JP2551049B2 - Alignment device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to two substrates (eg, a glass substrate such as a mask and a reticle and a photosensitive substrate) suitable for use in an exposure apparatus for manufacturing a VLSI, a liquid crystal screen and the like. ) Relates to a device for performing relative alignment (alignment).

[Conventional technology]

2. Description of the Related Art In recent years, miniaturization and densification have progressed in the manufacture of semiconductor elements (VLSI, etc.). It has been used a lot. This type of stepper has a reticle arranged above a projection lens and illuminates a circuit pattern area of the reticle with uniform exposure light from an exposure illumination system.
A projected image of the circuit pattern is projected and exposed on a partial area on the wafer. Since the dimension of the front surface of the wafer is usually larger than the dimension of the projected image of the circuit pattern, the stepper advances the two-dimensional moving stage on which the wafer is placed with respect to the projection lens by a fixed pitch to make the difference on the wafer different. The partial areas are sequentially exposed.

By the way, in the manufacture of a semiconductor device, different circuit patterns are sequentially superposed on each partial region on a wafer by using a large number of reticles to manufacture a device having a desired function. Therefore, it is indispensable to perform an alignment operation for accurately superimposing the circuit pattern transferred so far and the projected image of the circuit pattern of the reticle to be exposed for each partial area on the wafer. .

In order to automate this alignment work, what is called an automatic alignment device is incorporated in the stepper and is operating in an actual manufacturing line. The most reliable method for an automatic alignment device is TTR (Through Th
e Reticle), the mark area on the reticle and the mark attached to the partial area of the wafer are made substantially conjugate with each other by the projection lens, and the mark on the reticle and the mark on the wafer are detected at the same time. This is a method in which the deviation amount of both marks is obtained and the reticle or wafer is finely moved so that the deviation amount becomes zero. In this TTR method, the optical information from the mark on the wafer enters the alignment optical system through the projection lens and the transparent part (glass) in the mark area of the reticle, and the optical information from the reticle mark directly enters the alignment optical system. Is configured to be incident on. By simultaneously detecting the mark on the wafer and the mark on the reticle in this manner, one partial area (hereinafter, referred to as a shot area) on the wafer and the projected image of the reticle pattern are aligned in the positional relationship of the exposure state. You can
The exposure operation can be started immediately after the alignment is completed.

[Means for solving problems]

In such a conventional alignment apparatus, as the optical information corresponding to the mark on the wafer, the specular reflection light, the scattered light, or the diffracted light is photoelectrically detected, and the photoelectric signal waveform is electrically processed to detect the position of the mark. Has been identified. Therefore, if some distortion is added to the waveform of the photoelectric signal itself, an error occurs in specifying the mark position, and this error reduces the alignment accuracy. Recently, an ultraviolet stepper or the like capable of transferring a finer pattern (for example, a line width of about 0.6 μm) is being developed. In order to perform alignment of this fine pattern with sufficient accuracy, the line width A total alignment accuracy of about 1/5 to 1/10 (0.12 to 0.06 μm) is required. Considering that this requirement is satisfied, the distortion of the photoelectric signal waveform has become an error factor that cannot be ignored.

The waveform is distorted by the mark process as a cause of distortion, but in addition to this, it was found that irregularities in the optical characteristics of the transparent portion in the mark area of the reticle are also a cause.

This is because the reticle is a glass substrate with a certain thickness.Therefore, there is a non-uniform refractive index part in the glass in the transparent part in the mark area, internal stress and distortion are applied, and the refraction of light at that part This is because the index, reflectance, etc. slightly change. Therefore, if such an irregularity occurs in the transparent portion of the reticle, the light flux fluctuates when the optical information from the wafer mark passes through it, and the waveform of the photoelectric signal is distorted from the original. .

Therefore, an object of the present invention is to solve the above-mentioned problems and to obtain an alignment apparatus capable of detecting a position with higher accuracy.

[Means for solving problems]

In the present invention, the marks (WM, W
Illuminating means (8, 8) for illuminating a predetermined area including M ') by scanning with spot light (sheet-like beam) or by uniform illumination.
9, 10, 11, 12, 13, 14, 15) and optical information (scattered light, diffracted light, image-forming light flux, etc.) generated from a predetermined area of the second light transmission area of a second substrate such as a reticle or a mask. Photoelectric detection is performed via (mark-constituting window: RM), and the detection direction (scanning direction S) of the optical information that crosses the marks (WM, WM ') on the first substrate.
L) The photoelectric detection means (21, 31) for outputting a photoelectric signal corresponding to the intensity distribution, and the fluctuation of the light intensity distribution caused by the irregularity (40, 41) in the optical characteristics of the light transmission region of the second substrate (40, 41). And a variation extracting means (3, 21, 33) for extracting a variation signal corresponding to the above, and a correction for correcting the intensity of the photoelectric signal from the photoelectric detecting means at the time of detecting the position of the mark by the variation signal stored in the variation extracting means. And means (32) are provided.

[Action]

In the present invention, due to optical irregularities (distortion inside the glass, minute undulations on the surface, spots of oil on the surface, dust particles, etc.) generated in the transparent window area where the alignment marks of the reticle and mask are formed. Focusing on the fact that the optical information from the mark on the wafer is distorted, only the signal waveform component that causes distortion is extracted in advance, and the actual alignment signal waveform is corrected with the signal waveform component that causes distortion, and the reticle and mask The mark position was detected after correcting the signal waveform as if there were no optical irregularity.

〔Example〕

FIG. 1 is a diagram showing the configuration of a projection type exposure apparatus incorporating an alignment apparatus according to the first embodiment of the present invention.

The exposure light from the exposure illumination system 1 uniformly illuminates the reticle R, and the circuit pattern and the like formed on the reticle R are transferred as an image onto the wafer W by the projection lens 2. A photoresist layer having a thickness of about 1 μm is formed on the surface of the wafer W. In this embodiment, the reticle R and the wafer W are arranged so as to be optically conjugate with respect to the projection lens 2, and the projection lens 2 is a so-called bilateral telecentric system in which both the exit pupil and the entrance pupil are located at infinity. Composed. By the way, a reticle mark RM having a rectangular transmission window is formed outside the circuit pattern region of the reticle R, and the wafer mark WM is attached to one shot region of the wafer W.
Is formed. The two-dimensional moving stage 4 vacuum-sucks the wafer W and moves two-dimensionally along the projection image plane of the projection lens 2. The stage 4 is driven by a stage drive system including a motor, etc.
36, which controls stepping for step-and-repeat method and fine movement for alignment operation. Further, at a position different from the wafer W mounting surface of the stage 4, the reference plate 3 provided with a reference mark FM having the same shape as the mark WM on the wafer W should be almost at the same height as the front surface of the wafer W. It is fixed.

Now, the alignment apparatus of the present embodiment is provided with a laser light source 8 for irradiating the reticle mark RM, mark WM, reference mark FM, etc. with coherent laser light, and a beam expander for expanding the diameter of the laser light from the laser light source 8. 9, a cylindrical lens 10 for shaping the shape of the laser spot on the reticle R or the wafer W into an elliptical shape (slit shape), a mirror 11, a vibrating mirror 12 for oscillating the laser spot simply, a mirror 14, and A scanning beam transmitting system including a relay system including lenses 13 and 15 is provided as the illumination means of the present invention. Further, a mirror 18 obliquely provided above the reticle mark RM of the reticle R, a telecentric objective lens 17 horizontally arranged along the optical axis bent by the mirror 18, and a laser beam from a scanning beam transmission system And a beam splitter 16 for separating optical information from each mark, a relay system including lenses 19 and 20, and a mark detection optical system including a photoelectric converter 21 as photoelectric detection means of the present invention. There is. As is clear from the configuration of FIG. 1, the alignment objective lens 17 forms a part of the scanning beam sending system and also a part of the mark detecting optical system.

In the above structure, the laser light is stored as a slit-shaped spot on the reticle R or the wafer W, and is scanned by the vibrating mirror 12 in a direction orthogonal to the longitudinal direction of the spot light. The shake origin of the vibrating mirror 12 is determined by the lenses 13, 15 and the objective lens 17 and the like so as to be conjugate with the exit pupil (or entrance pupil) of the projection lens 2. The plane conjugate with the pupil of the projection lens 2 (pupil conjugate plane) is the objective lens 17 and the beam splitter.
It also exists between 16 and this pupil conjugate plane is further lens 19,
It is conjugated with the light receiving surface of the photoelectric converter 21 by a relay system of 20. Therefore, all light distributions on the pupil plane of the projection lens 2 are imaged on this light receiving surface.

Further, at the pupil conjugate position (light receiving surface) of the photoelectric converter 21, specularly reflected light from the reticle R, the wafer W, or the reference plate 3 (0
A spatial filter 21a that blocks (next light) is provided, and the diffracted light (or scattered light) other than the 0th light is photoelectrically detected by the light receiving elements 22a and 22b.

In the alignment apparatus configured as described above, the slit-shaped spot light SB by the laser light is simultaneously scanned in the direction crossing each of the mark WM on the wafer W and the mark RM of the reticle R, and generated from each mark. Diffracted light or scattered light is photoelectrically detected by the light receiving elements 22a and 22b, and a photoelectric signal whose intensity changes according to the position in the scanning direction is obtained. This photoelectric signal is amplified to a predetermined gain by an amplifier 30 and then input to an alignment waveform memory unit 31 including an analog-digital converter (ADC), a random access memory (RAM), and an address counter. The address counter of the alignment waveform memory unit 31 counts the up / down pulse UDP corresponding to the beam scanning position output from the drive system 35 of the vibrating mirror 12, and generates the count value as an address value of RAM. Therefore, the intensity distribution in the scanning direction of the diffracted / scattered light generated when the slit-shaped spot light scans the marks RM and WM is converted into digital information by the ADC and stored in this RAM. On the other hand, the fluctuation waveform memory unit 33 stores, as digital information, the distribution in the scanning direction of the fluctuation of the light intensity due to the irregular optical characteristics of the rectangular transparent window of the mark RM. The fluctuation waveform memory section 33 constitutes a part of the fluctuation extraction means of the present invention, and the fluctuation distribution waveform information stored therein is, for example, every time the reticle R is exchanged, the mark RM of the reticle. Updated for transparent windows. It is sufficient to perform this updating operation once for the signal waveforms obtained from the light receiving elements 22a and 22b under the operation described later when the reticle R is replaced. Therefore, if the optical irregularity of the transparent window of the mark RM is not corrected at the time of detecting the mark during alignment, the waveform information in the variable waveform memory unit 33 may be flat (or all clear).

The differential operation section 32 is for correcting the intensity based on the waveform information of the photoelectric signals from the light receiving elements 22a and 22b detected at the time of alignment, and the fluctuation waveform information stored in the fluctuation waveform memory section 33. It consists of a value adder or subtractor. On the basis of the signal waveform thus intensity-corrected, the position calculation unit 34 obtains the shift amount between the mark RM and the mark WM in the scanning direction, and the movement amount of the stage 4 necessary to correct the shift amount is determined by the stage drive system. Output to 36.

Now, FIG. 2 shows the plane of the photoelectric converter 21 in FIG. 1, and the outer shape of the circular region which becomes the light receiving surface is almost the same as the projection lens 2
The same size as the pupil conjugate image of. In addition, the light receiving element 2 is placed on the left and right (in the x direction) across the light shielding part 21a in the center of the light receiving surface.
2a and 22b are arranged and the slit spot SB of the laser beam
Receives scattered light that occurs when illuminates the linear edge of each mark. Further, in this embodiment, the light receiving elements 22c and 22d are arranged vertically (y direction) across the central light shielding portion 21a of the light receiving surface.
Are arranged to receive the diffracted light generated when the slit-shaped spot SB illuminates the diffraction grating mark.

FIG. 3 shows a plane shape of the reference mark FM formed on the surface of the reference plate 3 in FIG. 1, the mark FMx is a bar shape linearly extending in the y direction, and the mark FMy is a straight line in the x direction. It has a bar shape that is elongated. These fiducial marks FMX, F
Each of My has exactly the same shape as the wafer mark WM. Further, in this embodiment, the diffraction grating marks FSy and FSx each having the overall shape extending like a slit in the x and y directions are provided on the reference plate 3. This diffraction grating mark FSy, F
Sx also functions as a reference mark like marks FMx and FMy. The entire surface of the reference plate 30 is processed into a mirror surface having a predetermined reflectance (for example, about the same as a wafer).

FIGS. 4 (A) and 4 (B) show the respective positional relationships between the reticle mark RM, the wafer mark WM and the spot light SB when the reticle R and the wafer W are aligned by the apparatus shown in FIG. And a general photoelectric signal waveform obtained in FIG. The spot light SB is extended in a slit shape in the y direction, and is scanned by the vibrating mirror 12 in the x direction along the scanning line SL. Edges RM ₁ and RM ₂ extending in the y direction of the left and right light-shielding areas (chrome layers) of the reticle mark RM, and left and right straight edges of the wafer mark WM formed in a slit shape in the y direction.
When each of WM ₁ and WM ₂ is irradiated with the spot light SB, each of the light receiving elements 22a and 22b receives the scattered light and has a peak at each edge position as shown in FIG. 4 (B). A photoelectric signal waveform is obtained. Such a signal waveform is generated by the vibration drive system 35.
By digitally sampling the signal level from the amplifier 30 by the up / down pulse UDP from
It is stored in the RAM in the alignment waveform memory unit 31. It is desirable that the sampling resolution in the scanning direction (scanning amount per pulse of pulse UDP) be as high as possible.
It is performed every 01 to 0.05 μm. The basic alignment operation based on this signal waveform begins with the peak positions XR ₁ and XR ₂ on the waveform corresponding to the edges RM ₁ and RM ₂ of the reticle mark RM.
Then, the center position CR of these positions XR ₁ and XR ₂ is calculated. Next, the peak positions XW ₁ and XW ₂ on the waveform corresponding to the edges WM ₁ and WM ₂ of the wafer mark WM are obtained, and the center position CW thereof is obtained.
Then, the shift amount Δx between the center positions CR and CW in the x direction (scanning direction) is obtained, and the stage 4 is finely moved so that the shift amount Δx becomes substantially zero, which is completed.

However, the reticle mark RM rectangular transparent window (glass)
For example, if there is an irregularity in optical characteristics as shown in FIG. 5, the peak waveform of the photoelectric signal corresponding to the light passing therethrough, specifically, the edge scattered light from the wafer mark WM, is distorted. May occur. FIG. 5 shows a partial cross section of the reticle R in the vicinity of the mark RM and explains one phenomenon that causes waveform distortion. The laser light LB that forms the spot light SB on the slit is determined so as to form an image on the lower surface P ₂ of the reticle R. On this lower surface P ₂ , a light-shielding portion which forms a window of the mark RM with a chrome layer is vapor-deposited. Laser light
The LB is vertically incident on the lower surface P ₂ and is scanned in the x direction. However, when the laser light LB is incident on the parallel glass plate, a slight amount of reflected light is reflected on the upper surface P ₁ and the lower surface P _{2 of} the glass surface. Occurs, which becomes coast light and returns to the alignment optical system. Laser light incident on the reticle R
As for the ghost light reflected and returned in the light flux of LB, the light shielding part at the center of the light receiving surface of the photoelectric converter 21 shown in FIG.
Although relatively less problem because it is cut by 21a, the ghost light 51 that spreads and returns outside the luminous flux of the laser light LB cannot be completely cut by the light shielding portion 21a of the photoelectric converter 21, and the light receiving element 22a, May be accepted by 22b. However, the amount thereof is extremely small, and usually becomes a remarkably small constant background level as compared with the peak value of the peak waveform of the photoelectric signal.

However, if a distortion portion 41 due to internal stress or the like that deteriorates the uniformity of optical characteristics inside the glass plate, or a waviness portion 40 of minute unevenness on the surface is widely present in the longitudinal direction of the spot light SB, the laser light is generated here. The ghost light 51 generated when the light LB irradiates is reflected irregularly and returns. Therefore, the light receiving elements 22a and 22b of the photoelectric converter 21 receive the ghost light 51 whose light amount changes irregularly during the scanning of the laser light LB, and the background level of the photoelectric signal (which is originally constant) ( Nozzle level) will fluctuate. This occurs as a similar phenomenon when large dust particles firmly adhere to the transparent window of the mark RM, or when oil stains or the like adhere to a relatively wide area in the transparent window.

Information light such as edge scattered light generated when the laser light LB becomes spot light SB on the wafer W and scans the wafer mark WM returns to the alignment system through the transparent window of the reticle R mark RM. Therefore, the information light itself is also affected by the distorted portion 41 and the waviness portion 40, and the wafer mark
Fluctuations occur in the peak waveform corresponding to WM.

Further, since the information light and the ghost light 51 due to the reflection described above are coherent, they may interfere with each other to significantly distort the peak waveform corresponding to the wafer mark WM. In particular, this phenomenon is a phenomenon that cannot be avoided when coherent laser light is used as the illumination light for alignment.

Further, when laser light is used, interference may occur due to the internal reflection itself between the upper surface P ₁ and the lower surface P ₂ of the glass plate of the reticle R, and the signal waveform may be distorted due to the interference.

As described above, if there is an irregularity in the optical characteristics in the transparent window of the mark RM, distortion occurs in the signal waveform detected photoelectrically,
As a result, the unique relationship between the center of the peak waveform and the edge position shifts in the scanning direction (mark detection direction), and an error is included when the mark position is detected.

Therefore, in the first embodiment, after setting the reticle R, the reference plate 3 is placed at the position of the projected image of the transparent window of the mark RM.
The stage 4 is positioned so that the mirror surface without the mark pattern of 1 comes. And the slit-shaped spot light of laser light
SB is scanned in the same manner as shown in FIG. 4 (A). At this time, the scattered light received by the light receiving elements 22a and 22b of the photoelectric converter 21 is the fluctuation of the ghost light in the spot light scanning direction of the transparent window, the interference between the ghost light and the reflected light from the reference plate 3, or the reticle. The intensity changes due to interference due to internal reflection.

 This situation will be described with reference to FIG.

FIG. 6A shows a photoelectric waveform when the mirror surface of the reference plate 3 is scanned with the spot light SB, and the mark RM of the reticle R is shown.
Scanning is performed between the start position Xs and the end position Xe which are sandwiched between. In this signal waveform, only the peak waves corresponding to the respective positions XR ₁ and XR ₂ of the edges RM ₁ and RM ₂ of the mark RM appear, and the fluctuation of the signal level between these two peak waves is due to the transmission window of the mark RM. It is caused by irregularity of optical characteristics. This signal waveform is once stored in the alignment waveform memory unit 31. Here, the peak wave is similar to the intensity distribution (Gaussian distribution) of the spot light SB in the scanning direction, and the width of the tail portion of the peak wave is substantially equal to the width of the spot light SB in the scanning direction. The width of the spot light SB is about 5 μm,
When the sampling resolution is 0.01 μm, one peak wave is stored in RAM with a digital value of about 500 points. With this number of sampling points, the waveform can be reproduced almost faithfully. Incidentally, each sampling position and the RAM address are associated with each other on a one-to-one basis.

Next, the alignment waveform memory unit 31 addresses the waveform data in the range from the positions D ₁ to D ₂ corresponding to the transparent window of the mark RM on the waveform of FIG. And transfer. The fluctuation waveform memory unit 33 stores data from positions Xs to D ₁ where data was not transferred, and positions D ₂
Store a certain level (numerical value) of LV data in the RAM at the address corresponding to the range from Xe to Xe. In this way, the fluctuation waveform data as shown in FIG. 6B is stored in the fluctuation waveform memory unit 33. The constant level LV is set to a value (including zero) smaller than the minimum value of the level Ic on the waveform from the positions D ₁ to D ₂ . This level LV determines the base level after intensity correction of the signal waveform of the wafer mark WM during alignment.

Next, one shot area on the wafer W is the projection lens 2
Step 4 of the stage 4 so that the stage 4 is roughly positioned below. At this time, the reticle mark RM for x-direction alignment and the wafer mark WM are aligned as shown in FIG. 4 (A). And mark RM with spot light SB,
When WM is scanned, the photoelectric signal waveform corresponding to the amount of scattered light detected by the light receiving elements 22a and 22b becomes as shown in FIG. 6 (C) as in FIG. 4 (B). Scanning position D ₁ in FIG. 6 (C)
Two peak waveform generated between the D ₂ is, the wafer mark WM
It corresponds to each of the edges WM ₁ and WM ₂ . The signal waveforms from positions Xs to Xe in FIG. 6C are stored in the RAM in the alignment waveform memory unit 31. After this waveform is captured, the operation calculation unit 32 operates as an alignment waveform memory unit.
The waveform data of FIG. 6 (C) in 31 is changed to the variable waveform memory unit 33.
In order to correct the level with the waveform data shown in FIG. 6 (B), Is (x) = Ia (x)-{Ic (x) -LV} is subtracted at each sampling position (address position). . Here, Ia (x) represents the level of the photoelectric signal waveform at the time of alignment in FIG. 6 (C), and Ic (x)
Represents the level of the waveform in FIG. 6 (B). The waveform created by the calculated level Is (x) is shown in FIG. 6 (D).
become that way. As is clear from the waveform of FIG. 6 (D), the waveform between the positions Xs and D ₁ and the waveform between the positions D ₂ and Xe are not level-corrected under the above subtraction formula. And between positions D ₁ and D ₂ , FIG. 6 (B)
Only the fluctuations from the level LV of the fluctuation waveform shown in are canceled, and a waveform with an almost constant base level LV is obtained. At this time, the edges WM ₁ and WM _{2 of} the wafer mark WM
The distortion component included in the peak waveform corresponding to is also corrected,
It is shaped into a relatively clean Gaussian distribution waveform.

Next, the position calculation unit 34, based on the signal waveforms corrected as shown in FIG. 6D, sets the center positions XR ₁ and XR of the respective peak waveforms.
R ₂ , XW ₁ , and XW ₂ are obtained, and the shift amount Δx is obtained as described with reference to FIG. When using the integration method as one of the methods for obtaining the center position of each peak waveform, for the peak waveform corresponding to the edge of the reticle mark RM, the slice level Vr ₁ near the skirt portion is set, For the peak waveform corresponding to the edge of the wafer mark WM, the slice level Vr ₂ near the skirt portion is set. Then, the waveform portion higher than each slice level Vr ₁ and Vr ₂ is sequentially integrated from one direction to obtain the area, and the integration position at which the area is halved is set as the center position of the peak waveform.
Even if the symmetry with respect to the peak point of the waveform is broken, this method focuses on the center of gravity of the area of the entire peak waveform, so that the mark position can be detected with relatively high accuracy.
Of course, other than such integration method, slice level
The center of the waveform width when the peak waveform is sliced with Vr ₁ and Vr ₂ may be the mark positions XR ₁ , XR ₂ , XW ₁ , and XW ₂ .

In the first embodiment described above, the waveform data in FIG. 6B is a photoelectric signal waveform obtained from the light receiving elements 22a and 22b for detecting scattered light by scanning the mirror surface of the reference plate 3. That is, since the waveform data of FIG. 6 (B) is captured by a mirror surface different from the actual surface of the wafer W, it may not always be canceled cleanly due to a difference in reflectance or the like. Therefore, the equation for obtaining the waveform data in FIG. 6D is modified as follows.

Is (x) = Ia (x) −K · {Ic (x) −LV} Here, K is a constant that gives strength to how to correct the level, and is zero when not corrected, and the average value is 1. is there.
If K is larger than 1, the correction is strong, and if it is smaller than 1, the correction is weak. The optimum value of K should be experimentally obtained in advance.

The wafer mark WM in the first embodiment is a slit-shaped bar mark, but the diffraction grating mark W as shown in FIG.
The same effect can be obtained with M '. The diffraction grating mark WM 'has the same shape as the mark FSx on the reference plate 3, and when the slit spot light SB scans this, a Gaussian diffracted light amount distribution like a waveform PF is generated. In this case, the diffracted light (± 1st order, ± 2nd order, ± 3rd order, etc.) is the mark WM ′.
2 occurs in the arrangement direction (y direction) of the lattice elements, and is received by the light receiving elements 22c and 22d of the photoelectric converter 21 shown in FIG. Therefore, when the diffraction grating mark WM 'is used as the wafer mark, the photoelectric signals from the light receiving elements 22c and 22d and the photoelectric signals from the light receiving elements 22a and 22b that detect the edge scattered light of the reticle mark RM are one signal. A pair of circuits for synthesizing into a waveform or separately processing may be provided. In this way, when the alignment is performed using the diffraction grating mark WM ', the mark detection rate and the position detection accuracy are improved. This is because the diffraction grating mark WM 'has a periodic structure, so that no signal can be obtained unless there are uneven patterns having substantially the same period. Therefore, the photoelectric signal waveform PF
Can have a cleaner Gaussian distribution than the waveform due to edge scattered light. Therefore, when the diffraction grating mark WM 'is used, the mirror surface of the substrate 3 is scanned with the spot SB to capture the waveform data corresponding to the irregularity of the optical characteristics of the transmission window of the mark RM shown in FIG. 6 (A). The photoelectric signals of the light receiving elements 22c and 22d are stored in the variable waveform memory unit 33. Then, the subsequent processing is executed in exactly the same manner.

Next, a second embodiment of the present invention will be described with reference to FIG. In this embodiment, only the method of storing the fluctuation waveform in the fluctuation waveform memory unit 33 and the calculation at the time of level correction are different,
Others are the same. In this embodiment, it is assumed that the diffraction grating mark WM 'shown in FIG. 7 is used for alignment. In this embodiment, the reticle mark RM
Place the grid mark FSx on the reference plate 3 under the transparent window of
The mark FSx is moved by a constant pitch in the same direction (arrow SD) as the scanning direction (x direction) to receive the light receiving elements 22c, 22d.
The peak waveform PF of the photoelectric signal waveform from is extracted, and the peak value Ip is sequentially stored for each moving position. Explaining in a specific example, the spot light SB reciprocally scans a certain area including the reticle mark RM by the vibrating mirror 12, but the peak waveform PF is always extracted during the scanning in one direction, and the mark FSx is Reticle Mark RM Edge RM ₁ to RM ₂
The stage 4 is controlled so as to continuously move at a constant speed up to. The fluctuation waveform memory unit 33 has a circuit configuration as shown in FIG. 9, for example, and a photoelectric signal I from the light receiving elements 22c and 22d is converted into a peak wave PF by a sample-and-hold type analog-digital converter (ADC) 100. Convert the peak value Ip of to a digital value. The digital value is the address counter
Memory (RAM) 102 whose access address is determined by 104
Are sequentially stored in. ADC100 and address counter 104 are
It operates in synchronization with the sample hold timing pulse SHP, and while the spot light SB scans in one direction, the ADC100
Becomes the sample mode, the maximum peak level during that time is sampled, and when the spot light SB starts scanning in the reverse direction, the ADC 100 enters the hold mode and converts the sampled peak level into a digital value. Then, the address counter 104 updates the address value by one each time the ADC 100 switches between the sample mode and the hold mode. With the circuit shown in FIG. 9, for example, every time the spot light SB is scanned by 0.5 μm on the reference plate 3, the peak value I of the waveform PF is calculated.
When sampling p and storing the waveform of the fluctuation distribution,
The frequency of the vibrating mirror 12 is set to about 200 Hz, and the stage 4 is set to about 0.5.
/ 200 (μm / Sec), that is, 25 μm / Sec, it is sufficient to move at a constant speed. Of course, if the peak value Ip of the peak wave PF is sampled on both the forward and backward paths of the reciprocal scanning of the spot light SB, the stage 4 may be moved at a speed of about 50 μm / Sec, which is about twice the value. Further, since the resolution of the up / down pulse UDP from the vibration drive system 35 is about 0.02 μm on the wafer, even if the variation distribution waveform for every 0.5 μm is stored in the RAM 102,
The operation calculation unit 32 cannot directly perform the subtraction for each RAM address. Therefore, it is necessary to perform interpolation processing to convert the waveform stored in the RAM 102 to the same resolution (0.02 μm). Of course, the speed of the stage 4 may be determined so that one count of the address counter 104 corresponds to one pulse of the up / down pulse UDP. That is, assuming that the frequency of the vibrating mirror 12 is 200 Hz and the speed of the stage 4 is 1 μm / Sec (or 2 μm / Sec), the RAM 102 stores the peak value Ip for each 0.02 μm movement of the mark FSx in the order of addresses. The waveform data in the RAM 102 and the waveform data in the alignment waveform memory unit 31 can be directly processed.

FIG. 10A shows an example of the fluctuation waveform Ip (x) stored in the RAM 102 by moving the mark FSx between the positions D _{1 and} D ₂ . This fluctuating waveform is obtained by shifting the positions of the completely same mark FSx, and should be stable at a certain level originally. Therefore, in this embodiment, the waveform Ip (x)
The maximum value It and the minimum value Ib of the reticle are obtained, and the level variation between them is assumed to be due to the optical irregularity of the transparent window of the reticle. FIG. 10 (B) is a waveform Ic ₁ (x) obtained by subtracting the level of the waveform Ip (x) in FIG. 10 (A) from the maximum value It, and this waveform Ic
_The waveform Ic ₂ (x) is obtained by multiplying the level of ₁ (x) by K (K> 1 in the figure). This waveform Ic ₁ (x) (or Ic ₂
(X) is a variation distribution corresponding to the waveform data shown in FIG. 5 (B), but the meaning of the waveform is different from that in the case of FIG. 6 (B). That is, the waveform between the position D ₁ to D ₂ in FIG. 6 (B), while an extra information rather than those to be detected originally, the position D ₁ to D of FIG. 10 (B) The waveform Ic ₁ (x) between ₂ is information about the amount of what was originally detected as a constant value but decreased. Therefore, the calculation processing in the differential calculation section 32 is performed by the position D ₁ of the signal waveforms (similar to FIG. 7) from the light receiving elements 22c and 22d stored in the alignment waveform memory section 31
The portion of the to D ₂ are changed to adding the level of the waveform Ic ₁ (x) of FIG. 10 (B).

According to the present embodiment, more practical correction is possible as compared with the first embodiment in which the mirror surface is scanned. This is because if there are minute irregularities or undulations on the surface in the mirror surface area to be scanned, they will be included in the fluctuation waveform and the correction will be inaccurate. Therefore, each square grid element of the mark FSx shown in FIG. 8 is linearly extended in the x direction (spot scanning direction), and the grid pattern GS is formed in the entire transparent window of the mark RM as shown in FIG. If you unfold it, stage 4
The waveform of FIG. 10 (A) can be obtained by a single scan of the spot light SB similar to that of the first embodiment, without moving at a constant speed, but if a part of the grating pattern GS is irregular. , The fluctuation waveform becomes inaccurate in that part. Of course, if it is guaranteed that the mirror surface of the reference plate 3 and the lattice pattern GS spread over the entire window of the mark RM are free of partial irregularities, it is better to use them to move the stage 4 at a constant speed. It is possible to extract the fluctuating waveform at a higher speed than moving forward.

When the stage 4 is moved at a constant speed and the wafer mark is the bar mark shown in FIG. 4A, the reference mark FMx on the reference plate 3 may be used instead of the lattice mark FSx.

When the fluctuation waveform can be extracted by one scan as in the first embodiment, the fluctuation waveform can be extracted and averaged for each of a plurality of scans to reduce the influence of random noise.
The accuracy of the waveform is improved. Further, in the case of performing such averaging, it is also effective to slightly move the mirror surface portion (or the expanded grating pattern GS) with respect to the window of the mark RM by slightly moving the stage 4 for each scanning of the spot light SB. Is.

Although the respective embodiments of the present invention have been described above, specular reflection light (0th order light) returning to the light receiving surface of the photoelectric converter 21 may be used to extract the fluctuation waveform. In this case, the light shield on the light receiving surface
At the light receiving position of the specularly reflected light of 21a, if a transmissive portion having a size equal to or smaller than the light flux cross-sectional dimension of the specularly reflected light is formed, and the amount of specularly reflected light passing through this transmissive portion is photoelectrically detected. Good. The signal waveform at this time is obtained in the same manner as in FIG. 10 (a).

Further, the present invention is not limited to the method of photoelectrically detecting the reticle mark and the wafer mark by scanning the laser light, and irradiating the reticle mark and the wafer mark with substantially uniform illumination light to form a superimposed image of both marks on a television camera or the like. In addition to observing the bright field (or dark field) with the use of an alignment system that automatically moves the stage 4 by obtaining the positional deviation amount of both marks in the raster scanning direction based on the image signal from the TV camera. In this case, the same can be done. In this case, if the level of the image signal obtained for each raster scan (or the average level of the image signals of a plurality of rasters) is corrected in advance with the waveform of the fluctuation distribution obtained in advance, the same effect can be obtained. can get.

Further, in each of the embodiments of the present invention, the wafer W is finely moved at the time of alignment, but the reticle R may be finely moved. In this case, in the step-and-repeat method, when alignment is performed for each shot area on the wafer, the reticle R may be finely moved each time, which causes the transparent window of the reticle mark RM and the spot light SP to move. The scanning range changes relatively.
For this reason, the reticle R does not correspond to the position of the reticle R when the fluctuation distribution waveform due to the optical irregularity of the transparent window is stored.
Is slightly deviated from the alignment signal waveform in the alignment waveform memory unit 31 and the fluctuation waveform in the fluctuation waveform memory unit 33 by the amount of slight movement in the scanning direction (mark detection direction) of the spot light SB. So TT that makes the reticle R move slightly
In the R alignment method, the initial position of the reticle R when the fluctuation waveform is stored is stored, and immediately before the mark detection operation during alignment, the current position of the reticle R at that time is measured with a laser interferometer, etc. The level correction may be performed after the fluctuation waveform data and the alignment signal waveform data are relatively displaced in the RAM address direction by the amount of deviation between the position and the current position.

In the first and second embodiments, the laser light source 8 is assumed to generate laser light having a wavelength close to that of the exposure illumination light. However, laser light having a wavelength longer than that of the exposure illumination light (argon laser) is used. Etc.) may be used so that the absorption by the photoresist is reduced. In this case, the projection lens 2
Since it becomes difficult to form the spot light SB on both the reticle R and the wafer W due to the chromatic aberration, the correction optical system is provided in the alignment optical path between the reticle R and the projection lens 2, or the alignment optical system for alignment is used. A bifocal element (a birefringent element) between the objective lens 17 and the beam splitter 16.
By inserting the S-polarized light and the P-polarized light of the laser light LB by shifting the image forming points in the optical axis direction by an interval corresponding to the amount of chromatic aberration, and forming one as spot light (sheet beam) on the pattern surface of the reticle R. Then, the other side may be imaged as spot light on the wafer conjugate plane (the conjugate plane under the waveform of the laser beam) which is apart from the pattern surface of the reticle R by the chromatic aberration amount.

Further, the present invention can be applied to an exposure apparatus of a type in which a wafer is observed through a mask substrate, such as a proximity type aligner, in addition to the projection type exposure apparatus.
Further, if an antireflection film for the irradiation light (spot light) from the alignment system of the TTR system is formed on at least one of the upper surface and the lower surface of the reticle R, a better effect can be obtained.

〔The invention's effect〕

As described above, according to the present invention, when a mark on a first substrate such as a wafer is detected through a transparent portion of a second substrate such as a mask (reticle), optical information from the mark is generated due to irregularities in optical characteristics of the transparent portion. Even if is detected with distortion, the distortion can be corrected and a correct signal waveform can always be obtained. For this reason,
In particular, the position detection accuracy of the mark on the first substrate is improved, and the alignment accuracy can be further improved.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a projection type exposure apparatus incorporating an alignment apparatus according to a first embodiment of the present invention, FIG. 2 is a plan view showing a structure of a light receiving surface of a photoelectric converter, and FIG. 3 is a reference plate. A plan view showing the arrangement of the various marks above, FIG. 4 (A),
FIG. 6B is a layout view of the reticle mark and the wafer mark during alignment and a waveform diagram of photoelectric signals obtained correspondingly. FIG. 5 is a cross-sectional view of a glass portion near the alignment mark of the reticle. A), (B), (C),
(D) is a waveform diagram showing how the level of the photoelectric signal waveform during alignment is corrected, FIG. 7 is a plan view showing another shape of the wafer mark, and FIG. 8 is a variation according to the second embodiment of the present invention. FIG. 9 shows a waveform detection operation, FIG. 9 is a circuit block diagram showing a concrete configuration of the variable waveform extracting section in the second embodiment, and FIGS. 10 (A) and 10 (B) are the same as those in the second embodiment. FIG. 11 is a waveform diagram showing an example of the obtained fluctuation waveform, and FIG. 11 is a plan view showing the relationship between the diffraction grating pattern provided on the reference plate and the reticle mark. [Description of symbols of main parts] 2 ... Projection lens, 3 ... Reference plate 4 ... Stage, 8 ... Laser light source 12 ... Oscillating mirror, 17 ... Objective lens 21 ... Photoelectric converter 22a, 22b, 22c, 22d …… Light receiving element 31 …… Alignment waveform memory section 32 …… Fluctuation calculation section, 33 …… Fluctuation waveform memory section 40 …… Waviness, 41 …… Distortion 51 …… Ghost light, R …… Reticle W …… Wafer, SB …… Spot light RM …… Reticle mark WM, WM ′ …… Wafer mark FMx, FMy, FSx, FSy …… Reference mark

Claims (4)

(57) [Claims] 1. When the first substrate having alignment marks and the second substrate having at least a region corresponding to the marks of the first substrate as a light transmitting portion are relatively aligned, the second substrate is used. In an alignment apparatus that optically detects a mark on the first substrate via a light transmission region of the substrate, an illumination unit that illuminates a predetermined region including the mark on the first substrate; and optical information generated from the predetermined region. Photoelectric detection means for performing photoelectric detection through the light transmission region of the second substrate, and outputting a photoelectric signal corresponding to the intensity distribution of the optical information in the detection direction across the mark of the first substrate; Fluctuation extracting means for extracting in advance a fluctuation signal corresponding to fluctuations in the distribution of light intensity generated in the detection direction due to irregularities in the optical characteristics of the light transmitting region; and photoelectric fluctuation signals from the photoelectric detecting means as the fluctuation extracting means. An alignment apparatus, comprising: a correction unit that corrects the intensity by a fluctuation signal from a step to produce a corrected signal, and determines the position of the mark in the detection direction based on the corrected signal. 2. The movement extracting means relatively moves a mark on the first substrate, or a reference mark substantially equivalent to the mark, and a light transmitting region on the second substrate by a predetermined amount in the detection direction. A storage circuit that sequentially stores, as the fluctuation signal, a member and a level corresponding to the mark of the first substrate of the photoelectric signal output from the photoelectric detection unit or the reference mark every time the member moves by a predetermined amount. A device according to claim 1, characterized in that 3. The variation extracting means comprises a reflecting member arranged at a position corresponding to a light transmitting region of the second substrate instead of the first substrate, and a reflecting member reflected by the reflecting member and arranged on the second substrate. A storage circuit that detects the intensity distribution of the optical information that has passed through the light transmission region in the detection direction by the photoelectric detection means and stores the level of the output photoelectric signal as the fluctuation signal. The device according to claim 1. 4. The illuminating means comprises a beam irradiation system for irradiating a beam spot on the first substrate through a light transmitting region of the second substrate and scanning the beam spot in the detection direction. The device according to any one of claims 1 to 3, which is characterized in that: