate position of the orthogonal coordinate system xy, is determined by a coordinate measuring device 39 using a laser interferometer or an encoder.
The detected coordinate positions are processed by the control circuit 31 and used for positioning the stage 10.

Now, FIG. 4 is a circuit block diagram showing an example of a specific circuit of the signal processing circuit 30. The magnitude (level) of the image signal from the imaging device 21 is determined by an analog-to-digital converter (hereinafter referred to as ADC) 40.
The digital value is converted into a digital value by a randomly accessible memory circuit (hereinafter referred to as RAM).
) 41. Also, from the imaging device 21, a synchronization signal (horizontal periodic signal) for the horizontal scanning line is sent.
HS is output, and the counter 42 generates 1024 pulses during one scanning period based on this horizontal periodic signal HS so as to divide one scanning line into, for example, 1024 pixels, and sequentially counts the pulses. The counted value is applied as an address signal to the RAM 41, and as a result, the luminance level of each pixel of the image signal corresponding to one scanning line is stored in the RAM 41 in address order. Microprocessor (hereinafter referred to as MPU)
43 controls writing and reading of the RAM 41, and calculates the amount of positional deviation between the mark B and the mark A based on the image data stored in the RAM 41. Also, from the imaging device 21, 2
For dimensional scanning, a sawtooth waveform vertical scanning signal VS is also output, and is applied to a television cathode ray tube (hereinafter referred to as CRT) 44 together with the aforementioned image signal and horizontal periodic signal HS.

The CRT 44 displays the captured image and also displays a horizontal cursor line at a predetermined position in the vertical direction of the screen. The cursor line is a cursor signal generator 45 that generates a pulse-like cursor signal CS when the set voltage Vr corresponding to the vertical position of the cursor sensor matches the voltage of the vertical scanning signal VS.
and an adder 46 that adds the cursor signal CS to the image signal. The cursor signal CS is also input to the counter 42, and for one horizontal scanning line that coincides with the cursor line, the counter 4
Control is performed so that a count of 2 is performed. For this reason
The RAM 41 stores only the image signal at the display position of the cursor line.

Figure 5 shows the reticle 3 used when overlapping exposure with a predetermined pattern (chip) on the wafer 7.
FIG. The reticle 3 is a rectangular glass plate with a thin metal film such as chromium deposited on the periphery to form a light shielding part 3a, and a desired circuit pattern is formed in an area 3b inside the light shielding part 3a. When a coordinate system xy is defined with the center RC of the reticle 3 as the origin, a mark B for positioning in the x direction is provided as a minute rectangular opening at the outer periphery of the area 3b on the x axis. Note that a mark for alignment in the y direction is similarly provided on the y axis. This mark B indicates the reflecting mirror 14 and the first objective lens 1 shown in FIG.
5. Observed by the second objective lens 16 and the imaging device 21. This reticle 3 is the central RC
is positioned so that it passes through the optical axis 9 of the projection lens 5.

The position of the mark B in the x direction with respect to the center RC is known in advance when the reticle 3 is designed. Therefore, the reticle 3 is placed as shown in Figure 1, and the reflector 1 is
4 and the first objective lens 15 in the direction of the optical axis 13, the mark B is observed by the imaging device 21. Note that the alignment mark in the x direction is similarly observed by another set of alignment optical systems.

Next, the operation of this apparatus will be explained. The wafer 7 is pre-aligned and placed on the stage 10, and the chips on the wafer 7 and the projected image of the pattern area 3b of the reticle 3 are roughly aligned. The state up to precise alignment will be described. Illumination light is emitted from the light guide 20 to illuminate the marks B and A with the chip and the projected image of the pattern area 3b substantially aligned. As a result, the light-receiving surface 21a of the imaging device 21 has a sixth
As shown in the figure, both images of the rectangular aperture mark B and the linear mark A on the wafer 7 are formed within the scanning area SA. Within the scanning area SA, both images of marks A and B are raster scanned. At this time, the horizontal scanning line is mark A.
The edges E ₁ and E ₂ on both sides of the mark B and the edges R ₁ and R ₂ on both sides of the mark B are almost perpendicular to each other. Further, on the outside of the edges E ₁ and E ₂ of the mark A, interference fringes 50 (bright and dark stripes) due to the influence of the register appear parallel to the edges E ₁ and E ₂ . These interference fringes 50 are caused by uneven thickness of the resist. For example, when mark A is formed in a convex shape as shown in FIG.
This is because the thickness d of the resistor FR changes continuously near the edges E ₁ and E ₂ (steps) of the mark A.
Now, in the state shown in FIG. 6, the RAM 41 of the signal processing circuit 30 in FIG. 4 samples and stores image signals for one scanning line for each pixel at the position where the cursor line CL is displayed. The stored image signal V has a waveform (represented by an envelope) with many peaks and bottoms (dips) depending on the mark A and the interference fringes 50, as shown in FIG. 8, for example. In Figure 8, the vertical axis represents the intensity of the image signal V, and the horizontal axis represents the intensity of the image signal V.
The address of the RAM 41 or the direction on the reticle 3 or wafer 7 of the horizontal scanning line, for example x
Represents the position of the direction. The image signal V also includes a plate shape corresponding to the edges R ₁ and R ₂ of the mark B on the reticle 3, but for the sake of simplicity in FIG.
Only the waveform near mark A from position x ₀ to x ₅ is shown. Here, the bottoms B ₁ , B ₂ , B ₃ ,
It is assumed that among B ₄ , the bottoms corresponding to the edges E ₁ and E ₂ of mark A are B ₂ and B ₃ , and the bottoms B ₁ and B ₄ are formed by the interference fringes 50 .

Next, the MPU 43 reads out the image signal V stored in the RAM 41 as shown in _FIG .
Differentiate the image signal V with respect to x ₅ and obtain the differential signal
By associating V′ with the position in the x direction and the address,
It is stored in an empty area of RAM41. This differential is
After roughly calculating the midpoint at which the waveform of the image signal V becomes symmetrical, a numerical filter is applied to the left and right from this midpoint (from the address corresponding to the midpoint of the RAM 41, in the direction in which the address increases and in the direction in which the address decreases). It is done using. The differential signal is
As shown in Fig. 9, V' becomes zero at characteristic waveform parts (peaks and bottoms) in the image signal V, resulting in a waveform that includes many peaks U ₁ to _{U 4} and bottoms D ₁ to _{D 4} . . In FIG. 9, the vertical axis represents the intensity I' of the differential signal V', and the horizontal axis represents the position (or address) in the x direction.Similar to the image signal V, the differential signal V' depends on the pixel position on the scanning line. Although it is a discrete data group, it is shown here as an envelope.

Next, the MPU 43 determines the position of the mark A, that is, the center point of the edges E ₁ and E ₂ in the x direction, based on the image signal V and the differential signal V' stored in the RAM 41, according to the flowchart of FIG. To detect.
This flowchart schematically represents the functions of the MPU 43. Below, steps 100 to 108 for achieving this function will be explained.

Step 100 First, the MPU 43 scans the scanning range x ₀ ~
x ₅ image signal V is read and the positions x ₁ to x ₄ of the bottoms B ₁ to _{B 4} are detected, and the sizes h ₁ , h ₂ , h ₃ , and h ₄ of the bottoms are also detected. To detect the bottom position, for example, find the position where the magnitude I' of the differential signal V' becomes zero, and determine whether the waveform of the image signal V corresponding to that zero position is a dip. . The size of _the bottom is as shown in FIG. 11. For example, after bottom B ₁ is detected at position x ₁ , the size of bottom B ₁ is set as I ₁ , and the image signal V is Check the size I. And at the size I ₂ ,
Since the dip is finished, MPU43 is the size.
It is obtained by storing the difference between I ₂ and I ₁ as the size h ₁ of the bottom B ₁ . As above,
The MPU 43 is located at each position x ₁ of the bottom B ₁ , B ₂ , B ₃ , B ₄ ,
x ₂ , x ₃ , x ₄ and bottom sizes h ₁ , h ₂ , h ₃ , h ₄ are detected and stored. In addition, the position x ₁ ~ _{x 4}
are stored as addresses in the RAM 41, respectively.

Step 101 Next, the MPU 43 reads the differential signal V' corresponding to the scanning range x ₀ to x ₅ and detects the position and magnitude of the slope (inclined portion) on the differential waveform. In this embodiment, the peak and bottom positions on the differential waveform are detected as the slope position, and the bottom size is detected as the slope magnitude. Of course, the zero point position on the differential waveform may be detected as the slope position, and the difference between the peak value and bottom value of one slope may be detected as the slope magnitude. Now, the MPU 43 detects four peak positions (addresses) from the differential signal V' shown in Figure 8.
U ₁ , U ₂ , U ₃ , U ₄ and four bottom positions (address)
D ₁ , D ₂ , D ₃ , and D ₄ are detected, and the sizes of the bottoms l ₁ , l ₂ , l ₃ , and l ₄ are also detected and stored. The reason why the bottom size on the differential waveform is detected as the slope size is because the program can be made more compact by using the bottom size detection program used in step 100 as is. be.

Step 102 (first extraction means 201a) Next, the MPU 43 detects the bottom of the waveform of the image signal V.
Create all combinations (pairs) of any two bottoms from _B1 to _B4 .

Here, (B ₁ , B ₂ ), (B ₁ , B ₃ ), (B ₁ , B ₄ ),
There are six pairs: (B ₂ , B ₃ ), (B ₂ , B ₄ ), and (B ₃ , B ₄ ).

Step 103 (second extraction means 201b) Next, the MPU 43 determines the slope positions on the waveform of the differential signal V', that is, the peaks U ₁ to _{U 4} and the bottom D ₁ to _{D 4}
Create all combinations (pairs) of any two peaks or bottoms for .

Here, regarding the peaks (U ₁ , U ₂ ), (U ₁ ,
U ₃ ), (U ₁ , U ₄ ), (U ₂ , U ₃ ), (U ₂ , U ₄ ), (U ₃ ,
Regarding the 6 pairs of U ₄ ) and the bottom (D ₁ , D ₂ ),
(D ₁ , D ₃ ), (D ₁ , D ₄ ), (D ₂ , D ₃ ), (D ₂ , D ₄ ),
There will be 6 pairs (D ₃ , D ₄ ).

Step 104 (first combination detection means 202a) Next, the MPU 43 detects the 6 images extracted from the image signal V.
One pair (P) that most satisfies a predetermined criterion is detected from among the two pairs. In this embodiment, it is determined whether the interval between the bottom positions of the pair is approximate to the width of the mark A, and the symmetry of folding and overlapping the image signals V is determined.

First, for the pair (B ₁ , B ₂ ), check whether the distance between the positions x ₁ and x ₂ of the bottom B ₁ and B ₂ is approximate to the width of mark A, and if it is not approximate, move to the next pair ( B ₁ , B ₃ ) are investigated in the same way. Pair B ₁ , B ₃
If it is determined that the mark width is similar to the mark width, the symmetry by folding is determined. Here, the center position between the position x ₁ of the bottom B ₁ and the position x ₃ of the bottom B ₃ is determined, the image signals V are folded back from the center position and superimposed, and a correlation value is determined by a correlation calculation. The above operation is performed in the same way for the six pairs, and here the pairs (B ₂ , B ₃ ) and (B ₂ ,
B ₄ ) is selected to determine the mark width, and the correlation value is determined by folding back each mark. And MPU43 has three pairs (B ₁ , B ₃ ), (B ₂ ,
Compare the correlation values for B ₃ ) and (B ₂ , B ₄ ),
Detect the most correlated pair, for example the pair (B ₂ , B ₃ ). Note that when performing pattern matching by folding, it is not necessary to fold all of the image signal V. For example, when calculating the correlation value for a pair (B ₁ , B ₃ ), the waveform from position x ₁ to x ₃ of the image signal V It may also be possible to perform wrapping only for.

Step 105 (Second combination detection means 202b) Next, the MPU 43 performs the above-mentioned mark width judgment and symmetry judgment from each of the six bottom pairs and six peak pairs extracted from the differential signal V'. , one pair (P′) is detected. Again, the mark width condition is satisfied for the pair (U ₁ , U ₃ , U ₂ , U ₃ , U ₂ , U ₄ ), for example, the pair (U ₁ , U ₃ )
, the correlation value is obtained by pattern matching by folding back the differential signal V' from the center point between the position of the peak U ₁ and the position of the peak U ₃ . and
MPU43 is a pair (U ₁ , U ₃ , U ₂ , U ₃ , U ₂ , U ₄ )
The pair with the highest correlation (good symmetry) among
For example, detect the pair (U ₂ , U ₃ ). Similarly, from the bottom six pairs, for example, the pairs (D ₁ , D ₃ ,
D ₂ , D ₃ , D ₂ , D ₄ ) are selected by mark width determination, and similarly, the pair (D ₂ , D ₃ ) with the highest correlation is detected by folding symmetry determination.

Step 106 (selection means 203b) Next, the MPU 43 selects the pair ((P), P') that has a larger bottom or slope. In the previous step 101, the differential signal is calculated as the magnitude of the slope.
Since we have detected the respective sizes l ₁ to _{l 4} of the bottoms D ₁ to D ₄ of V′, we will use the pairs (P) and (P′), that is, the pair (B ₂ ,
B ₃ ) and pair (D ₂ , D ₃ ), bottom B ₂ , B ₃
Compare the sizes h ₂ , h ₃ of the bottom D ₂ , D ₃ with the sizes l ₂ , l ₃ of the larger pair, e.g. (D ₂ , D ₃ )
Select.

Step 107 Now, when the pair (D ₂ , D ₃ ) is selected, the MPU
Step 43 checks whether the mark width is extremely close to the width of the designed mark A, which is stricter than the previous mark width determination. The pair (D ₂ , D ₃ ) selected up to this point corresponds to the two edges E ₁ and E ₂ of mark A with an extremely high probability.

Step 108 (fixed point position detection means 203a) Next, the MPU 43 calculates the center position between the bottom D ₂ position and the bottom D ₃ position of the pair (D ₂ , D ₃ ),
The position of the center line of edges _E1 and _E2 of mark A in the scanning line (cursor line) direction is detected. Here, it is sufficient to know at what pixel (pixel position) the center of mark A is located from the scanning start pixel of the scanning line. Also, from the image signal V, reticle 3
Find the pixel positions of both edges R ₁ and R ₂ of mark B,
The pixel position at the center of edges _R1 and _R2 is also detected. Then, the MPU 43 calculates the number of pixels that is the difference between the center pixel position of mark A and the center pixel position of mark B, and multiplies the pitch of one pixel on the wafer 7 by the number of pixels. The amount of positional deviation in the scanning line direction with respect to B is detected.

Through each of the above steps, the relative positional deviation in one dimension (x direction or y direction) between the wafer 7 and the reticle 3 is detected, and the control circuit 31 shown in FIG. , the stage 10 is positioned in one dimension by driving the motor 37 (or 38) so that the positional deviation is eliminated. Of course, the signal processing circuit 30 in FIG. 4 and the other set of alignment optical systems shown in the flowchart in FIG. is accomplished. In this way, the projected image of the pattern area 3a of the reticle 3 and the chips on the wafer 7 are accurately overlapped, so that the exposure light 1 is irradiated onto the reticle 3 and a desired pattern is transferred onto the wafer 7. When overlapping exposure is performed on the next chip using the step-and-repeat method, the mark A attached to the next chip and the mark B on the reticle 3 may be aligned in exactly the same manner as described above. Of course, it is reliable if the above alignment is performed for all chips on the wafer, but it is better to perform the above alignment for only one or more chips among all the chips, and perform the step-and-repeat step pitch method for the other chips. Alternatively, the positioning may be performed using .

As described above, in steps 102 and 103 of this embodiment, all pairs of arbitrary two positions of bottom and peak are created, but here too, the mark width is determined and only pairs that are close to the mark width are created. You can do it like this. Furthermore, depending on the process (etching, diffusion, etc.) of the wafer 7, the mark A may be deformed asymmetrically. In this case, when determining the center position of edges _E1 and _E2 in step 108, the mark It is preferable to set the mark A at a position shifted to either side from the center.

Next, a second embodiment of the present invention will be described using FIG. 12. In the second embodiment, the previous step 104,
While symmetry was used in the judgment of 105, a method called template matching is used. This template is an image signal V of mark A on wafer 7.
It is only necessary to store the waveform of in advance in memory.
However, it is also necessary to find the center position of the mark A on the waveform. When actually creating a template, it is sufficient to extract and store the image signal V of mark A from a group of multiple wafers having the same pattern on the same base, that is, the first wafer of a lot (ROT). . Normally, all the wafers in a lot are subjected to the same process, so the resist thickness d is uniform among the wafers, and the occurrence of interference fringes does not vary greatly from wafer to wafer. Therefore, after aligning the part of the first wafer of the lot with the correct resist thickness, for example the chip near the center of the wafer, and the pattern area 3 using the stage 10, the mark A attached to that chip is imaged. Then, a characteristic waveform near the mark A of the image signal V as shown in FIG. 8 is stored in the RAM 41 as a template Tp as shown in FIG. Furthermore, the center position (address) Tc of mark A is also detected and stored in accordance with the flowchart shown in FIG.

During actual alignment, in the judgment at step 104 in the flowchart of FIG.
Three pairs (B ₁ , B ₃ , B ₂ ,
After B ₃ , B ₂ , B ₄ ) are selected, template matching is performed on the three pairs. For the pair (B ₁ , B ₃ ), the waveform of the image signal V and the template TP are adjusted so that the center position of the bottom B ₁ position x ₁ and the bottom B ₃ position x ₃ matches the center position Tc of the template TP. The correlation value is calculated by superimposing the Correlation values are similarly calculated for the pairs (B ₂ , B ₃ , B ₂ , B ₄ ). Then, if you select the pair with the highest correlation from those calculated correlation values,
The target pair (B ₂ , B ₃ ) is found. In this embodiment, when creating the template TP, a template is also prepared for the waveform of the differential signal, and one pair (U ₂ , U ₃ and (D ₂ , D ₃ ) is created by template matching in step 105. ) may be detected.

As described above, since a template is used in this embodiment, when detecting mark A for each chip on a wafer, there are slight differences in the state of resist application depending on the position on the wafer, and the image signal obtained for each chip is The waveform of is slightly different from the template. However, since the same templates for one image signal are simply superimposed at different positions (addresses), when one pair is selected from the preselected pairs, the possibility of selecting a wrong pair is extremely small.

As described above, in the first and second embodiments of the present invention, the center positions of the edges E ₁ and E ₂ of the mark A are detected using both the image signal V and the differential signal V'. The center position may be detected using only the signal. In addition, although an image pickup tube such as a television camera is used to obtain the image signal, a slit-shaped spot light (monochromatic light) parallel to the marks A and B is scanned, and the reflected light from the wafer 7 is detected. Even if photoelectric detection is performed with a single light-receiving element or a slit-scanning photoelectron microscope is used, a time-series image signal can be obtained as shown in FIG.

Furthermore, the present invention is applicable not only to optical exposure devices such as aligners and steppers, but also to optical alignment devices such as X-ray exposure devices and other alignment devices that need to detect marks, such as scanning electron microscopes and electron beam exposure devices. It is widely applicable.

In this case, a similar image signal can be obtained by scanning mark A on the wafer with a spot of an electron beam and detecting scattered electrons from the wafer surface.

(Effects of the Invention) According to the present invention, all the positions (scanning positions) of characteristic waveforms that change at the edges of a pattern signal are extracted, and then a predetermined waveform condition (symmetry or template matching) is most satisfied. Since the position of the pattern is determined by selecting only two combinations of two scanning positions, when detecting the alignment pattern (mark) on the semiconductor wafer, there are many interference fringes due to the shape of the resist and its underlying surface. Even if the lines appear asymmetrically, the accuracy of position detection does not decrease and high alignment accuracy can be obtained. Further, by restoring (extracting) the position of the characteristic waveform in advance in this way, there is an advantage that the number of times of arithmetic processing for symmetric pattern matching and template matching is reduced and the processing speed is increased.

[Brief explanation of drawings]

FIG. 1 is a claim correspondence diagram of the present invention, and FIG.
3 is a circuit block diagram of a control system, FIG. 4 is a specific circuit block diagram of a signal processing circuit, and FIG. FIG. 5 is a plan view of the reticle mounted on the exposure apparatus shown in FIG. 2, FIG. 6 is a diagram showing how mark A on the wafer is aligned with mark B on the reticle, and FIG. 7 is a diagram showing mark A on the wafer. 8 is a waveform diagram of image signal V, and FIG. 9 is a differential signal.
FIG. 10 is a flow chart of the first embodiment in which the position of mark A on the wafer is detected by a microprocessor, and FIG. 11 is a waveform diagram illustrating the bottom size of image signal V. Figure, 1st
FIG. 2 is a waveform diagram of a template according to the second embodiment. [Explanation of symbols of main parts], 3...Reticle,
7...Wafer, 21...Imaging device, 30...Signal processing circuit, A...Wafer mark, B...Reticle mark.

Claims (1)

[Scope of Claims] 1. A pattern detection means for scanning an alignment pattern having at least two edges and generating a time-series pattern signal according to the pattern; extracting means for extracting all scanning positions that have a predetermined shape according to the above; creating a combination of any two scanning positions from among the plurality of extracted scanning positions; combination detection means for determining how much the pattern signal satisfies a predetermined waveform condition and detecting one combination that most satisfies the waveform condition; two scanning positions in the detected one combination; A pattern position detection device comprising: position detection means for detecting a predetermined position that divides the gap into two at a predetermined ratio as the position of the pattern. 2. The pattern detection means includes an image detector that outputs a time-series image signal that has a characteristic waveform such as a bottom or slope at the edge position of the pattern, and a differentiation means that outputs a differential signal of the image signal. The extracting means includes a first extracting means for extracting all scanning positions according to a characteristic waveform in the image signal, and a first extraction means for extracting all scanning positions according to a characteristic waveform such as a bottom or a slope in the differential signal. a second extraction means, the combination detection means detects one combination (P) by the determination from the plurality of scanning positions extracted by the first extraction means; a second combination detection means for detecting one combination (P') from the plurality of scanning positions extracted by the extraction means, and the position detection means detects the two combinations (P), (P' 2. The apparatus according to claim 1, wherein the predetermined position is detected using one of the following. 3. The position detection means includes selection means for selecting one of the combinations (P) and (P') based on the magnitude relationship between the characteristic waveform in the image signal and the characteristic waveform in the differential signal. A device according to claim 2, characterized in that: 4. The combination detecting means includes a storage means that stores in advance a standard waveform of the pattern signal as a template, and, at the time of the determination, a center point of the two combined scanning positions and two edge positions of the template. The method according to the present invention further comprises a correlation detecting means for superimposing the waveform of the pattern signal and the template so that the center points coincide with each other, performing a correlation calculation, and detecting one combination with the best correlation. A device according to scope 1. 5. The combination detecting means includes a pattern width determining means for selecting a combination in which the interval between the two combined scanning positions is approximately equal to the interval between the two edges of the pattern, and the correlation calculation is performed on the selected combination. The apparatus according to claim 4, characterized in that the apparatus performs the following. 6 Comprising a pattern detection means that scans a local area on the substrate including an alignment pattern having at least two edges and generates a time-series image signal according to optical characteristics within the local area. , a device for detecting the center position of the pattern within the local range by analyzing the waveform of the image signal, the intensity of the image signal being converted into a digital value pixel by pixel in the scanning direction of the local range. original waveform data generation means for generating digital data representing the waveform of the image signal by converting the digital values and sequentially storing the digital values in a memory having an address corresponding to the position of the pixel; template information storage means for storing in advance template information having level changes corresponding to optical characteristics of the edge portion; reading digital data from the memory;
correlation calculation means for determining the position of a partial waveform most similar to a partial waveform that may occur at an edge portion of the pattern by correlation calculation with the template information; A center position calculation that compares the characteristics of the partial waveform with reference data determined by the shape of the pattern and, based on the comparison result, accurately calculates the center position of the pattern using the determined partial waveform. A pattern position detection device comprising: means.