JP3259341B2 - Alignment method, exposure method using the alignment method, and device manufacturing method using the exposure method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of aligning a mask with a photosensitive substrate suitable for an exposure apparatus used in a lithography process for manufacturing semiconductor devices and liquid crystal devices.

[0002]

2. Description of the Related Art In recent years, in a lithography process for manufacturing a semiconductor device, a step-and-repeat type reduction projection type is used as an apparatus for transferring a pattern formed on a mask or a reticle onto a photosensitive substrate (wafer, glass plate, etc.) with high resolution. Exposure apparatuses (steppers) are frequently used. This type of stepper has a TTL (for example, as disclosed in Japanese Patent Application Laid-Open No. 2-272305) as an alignment sensor for accurately superimposing a projected image of a reticle pattern and a circuit pattern formed in a matrix on a wafer. Through The Lens) type alignment sensor is provided. This alignment sensor irradiates an elongated band-shaped spot light on an alignment mark (diffraction grating mark) on a wafer and photoelectrically detects a diffracted light generated from the mark, and a laser step alignment (LSA) system. On the other hand, two laser beams are irradiated from two different directions to form a one-dimensional interference fringe, and an alignment (La) for photoelectrically detecting interference light of diffracted light generated from the mark in substantially the same direction.
This is a combination of a ser interferometric alignment (LIA) system so that the optical members can be shared as much as possible.

At present, the alignment method of the stepper is as follows.
For example, JP-A-61-44429 and JP-A-62-8
An enhanced global alignment (EGA) as disclosed in Japanese Patent No. 4516 or the like has become mainstream. The EGA method means that at least three of a plurality of exposure regions (shot regions) formed on a wafer, for example, near the center of the wafer and its periphery, before performing overlapping exposure on one wafer 7 sets of shot areas located in the area, and two sets (X, Y directions) of alignment marks (wafer marks) attached to each shot area
Is measured (sample alignment) by the alignment sensor. Thereafter, based on these mark positions (measured values) and design values, error parameters relating to the arrangement characteristics of the shot areas on the wafer, that is, the offset of the wafer center position (X and Y directions) and the degree of expansion and contraction of the wafer (X , Y direction), the remaining rotation amount of the wafer, and the orthogonality of the wafer stage (or the orthogonality of the shot arrangement) in total of 6
The two parameters are determined by a statistical operation (such as a least square method). Then, based on the determined parameter values, the design coordinate values of all shot areas on the wafer are corrected, and the wafer stage is stepped according to the corrected coordinate values (calculated values). It is a way to go. As a result, the projected image of the reticle pattern and each of the plurality of shot areas on the wafer are accurately overlapped and exposed.

When a reticle pattern is exposed while being superimposed on a shot area on a wafer, the projected image of the reticle pattern may be distorted due to fluctuations in the reticle position and baseline, fluctuations in the projection magnification and distortion of the projection optical system, and the like. There has been a problem that the shot region is displaced in the X and Y directions, and sufficient overlay (alignment) accuracy cannot be obtained. Therefore, conventionally, prior to overlay exposure (main exposure) on a wafer on which a shot region is formed, overlay exposure (test printing) of a reticle pattern (vernier pattern, etc.) on a pilot wafer.
Then, the amount of deviation between the two resist images (main scale pattern and sub scale pattern) formed on the wafer by the development process is detected by an alignment sensor or a dedicated inspection device. When the overlay exposure is actually performed, the projected image of the reticle pattern and the shot area can be accurately overlapped by giving the offset amount to the measurement result of the alignment sensor as an offset.

However, the above method requires a dedicated pilot wafer for measurement, and has a problem that the measurement time is lengthened by the development process. For this reason, for example, as disclosed in Japanese Patent Application Laid-Open No. 61-201427, there has been proposed a method of using a latent image formed on a resist layer by irradiating a reticle mark with exposure light. In the method disclosed in the above publication, after a reticle pattern is aligned with one shot area on a wafer using an alignment sensor, exposure light is applied to the reticle mark prior to the main exposure to expose the reticle mark to the resist layer. The latent image is formed. Then, the same alignment sensor is used to simultaneously detect the wafer mark (base mark) already formed along with the shot area and the latent image of the reticle mark formed in the vicinity of the mark. Find the shift amount. Further, the reticle and the wafer are relatively moved in accordance with the positional deviation amount, that is, realignment is performed. Thereafter, by starting the main exposure, the projected image of the reticle pattern and the shot area on the wafer can be accurately overlapped and exposed.

The above method does not require a pilot wafer,
In addition, there is an advantage that a position shift (offset) can be measured in a short time without performing a developing process. In addition, if the amount of misregistration using the latent image as described above is measured for each shot area on the wafer, the processing time for one wafer becomes longer, and the throughput of the apparatus may decrease. For this reason, for example, the above-described measurement is performed only on the first (first) shot area on the wafer, and the reticle pattern and the shot are determined on the second and subsequent shot areas by using the previously measured positional shift amount. What is necessary is just to perform the alignment (alignment) with the area.

[0007]

However, in the prior art as described above, when alignment of all shot areas on a wafer is performed using the positional deviation measured in the leading shot area, the expansion and contraction (scaling) of the wafer is performed.
If the shot area is misaligned due to the above-described factors, there is a problem that accurate superposition cannot be performed even if the above-described misalignment amount is given as an offset. In addition, depending on the measurement accuracy (reproducibility) of the alignment sensor, it is difficult to perform accurate overlay even if the offset correction is performed based on the amount of positional deviation. Further, when a latent image is detected by an alignment sensor that uses illumination light having a wavelength substantially the same as the exposure light, the optical characteristics of the latent image change due to the irradiation of the illumination light, which adversely affects the measurement accuracy. There is. In addition, the latent image of the reticle mark is always formed near the wafer mark (base mark), which greatly restricts the formation position. In the next layer, the latent image is formed near the wafer mark as the base mark. Therefore, there is also a problem that the alignment accuracy in the next layer may be reduced by the base mark (pseudo alignment mark) due to the latent image.

The present invention has been made in view of the above points, and a latent image of a reticle mark can be formed at an arbitrary position on a photosensitive substrate. It is an object of the present invention to provide an alignment method capable of performing alignment with the alignment.

[0009]

According to the present invention, in order to solve the above-mentioned problems, each of a plurality of exposure areas (SA) arranged on a substrate (W) is moved to a stationary position which defines a moving position of the substrate. In a method of aligning with a predetermined exposure position in a coordinate system, position information of at least three sample areas (sample shots) of the plurality of exposure areas (SA) in the stationary coordinate system is measured. The regularity of the arrangement of the plurality of exposure areas, which is calculated by performing statistical operation on the measured position information of the plurality of sample areas.
A first step of obtaining position information of at least a plurality of desired exposure areas (virtual shots) in the stationary coordinate system using a plurality of first parameters (Ag, Og), A second step of exposing a specific mark (RM) to each of the plurality of desired exposure regions (virtual shots) positioned at the exposure position based on position information of the desired exposure region (virtual shot); By measuring position information of at least three specific marks in the stationary coordinate system among the specific marks and statistically calculating the position information of the measured plurality of specific marks, the position information of each of the plurality of specific marks is obtained. Rules for arranging the specific marks used for calculating position information in a stationary coordinate system
A third step of calculating a plurality of second parameters (At, Ot) relating to sex, and each of the plurality of exposure areas (SA) is aligned with the exposure position based on the first parameter and the second parameter. And a fourth step of determining position information for the second position.

[0010]

According to the present invention, a plurality of first parameters calculated by statistically calculating position information in at least three sample areas of the plurality of exposure areas on the substrate in the stationary coordinate system are used. Position information of at least a plurality of desired exposure regions in a stationary coordinate system is obtained, and specific marks are exposed at a plurality of positions on the substrate according to the calculated position information. Then, by statistically calculating the positions of the latent images of at least three specific marks in the static coordinate system among the images (latent images) of the plurality of specific marks, the static coordinates of each of the plurality of specific marks (latent images) are calculated. Calculate a plurality of second parameters used to determine the 'position on the system. Then, based on the first parameter and the second parameter, each of the plurality of exposure regions on the substrate is aligned with the exposure position. For this reason, regardless of the measurement reproducibility of the alignment sensor, the amount of positional deviation at the time of the overlay exposure can be accurately measured, and high-precision alignment can be always performed.

According to this embodiment, the specific mark can be exposed at an arbitrary position on the substrate, and there is no restriction on the position where the latent image is formed. In particular, if the latent image of the specific mark is formed in an unexposed area near the outer periphery of the substrate, no latent image is formed near the alignment mark (base mark) attached to the exposed area, and the latent image is not formed. A decrease in alignment accuracy in the next layer due to image formation can also be prevented.

[0012]

FIG. 1 is a diagram showing a schematic configuration of a projection exposure apparatus suitable for applying a positioning method according to a first embodiment of the present invention, and FIG. 2 is a TTL type alignment shown in FIG. FIG. 3 shows a specific configuration of the sensor, and FIG. 3 is a block diagram of a control system of the device shown in FIG. Hereinafter, before describing the alignment method of the present embodiment, a schematic configuration of a projection exposure apparatus will be described.

In FIG. 1, illumination light IL generated from an ultra-high pressure mercury lamp 1 is reflected by an elliptical mirror 2 and once collected at a second focal point thereof.
The light enters an illumination optical system 3 including an optical integrator (fly-eye lens) and an aperture stop (σ stop).
Although not shown, the fly-eye lens is arranged in an in-plane direction perpendicular to the optical axis AX such that its reticle-side focal plane substantially matches the Fourier transform plane (pupil conjugate plane) of the reticle pattern. Further, a shutter (for example, a rotary shutter with four blades) 37 for closing and opening the optical path of the illumination light IL by a motor 38 is arranged near the second focal point. The exposure illumination light source 1 may use a laser light source such as a KrF or ArF excimer laser, or a harmonic of a metal vapor laser or a YAG laser, in addition to a bright line such as a mercury lamp.

Most of the illumination light (i-line or the like) IL in the wavelength range for exposing the resist layer emitted from the illumination optical system 3 is reflected by the beam splitter 4 and relay lenses 5, 7,
After passing through a variable field stop (reticle blind) 6 and reaching a mirror 8 where the light is reflected almost vertically downward, the pattern area PA of the reticle R is illuminated with a substantially uniform illuminance via a main condenser lens 9. I do. Since the surface of the reticle blind 6 is in a conjugate relationship (imaging relationship) with the reticle R, the drive system 36 opens and closes a plurality of movable blades constituting the reticle blind 6 to change the size and shape of the opening. , The illumination visual field of the reticle R can be set arbitrarily.

FIG. 4A is a diagram showing a specific configuration of the reticle R. In this embodiment, four reticle marks are provided at substantially the center of each of four sides of the pattern area PA surrounded by the light-shielding band LSB. RMx ₁ , RMx ₂ , RMy ₁ , and RMy ₂ are formed. These reticle marks are formed as latent images on the resist layer, and in this embodiment, are also used as alignment marks for positioning with the shot area. Further, the four reticle marks are the same configuration, it will be described here only marks RMx _1. Figure 4 alignment marks RMx ₁ as shown in (b)
Is a matrix (multi-mark) in which five diffraction grating marks composed of seven dot marks arranged in the Y direction are arranged at predetermined intervals in the X direction (multi-marks). It is formed by a light shielding part such as chrome. Further, on the reticle R, two sets of alignment marks RAM ₁ and RAM ₂ are formed to face each other in the vicinity of the outer periphery. The alignment marks RAM ₁ and RAM ₂ are both cross-shaped light-shielding marks, and are used for alignment of the reticle R (positioning with respect to the optical axis AX).

The reticle R is mounted on a reticle stage RS which can be finely moved in the optical axis direction of the projection optical system 13 by a motor 12 and which can be two-dimensionally moved and minutely rotated in a horizontal plane. At the end of the reticle stage RS, a movable mirror 11m for reflecting a laser beam from a laser light wave interferometer (interferometer) 11 is fixed, and the two-dimensional position of the reticle stage RS is determined by the interferometer 11. For example, it is always detected with a resolution of about 0.01 μm. Reticle alignment (RA) systems 10A and 10B use reticle R
The two sets of alignment marks (cross-shaped marks RAM ₁ , RAM ₂ in FIG. 4A) formed near the outer periphery of FIG. By finely moving reticle stage RS based on detection signals from RA systems 10A and 10B, reticle R is positioned such that the center point of pattern area PA coincides with optical axis AX.

The illumination light IL that has passed through the pattern area PA is incident on a projection optical system 13 that is telecentric on both sides, and the projection optical system 13 reduces the projected image of the circuit pattern of the reticle R to 1/5, A resist layer is formed on the
The surface is superimposed and projected (imaged) on one shot area on the wafer W held so that its surface substantially coincides with the best imaging plane. FIG. 5A is a diagram showing an arrangement relationship between one shot area SA on the wafer W and the wafer marks MX and MY, and FIG. 5B is a diagram showing a configuration of the wafer mark MX. In FIG. 5A, four sides of the shot area SA are surrounded by street lines ST, and alignment marks MX and MY are formed at approximately the center of each of the two side street lines ST orthogonal to each other. . Each of the alignment marks MX and MY is a one-dimensional diffraction grating mark and is formed as a concave portion or a convex portion on a base. As shown in FIG. 5B, the wafer mark MX is
Five bar marks extending in the Y direction are arranged at a predetermined pitch in the X direction (measurement direction).

The wafer W is vacuum-sucked on a micro-rotatable wafer holder (not shown), and is held on the wafer stage WS via this holder. Wafer stage WS
Is 2 in a step-and-repeat manner by the motor 16.
When the transfer exposure of the reticle R to one shot area on the wafer W is completed, the reticle R is stepped to the next shot position. A movable mirror 15m for reflecting the laser beam from the interferometer 15 is fixed to an end of the wafer stage WS.
The two-dimensional position of S is determined by the interferometer 15 using, for example,.
It is always detected with a resolution of about 01 μm.

On the wafer stage WS, a reference member (glass substrate) 14 having fiducial marks used at the time of baseline measurement or the like is provided so as to substantially coincide with the surface position of the wafer W. . As the fiducial mark, the reference member 14 has a slit pattern FMa composed of five sets of light-transmissive L-shaped patterns as shown in FIG. The duty ratio is 1: 1) FMx and FMy are provided. In the reference pattern FMx, three diffraction grating marks in which seven dot marks are arranged in the Y direction and 12 bar marks extending in the Y direction are arranged in the X direction around the three diffraction grating marks. Things. Note that the reference pattern FMy has the same configuration as the pattern FMx, and a description thereof will be omitted.

The slit pattern FMa is configured to be illuminated from below (inside the wafer stage) by illumination light (exposure light) transmitted below the reference member 14 using an optical fiber (not shown) or the like. I have. The illumination light transmitted through the slit pattern FMa forms a projection image of the slit pattern FMa on the back surface (pattern surface) of the reticle R via the projection optical system 13. Furthermore, reticle mark R
Illumination light that has passed through any of Mx ₁ , RMx ₂ , RMy ₁ , and RMy ₂ passes through the condenser lens 9, the relay lenses 5 and 7, and the beam splitter 4, and then forms a pupil conjugate plane of the projection optical system 13. The light is received by a photoelectric detector 35 disposed in the vicinity, and the photoelectric detector 35 outputs a photoelectric signal SS corresponding to the intensity of the illumination light to the main control system 18. Hereinafter, the optical fiber, the reference member 14, and the photoelectric detector 35 are collectively referred to as an ISS (Imaging Slit Sensor) system.

In FIG. 1, there is also provided an image forming characteristic correcting section 19 capable of adjusting the image forming characteristics of the projection optical system 13.
The correction unit 19 in the present embodiment independently drives a part of the lens elements constituting the projection optical system 13, in particular, each of a plurality of lens elements close to the reticle R using a piezoelectric element such as a piezo element (optical axis). By performing parallel movement or tilting with respect to AX, the imaging characteristics of the projection optical system 13, such as the projection magnification and distortion, are corrected.

Next, two sets of alignment sensors provided in the apparatus having the above configuration will be described. Off-axis alignment sensor (Field Image Alignment;
The FIA system guides light generated by the halogen lamp 20 to the interference filter 23 via the condenser lens 21 and the optical fiber 22, and cuts light in the photosensitive wavelength region and the infrared wavelength region of the resist layer. Interference filter 2
The light transmitted through 3 enters a telecentric objective lens 27 via a lens system 24, a beam splitter 25, a mirror 26, and a field stop BR. Further, the light is reflected by a prism (mirror) 28 fixed around the lower part of the lens barrel so as not to block the illumination visual field of the projection optical system 13 and irradiates the wafer W almost vertically.

The light from the objective lens 27 is applied to a partial area including the wafer mark (base mark) MX, and the light reflected from the area is reflected by the prism 28, the objective lens 27, the field stop BR, the mirror 26, the beam splitter 25, And a guide plate 30 through a lens system 29. Here, the index plate 30 is arranged in a plane conjugate with the wafer W by the objective lens 28 and the lens system 29, and an image of the alignment mark MX on the wafer W is formed in a transparent window of the index plate 30. You. Further, the index plate 30 is formed such that two linear marks extending in the Y direction are arranged at predetermined intervals in the X direction as index marks in the transparent window. The light that has passed through the index plate 30 is transmitted to the image pickup device (CC
D, etc.), and an image of the alignment mark MX and an image of the index mark are formed on the light receiving surface thereof.
The image signal SV from the image sensor 34 is sent to the main control system 18, where the position (coordinate value) of the alignment mark MX in the X direction is calculated. Although not shown in FIG. 1,
In addition to the FIA system (X-FIA system) having the above-described configuration, another FIA system (Y-system) for detecting a mark position in the Y direction is used.
FIA system) is also provided.

Next, the TTL type alignment sensor 17 will be described with reference to FIG. The alignment sensor 17 is a combination of an LIA system and an LSA system with the optical members being shared as much as possible, and the specific configuration is disclosed in Japanese Patent Application Laid-Open No. 2-272305. explain. In FIG. 2, a laser beam emitted from a light source (He-Ne laser or the like) 40 is split by a beam splitter 41, and the reflected beam is transmitted to a first beam shaping optical system (LIA optical system) 45 via a shutter 42. Incident. On the other hand, the beam transmitted through the beam splitter 41 enters a second beam shaping optical system (LSA optical system) 46 via a shutter 43 and a mirror 44. Therefore, by appropriately driving the shutters 42 and 43, it is possible to switch between the LIA system and the LSA system for use.

The LIA optical system 45 includes two sets of acousto-optic modulators and emits two beams having a predetermined frequency difference Δf so as to be substantially symmetric with respect to the optical axis. Further, the two beams emitted from the LIA optical system 45 enter a beam splitter 49 via a mirror 47 and a beam splitter 48, and the two beams transmitted therethrough are transmitted through a lens system (inverse Fourier transform lens) 53, Then, the light enters the reference diffraction grating 55 fixed on the apparatus from two different directions at a predetermined intersection angle via the mirror 54 and forms an image (intersects). The photoelectric detector 56 receives the interference light of the diffracted light passing through the reference diffraction grating 55 and generated in substantially the same direction, and outputs a sinusoidal photoelectric signal SR corresponding to the intensity of the diffracted light to the main control system 18, ie, Output to the LIA operation unit 58.

On the other hand, the two beams reflected by the beam splitter 49 are applied to a field stop 51 by an objective lens 50.
Once, the light enters the projection optical system 13 via a mirror M ₂ (the mirror M _{1 in} FIG. ₁ is not shown). further,
After being condensed once in the form of a spot on the pupil plane of the projection optical system 13 so as to be substantially symmetric with respect to the optical axis AX, symmetrically with respect to the pitch direction (Y direction) of the alignment mark MY on the wafer W with respect to the optical axis AX. A parallel light beam inclined at an appropriate angle enters the wafer mark MY from two different directions at a predetermined intersection angle. On the wafer mark MY, the frequency difference Δf
A one-dimensional interference fringe moving at a speed corresponding to
± 1st-order diffracted light (interference light) generated in the same direction from the mark, here in the optical axis direction, is received by the photoelectric detector 52 via the projection optical system 13, the objective lens 50, and the like. And outputs a sinusoidal photoelectric signal SDw corresponding to the period of the light / dark change to the LIA operation unit 58. LI
The A operation unit 58 calculates the position shift amount of the wafer mark MY from the phase difference on the waveforms of the two photoelectric signals SR and SDw, and uses the position signal PDs from the interferometer 15 to reduce the position shift amount to zero. The coordinate position of wafer stage WS at the time of is obtained, and this information is output to alignment data storage unit 61 (FIG. 3).

The LSA optical system 46 includes a beam expander, a cylindrical lens, and the like, and a beam emitted from the LSA optical system 46 enters an objective lens 50 via a mirror 48 and a beam splitter 49. Further, the light is converged once in a slit shape by the field stop 51 by the objective lens 50,
After passing through the approximate center of the pupil plane is incident on the projection optical system 13 via a mirror M _2, extending in the X direction in the image field, and on the wafer W as an elongate strip spotlight as toward the optical axis AX Is formed. When the spot light and the wafer mark (diffraction grating mark) are relatively moved in the Y direction, light generated from the mark is received by the photoelectric detector 52 via the projection optical system 13, the objective lens 50, and the like.
The photoelectric detector 52 outputs ± 1 to ± 1 of the light from the wafer mark.
Only the third-order diffracted light is photoelectrically detected, and a photoelectric signal SDi corresponding to the light intensity is output to the main control system 18, that is, the LSA operation unit 57. The LSA operation unit 57 includes the interferometer 1
5 is also input, and the wafer stage WS
The photoelectric signal SDi is sampled in synchronization with an up / down pulse generated for each unit movement amount. Further, after converting each sampling value into a digital value and storing it in a memory in the order of addresses, the position of the wafer mark in the Y direction is calculated by a predetermined arithmetic processing, and this information is output to the alignment data storage unit 61.

By the way, in this embodiment, prior to the overlay exposure (main exposure), a reticle mark (diffraction grating mark) is exposed to the wafer W by exposure light to form a latent image on the resist layer. An LSA system is mainly used as a sensor for detecting a latent image. Therefore, LSA
The system and LSA operation unit 57 detects the latent image of the reticle mark by exactly the same operation as the wafer mark (base mark), and outputs the position information to the alignment data storage unit 61. Although not shown in FIG. 1, the alignment sensor having the above configuration (Y-LIA system, Y-LSA system)
In addition to the above, another set of TTL alignment sensors (X-LIA system,
X-LSA system) is also provided.

Next, with reference to FIG. 3, a description will be given of the main control system 18 for integrally controlling the above apparatus. In this embodiment, L
The main control system 18 is configured from the SA operation unit 57 to the system controller 65 in code order. In FIG.
The FIA operation unit 59 calculates a positional shift amount of the image of the wafer mark with respect to the index mark by a predetermined operation based on the waveform of the image signal SV from the image sensor 34. Further, the position signal PDs from the interferometer 15 is also input, and the coordinate position of the wafer stage WS when the image of the wafer mark is accurately located at the center of the index mark (the amount of displacement is zero) is obtained. The information is output to the alignment data storage unit 61. The ISS calculation unit 60 also outputs the position signal PDs from the interferometer 15 together with the photoelectric signal SS output from the photoelectric detector 35 when the reference member 14 (slit pattern FMa) and the reticle R are relatively moved.
Also enter The photoelectric signal S is synchronized with an up / down pulse generated for each unit movement amount of the wafer stage WS.
S is sampled, each sampled value is converted into a digital value, and stored in the memory in the order of addresses. Then, the position of the reticle mark in the X or Y direction (a rectangular coordinate system defined by the interferometer 15) XY coordinates are calculated, and this information is output to the alignment data storage unit 61.

The alignment data storage unit 61 stores
The mark position information (including the position of the latent image) from each of the three arithmetic units 57 to 60 can be input. EG
The A operation unit 62 includes four operation units 57 to 60
Various calculations are performed by using the position information from the computer, and the calculation results are sent to the storage unit 63 and the system controller 65. As an example, based on the mark position information stored in the alignment data storage unit 61, an array coordinate value of the shot area on the wafer W is calculated by a statistical calculation method. The EGA calculation unit 62 calculates calculation parameters prior to the calculation of the array coordinate values as described above, that is, the offset of the wafer center position (X and Y directions), the degree of expansion and contraction of the wafer (X and Y directions), the residual rotation error of the wafer, The orthogonality of the wafer stage or the orthogonality of the shot array (conversion matrices A and O described later) is also calculated, and this information is also stored in the storage unit 63.

Further, the shot map data section 64 includes a design exposure position (array coordinate value Dn) of a shot area to be exposed on the wafer, and the position and number of shot areas to be sample-aligned (hereinafter, referred to as sample shots). ,
The position and the number of the reticle mark where the latent image is to be formed are stored. These data are sent to the EGA operation unit 62 and the system controller 65. Further, the system controller 65 determines a series of procedures for controlling the movement of the wafer stage WS at the time of alignment, at the time of forming a latent image, at the time of exposure in a step-and-repeat method, and the like based on the above data. Further, FIG. 3 also shows a wafer stage controller 66 and a reticle stage controller 67.

The positions and numbers of sample shots and the positions and numbers of latent images are determined in advance so as to satisfy both alignment accuracy, calculation accuracy of array offsets (ΔOx, ΔOy) described later, and throughput. These data are set in advance in the shot map data section 64 by an operator using an input device (such as a keyboard (not shown)), or are automatically selected and set by the system controller 65 according to a predetermined program. It is good to keep it. Further, the above data may include, for example, the types of wafers, underlayers, resists (and their film thicknesses), processing conditions (heating temperature, time, etc.) in the wafer processing process, or alignment sensors used to detect underlying marks and latent images. It may be determined empirically, by experiment, simulation, or the like, based on the type or the like, and may be different for each unit number of wafers or for each lot.

Next, an alignment method according to the present embodiment will be described. In this embodiment, the alignment of the EGA system is performed using the FIA system at the time of the overlay exposure (main exposure), and the LSA system is used for detecting the latent image. Further, as shown in FIG. 7, a plurality of circuit patterns (shot areas SAij) and a wafer mark (MX)
ij, MYij) are formed in a matrix, and the surface thereof is coated with a resist having a substantially constant thickness (about 0.5 to 2 μm). In FIG. 7, a shot area of an i-th row (the i-th shot row arranged in the X direction from the top of the paper), a j-th row (the j-th shot row arranged in the Y direction from the left in the paper), and its wafer The mark is defined as a shot area SAij, a wafer mark MXij,
MYij.

In the stepper shown in FIGS. 1 to 3, the reticle R (FIG. 4) is loaded on the reticle stage RS, and the reticle R is moved so that its center coincides with the optical axis AX.
After rough alignment is performed by the set of RA systems 10A and 10B, the RA system 10A and 10B are vacuum-sucked to the reticle stage RS. Further, the main control system 18 uses the reference member 14 to control the reticle R
Perform fine alignment. That is, four reticle marks RMx ₁ , R
Each of Mx ₂ , RMy ₁ and RMy ₂ is detected. The EGA operation unit 62 calculates the value of 4 calculated by the ISS operation unit 60.
From the two coordinate values, the reticle R is calculated with respect to the orthogonal coordinate system XY defined by the interferometer 15 in terms of X, Y, and the amount of displacement in the rotational direction. Then, the system controller 65 drives the reticle stage RS while monitoring the output of the interferometer 11 with the stage controller 67 so that the amount of displacement becomes substantially zero. As a result, the reticle R is accurately positioned with respect to the optical axis AX, and the reticle alignment ends.

Next, the main control system 18 uses the reference member 14 to measure the baseline of the X-FIA systems 20 to 34. First, to detect the reticle mark RMx ₁ using a slit pattern FMa, ISS calculation unit 60 outputs the coordinate value to the alignment data storage unit 61. Thereafter, the reference pattern FMx is detected using the X-FIA system, and the FIA operation unit 59 stores the coordinate values in the storage unit 61.
Output to The EGA operation unit 62 calculates a base line ΔBx in the X direction from the two coordinate values stored in the storage unit 61.
Is obtained and stored in the storage unit 63. Note that a baseline measurement is also performed by a Y-FIA system (not shown), and the value ΔBy is stored in the storage unit 63.

After the baseline measurement as described above is completed, the first wafer W in the lot (FIG. 7) is loaded on the wafer stage WS, and the system controller 65 uses the two sets of FIA systems to load the wafer W. Perform pre-alignment. First, Y-FIA system in the vicinity of the outer periphery of the wafer W, the wafer center two shot areas formed in substantially symmetrical positions with respect to (the area shot center of the SA _56), for example, the shot area SA in Fig. 7 _72, SA detecting the wafer mark MY _72, MY ₇₈ of _78, FIA computation unit 59 calculates the position of the Y-direction (coordinate values). on the other hand,
The X-FIA system detects a shot area which is almost equidistant from the two shot areas SA ₇₂ and SA ₇₈ , for example, a wafer mark MX ₂₄ of the shot area SA _{24 in} FIG. In the arithmetic unit 59, the X
The position in the direction is calculated.

The EGA calculation unit 62 calculates the amount of displacement (including the rotation error) of the wafer W with respect to the rectangular coordinate system XY based on the mark position information of the three shot areas stored in the alignment data storage unit 61. . Thereafter, the system controller 65 drives the wafer stage WS by the stage controller 66 so that the amount of displacement becomes substantially zero, thereby completing the pre-alignment of the wafer W. As a result, the relative displacement between reticle R and wafer W (shot area SAij) is 1 μm.
It is corrected with an accuracy of m or less.

Next, the above-mentioned JP-A-61-44429 is described.
In accordance with the method disclosed in Japanese Patent Application Laid-Open Publication No. H11-207, array coordinate values (Fgx, Fgy) of all the shot areas SAij on the wafer W are calculated. First, the system controller 65 has two sets of FIAs.
Perform sample alignment using the system. Therefore,
Data (position, number) relating to the sample shot is read out from the shot map data section 64, and this data, that is, the design array coordinate values (Dx, D
y), the baseline ΔBx stored in the storage unit 63,
The wafer stage WS is stepped according to ΔBy. Thereby, the wafer mark MX of the sample shot is provided for each of the X-FIA system and the Y-FIA system.
ij and MYij are sequentially positioned. Here, among the plurality of shot areas ij on the wafer W, the number of sample shots in this embodiment is m (m ≧ m) located near the outer periphery thereof.
3) shot areas, for example, the seven shot areas SA ₁₃ , SA ₂₆ , SA ₃₂ , SA ₅₉ , SA ₆₂ , SA ₇₇ , and so on in FIG.
It is assumed that SA ₈₂ has been set. As a result, FIA
The arithmetic unit 59 calculates the mark position by a predetermined waveform processing, and the position information is based on the array coordinate values (Mgx, Mg).
gy) is sent to the alignment data storage unit 61.

Here, in this embodiment, two sets (X, Y directions) of the wafer mark MX of each of the seven shot areas SAij are set.
Although the positions of ij and MYij are measured, marks that cannot be measured or marks whose measured values are extremely different from the design values (hereinafter referred to as defective marks) are rejected.
Only the remaining mark positions may be used. At this time, the shot area itself having the defective mark may be rejected. Alternatively, a shot area near the shot area to be rejected may be selected as an alternative shot, and its mark position may be measured and used. In the alternative shot, only one of the wafer marks MXij and MYij corresponding to the defective mark may be measured. For example, when the wafer mark MX ₈₂ of the shot area SA ₈₂ is defective, select the shot area SA ₇₃ alternatively shots, the wafer mark MX _73, MY
_{It is sufficient} to measure the positions of both the marks ₇₃ or only the mark MX ₇₃ . Further, a shot area in which both the positions of the wafer marks MXij and MYij are measured is regarded as one sample shot, and in the present embodiment, it is necessary to perform mark measurement on at least three sample shots. When only one mark (for example, in the X direction) is measured in one shot area, and when the other (Y direction) mark is measured in the substitute shot, these marks are counted as one sample shot. You may do it.

The EGA operation unit 62 calculates the measured array coordinate values (Mgx, Mgy) of the seven sample shots stored in the alignment data storage unit 61 and the design coordinate values stored in the shot map data unit 64. Based on the array coordinate values (Dx, Dy), the regularity of the shot array on the wafer W to be aligned by the step-and-repeat method is determined. That is, the transformation matrices Ag and Og in the mapping relational expression (determinant Mg = Ag · D + Og) shown in the following Expression 1 are determined. However, the transformation matrices Ag and Og in the above relational expressions are the remaining rotation error θ _g , the orthogonality ω _g and the scaling errors Rgx and Rgy, and the offset error Og
x and Ogy are included as parameters, and the transformation matrix A
g is a matrix of 2 rows and 2 columns, and Og is a matrix of 2 rows and 1 column.

[0041]

(Equation 1)

The transformation matrices Ag and Og are given by
It is represented by 3.

[0043]

(Equation 2)

[0044]

(Equation 3)

Here, the shot area on the wafer has a residual term (εXg, εYg) with respect to the measured array coordinate values (Mgx, Mgy) and the designed array coordinate values (Dx, Dy). Equation 1 above can be rewritten as Equation 4 below.

[0046]

(Equation 4)

[0047] Thus, EGA calculation unit 62 as described above residue Sako (i.e. the sum of the squares of residuals) is minimized, transform matrix Ag, calculates the values of the parameters a _g ~f _g of Og (least square method ). The conversion matrices Ag and Og calculated as described above are stored in the storage unit 63. Further, the EGA operation unit 62 calculates the array coordinate values (Fgx, Fgx,
Fgy) is calculated, and the calculation result is sent to the storage unit 63 and the system controller 65.

In this embodiment, as will be described later, an area near the outer periphery of the wafer W and on which no circuit pattern or alignment mark is formed (unexposed area excluding shot areas and street lines), for example, as shown in FIG. the region ND ₁ to ND ₅ of 7 to form a latent image of the reticle mark. These unexposed areas are areas where the reticle pattern has not been transferred because part of the shot area on the wafer W is missing as shown by the dotted line in FIG. For this reason,
In this embodiment, when calculating the coordinate values, the unexposed area N
N (n ≧ ₅ ) virtually set for each of D _{2 to} ND ₅
3) Shot areas, for example, eight virtual shot areas (hereinafter, referred to as virtual shots) SA _{L1 to} SA _L shown in FIG.
For each of A _L8 , the array coordinate values in design are known in advance (stored in the data section 64), so the array coordinate values (Fgx, Fgy) are calculated by the above calculation,
This calculation result is also stored in the storage unit 63 and the system controller 65.
And send to The wafer W shown in FIG. 8 is the same as the wafer shown in FIG.

[0049] Incidentally, by running the sample alignment as described above, the transform matrix Ag by the method of least squares, Og (parameters a _g ~f _g) array coordinate values from seeking (Fgx, Fg
A series of operations up to the calculation of y) will be simply referred to as g−
Let's call it EGA. Further, g-EGA the conversion matrix Ag, Og (parameters a _g ~f _g) and may only be calculated only array coordinate values of the eight virtual shots. By the way, the array coordinate value calculated by g-EGA is represented by (Fgx, Fgx
gy) is calculated by applying the least squares method to the calculated array coordinate values (Fgx, Fgy) and the measured array coordinate values (Mgx, Mgx).
gy) does not always match in all shot areas on the wafer.

Next, the system controller 65 steps the wafer stage WS in accordance with the previous arrangement coordinate values (Fgx, Fgy), especially the arrangement coordinate values of the virtual shots, and unexposed areas ND _{2 to} ND ₅ on the wafer W. Is exposed to a reticle pattern (FIG. 4A). At this time, the system controller 65 sets the reticle marks RMx ₁ , RM
As y ₁ only the exposure light IL is irradiated in advance to adjust the illumination field by driving the advance reticle blind 6. As a result, images of the reticle marks RMx ₁ and RMy ₁ , that is, latent images are formed on the resist layer. In forming the latent image as described above, an exposure amount of about 2 to 3 times the saturation energy of the resist to be used is applied to the resist layer. In the following, the reticle mark R attached to the virtual shot SA _L1
Each latent image mx _1, RMy ₁ is referred to as a latent image mark LX _L1, LY _L1. Although not shown in FIG. 8, the same applies to other virtual shots.

Here, when reticle marks RMx ₁ and RMy ₁ are accurately superimposed on reticle R and shot area SAij, their projected images coincide (overlap) with wafer marks MXij and MYij. In addition, reticle mark RM
The reticle marks RMx ₁ , RMy ₁ may be arranged so that the projection images of x ₁ , RMy ₁ and the wafer marks MXij, MYij are formed apart from each other by a predetermined interval. Since it is difficult to shield the area other than the reticle marks RMx ₁ and RMy ₁ using the reticle blind 6, for example, an auxiliary light shielding plate is arranged to shield the pattern area PA, and the reticle marks RMx ₂ and RMy ₂ are Light may be blocked by the blind 6. However, since in this embodiment it was decided to form a latent image in association with a virtual shot region of each of the unexposed area ND ₂ to ND _5, without shielding the region other than the reticle mark RMx _1, RMy ₁ In both cases, the entire surface of the reticle may be irradiated with the exposure light IL. At this time, the reticle mark RMx ₂ or RMx is used for the wafer mark in the shot area adjacent to the virtual shot.
y ₂ can be projected. For example, when forming a latent image with the virtual shot SA _L2 , the wafer mark MY in the shot area SA ₂₇
Reticle mark RMy ₂ is overlapping exposed to _27. For this reason, it is desirable that the reticle marks RMx ₂ and RMy ₂ be shielded from light by the reticle blind 6.

Now, as described above, the reticle mark RMx ₁ ,
When the latent image of RMy ₁ is formed, the system controller 6
Reference numeral 5 denotes an alignment sensor most suitable for detecting a latent image, in particular, an LSA system is selected in this embodiment, and a latent image mark is detected using two sets of LSA systems. First, the system controller 65
Is data (position,
) Is read out, and the wafer stage WS is stepped according to the calculated array coordinate values (Fgx, Fgy) of each virtual shot, and the X-LSA system and Y-
Each spot light of the LSA system and two sets of latent image marks (LX _L1 ,
LY _L1 ). As a result, LS
The A operation unit 57 calculates the mark position by a predetermined waveform processing, and these position information is based on the array coordinate values (Mtx,.
Mty) is sent to the alignment data storage unit 61.

In this embodiment, when detecting a latent image, a mark that cannot be measured or a mark whose measured value is extremely different from the design value is rejected, and only the remaining mark positions are used. At this time, the virtual shot itself having the bad latent image mark may be rejected. Further, among the plurality of virtual shots on the wafer W shown in FIG. 8, it was chosen eight virtual shots together can form a latent image of the reticle mark RMx ₁ and RMy ₁ in this embodiment, either one A virtual shot that can form only the latent image may be selected. Further, in preparation for the reject as described above, in addition to the eight virtual shots selected previously, for example, at least one virtual shot near the selected shot is designated as a substitute shot to form a latent image, and the mark position can be substituted. You may do it. In addition, a virtual shot in which the positions of the two sets of latent image marks are measured together is taken as one.
In this embodiment, it is necessary to detect latent images for at least three virtual shots. If only one latent image mark is measured in one virtual shot (for example, in the X direction) and if the other latent image mark is measured in the alternative shot (in the Y direction), these are regarded as one virtual shot. You may count.

Next, the EGA calculation unit 62 calculates the measured array coordinate values (Mtx, Mty) of the eight virtual shots stored in the alignment data storage unit 61 and the g-EGA.
The calculated array coordinate values (Fg
x, Fgy), based on the regularity of the arrangement of the latent images on the wafer W, that is, the transformation matrix At in the mapping relational expression (determinant Mt = At · Fg + Ot) shown in the following Expression 5.
Calculate Ot. Here, the regularity of the arrangement of the latent images corresponds to the regularity of the arrangement of the pattern images when the overlay exposure is actually performed. However, the transformation matrices At and Ot in the above relational expression include, as parameters, the remaining rotation error θ _t , the orthogonality ω _t and the scaling errors Rtx and Rty, and the offset errors Otx and Oty, and the transformation matrix At has two rows. Two columns, Ot is a matrix of two rows and one column.

[0055]

(Equation 5)

The transformation matrices At and Ot are given by the following equation (6).
It is represented by 7.

[0057]

(Equation 6)

[0058]

(Equation 7)

Here, the shot area on the wafer has residual terms (εXt, εYt) with respect to the measured array coordinate values (Mtx, Mty) and the calculated array coordinate values (Fgx, Fgy). Equation 5 is rewritten as Equation 8 below.

[0060]

(Equation 8)

[0061] Thus, EGA calculation unit 62 so that the residuals is minimized, transformation matrices At, is determined by calculating the values of the parameters a _t ~f _t of Ot (least squares method). The conversion matrices At and Ot calculated as described above are stored in the storage unit 63. Here, it is not necessary to calculate the array coordinate values of all the virtual shots on the wafer W by using Expression 5 above. The conversion matrix by the least square method from the latent image detection At, the series of operations until the calculated Ot (parameters a _t ~f _t), in the following simply referred to as t-EGA.

Now, the EGA operation unit 62 calculates g-EG
Is calculated in each of A and t-EGA, 2 sets of parameters a _g ~f _g stored in the storage unit 63 reads the a _t ~f _t. Here, the remaining rotation errors θ _g and θ _t and the orthogonality ω
_g, since the ω _t is a very small amount, g-EGA and t-EGA
The six error parameters at each of are:
It is represented by 10.

[0063]

(Equation 9)

[0064]

(Equation 10)

Next, the EGA calculation unit 62 calculates the difference between the position of the shot area on the wafer and the position of the latent image (ie, the actual exposure position of the reticle pattern to be overlaid and exposed), in other words, g-EGA. The difference (Ogx-Otx) and (Ogy-Oty) of the offset error with t-EGA are obtained,
This difference is stored in the storage unit 63 as a system offset. Further, the EGA calculation unit 62 calculates the scaling error (Rgx, Rgy) or (Rtx, Rty) based on:
The magnification error (size change amount) of the shot area on the wafer W is calculated, and this information is sent to the system controller 65. The system controller 65 adjusts the projection magnification of the projection optical system 13 using the imaging characteristic correction unit 19 so as to cancel the change in magnification of the shot area due to scaling, and adjusts the magnification error between the shot area and the projected image of the reticle pattern. Is set to almost zero.

As described above, the offset measurement before the main exposure is completed, and the system controller 65 starts the overlay exposure on the wafer W. In the present embodiment, the overlay exposure is performed by employing the EGA method based on the FIA system. However, since the sample alignment and the like have already been performed by g-EGA, all of the data on the wafer W is calculated using the following Expression 11. Are calculated (Fx, Fy). However, transform matrix Ag in Equation 11 (parameter a _g to d _g) is used as it is calculated in g-EGA. Therefore, the array coordinates (F
x, Fy) is obtained by correcting the array coordinates (Fgx, Fgy) obtained by g-EGA (Equation 1) with the offset errors (Otx, Oty) obtained by t-EGA.

[0067]

[Equation 11]

The system controller 65 steps the wafer stage WS in accordance with the calculation results (Fx, Fy) in the EGA calculation unit 62, and performs superposition exposure of the reticle pattern for each shot area. As a result, alignment errors (lateral displacement due to offset),
In addition, the magnification error can be reduced to almost zero, and high-accuracy overlay exposure is performed. Note that the wafer stage WS is calculated according to the array coordinate values (Fgx, Fgy) calculated by g-EGA without calculating the array coordinate values (Fx, Fy) from Expression 11.
May be stepped, and the position may be shifted by an offset (Otx, Oty) calculated by t-EGA for each shot area.

As described above, in this embodiment, since the system offset is calculated by applying the EGA method, high-speed and high-accuracy offset measurement is possible, and the latent image of the reticle mark can be placed at an arbitrary position on the wafer. It has the advantage that it can be formed and there are few restrictions on the place when forming a latent image. In particular, if a latent image is formed in an unexposed area, the latent image will not be disposed near the wafer mark, and the alignment accuracy in the next layer will be reduced due to a pseudo alignment mark that can be formed on the base by the latent image. The effect that it can be prevented is obtained. Furthermore, since the wafer mark (underlying mark) and the latent image mark can be detected by separate alignment sensors, that is, the optimum alignment sensor can be selected and used for detecting each mark. Becomes possible.

In this embodiment, g-EGA is performed by using the FIA system and t-EGA is performed by using the LSA system. However, in the case of the t-EGA, the type of the base and the resist, the shape of the latent image mark, etc. In this case, an optimal alignment sensor for detecting a latent image may be selected according to the conditions, and for example, an LIA system may be used. Further, the same alignment sensor may be used for g-EGA and t-EGA.
-In EGA, it is desirable to use an alignment sensor used in the main exposure. In particular, in the present embodiment, it is assumed that the alignment of the EGA method is performed at the time of the main exposure. In the main exposure, only the correction of the system offset is performed using the conversion matrix Ag calculated by g-EGA as it is. For this reason, the position and number of sample shots in g-EGA are
As in the case of normal main exposure, the arrangement coordinates of the shot area, that is, the conversion matrix Ag are determined so as to be calculated with sufficient accuracy. In other words, it is desirable that g-EGA be performed in exactly the same way as EGA performed prior to overlay exposure in a conventional apparatus.

Here, detection of a latent image using the LIA system will be briefly described with reference to FIG. In FIG. 9A, when two LIA beams BM ₁ and BM ₂ are incident on a diffraction grating wafer mark GA (pitch is 2P) at an intersection angle of 2 角₀ from two different directions, the mark GA
Above, an interference fringe IF (pitch: P) is formed by two beams. By the way, in the above apparatus (FIGS. 1 and 2), the beam BM ₁ generated along the optical axis AX of the projection optical system 13
And the -1st-order diffracted light B ₁ (-1), the interference light of the beam BM ₂ + 1-order diffracted light B ₂ (+1) is received by the photoelectric detector 52. The photoelectric detector 52 has three light receiving surfaces, that is, a first light receiving surface for receiving the interference light, and a second light receiving surface with the first light receiving surface interposed therebetween.
A third light receiving surface is provided. In particular, the second and third light receiving surfaces are for independently detecting + 1st to + 3rd order diffracted light and -1st to -3rd order diffracted light generated from the wafer mark by irradiation of the LSA system spot light.

[0072] From further wafer mark GA, ± 1-order diffracted light B ₁ (-1), with B ₂ (+1), a beam BM ₁ 0-order diffracted light B ₁ (0) and the beam BM ₂ +2 order diffracted light B ₂ (+2) are generated in the same direction, and the _second-order diffracted light B of the beam BM _1
1 (-2) and the zero-order diffracted light B ₂ (0) of the beam BM ₂ are generated in the same direction. The two interference lights are received by the second and third light receiving surfaces, and the photoelectric detector 52 also outputs a photoelectric signal obtained by adding the two. Therefore, the wafer mark (base mark)
In the measurement, the LIA operation unit 58 determines the mark position from the waveform phase difference between at least one of the two photoelectric signals from the photoelectric detector 52 and the photoelectric signal (reference signal) SR from the photoelectric detector 56. I do.

When the LIA system having the above configuration is used,
Using a latent image mark having the same pitch as the wafer mark GA, ±
Although the first-order diffracted light may be photoelectrically detected, it is more advantageous to receive the zero-order diffracted light in terms of contrast.
Therefore, in the LIA system, a latent image mark GB having a pitch P as shown in FIG. 9B is used, and at the time of position measurement, of the diffracted light generated from the mark, the beam BM ₁ generated in the same direction is used. Zero-order diffracted light B ₁ (0) and beam B
The interference light of the + _1st-order diffracted light B ₂ (+1) of M ₂ and the beam BM ₁
-1 receives the interference light of the diffracted light B ₁ (-1) and 0-order diffracted light B ₂ (0) of the beam BM _2. Here, these two interference lights are incident on the second and third light receiving surfaces of the photoelectric detector 52. In other words, by setting the pitch of the latent image marks to half of the pitch of the wafer marks, it becomes possible to improve the measurement accuracy of the latent image marks without changing the configuration of the LIA system. Therefore, when measuring the latent image mark,
The LIA operation unit 58 controls the second and third
The mark position is determined from the phase difference between the photoelectric signal from the light receiving surface and the reference signal SR.

Further, in the t-EGA in the above embodiment, the latent images are formed for eight virtual shots (FIG. 8). Should be determined. Further, when the unexposed area is small (the area is small) and the number of virtual shots can be reduced, at least one shot area is selected as a virtual shot from a plurality of shot areas on the wafer W,
The shot may be used as a discard shot for measurement (a part of the unexposed area). When a discard shot is used, it is desirable not to perform pattern exposure on the shot in the exposure step up to the previous layer. Further, it is desirable that the shot area designated as the discard shot is located on the outermost periphery on the wafer W. This is because the defect occurrence rate in the outermost shot area is relatively high.

Further, in the above embodiment, t-EGA is performed only once, but t-EGA may be performed two or more times. In this case, the average value of the measured values is used as the offset error. You can use it. At this time, the position and the number of virtual shots for performing the latent image detection are determined, for example, by one t-EG.
It may be different for each A. In the t-EGA, it is not necessary to calculate even the transformation matrix At, and the transformation matrix Ot, that is, only the offset error (Otx, Oty) may be calculated. Further, the number n of the virtual shots used in the t-EGA does not need to be the same as the number m of the sample shots used in the g-EGA, and the number and the position each correspond to the required calculation accuracy. Just do it.

In the above embodiment, the system offset (Ogx−Otx) and the first wafer in the lot are used.
Since (Ogy−Oty) was measured, it is not necessary to perform offset measurement particularly for the second and subsequent wafers, and alignment may be performed using the previous measurement value as it is. However, even for wafers in the same lot, if the type of base or resist, the type of wafer mark (shape, step, etc.) is different, or the alignment sensor is changed, the first wafer after the change It is desirable to measure the system offset in exactly the same manner as described above using. Further, even if the condition is not changed as described above, the system offset may be measured for each unit number of wafers, each lot, or each unit time, and the data stored in the storage unit 63 may be sequentially updated. good. Further, when the system offset is extremely large, it is desirable to execute again the reticle alignment, the baseline measurement, or the measurement of the imaging characteristics (magnification, distortion, etc.) of the projection optical system 13. Alternatively, the signal processing conditions (algorithm) of the alignment sensor used in g-EGA (i.e., main exposure) are changed, the alignment sensor itself is changed, the number and position of sample shots in g-EGA are changed, or t-EGA is changed. The alignment sensor to be used, its signal processing conditions, and the latent image mark forming conditions (pitch, shape, etc.) may be changed. Conversely, when the system offset is sufficiently smaller than the alignment accuracy, the array coordinates calculated by g-EGA may be used as they are in the main exposure.

Further, in the above embodiment, even in the main exposure, the EGA
Although the alignment of the method was performed, for example, TTR
Even when die-by-die alignment (DDA) is performed using an alignment sensor of the (Through The Reticle) method, the system offset is calculated using the alignment sensor in exactly the same manner as in the above embodiment,
When the DDA is executed, a system offset is used for each shot area, that is, the shot area may be shifted with respect to the projected image of the reticle pattern by the offset, and an improvement in alignment accuracy can be expected. This is particularly effective in an alignment sensor using non-exposure light as disclosed in, for example, Japanese Patent Application Laid-Open No. 4-7814.

In the above embodiment, g-EGA, tE
Among the parameters calculated by the GA, the scaling error and the offset error are used, but other errors, that is, the residual rotation error and the orthogonality are not particularly used. However, at least one of the orthogonalities ω _g and ω _t is an array coordinate value (Fx, Fx) for each shot area.
It may be used for correcting y). This is done by changing according to the transformation matrix Ag in Equation 11, i.e. the value of parameter a _g to d _g orthogonality. On the other hand, with respect to the residual rotation errors θ _g and θ _t , the array coordinate values (Fx, Fy) for each shot area may be corrected by changing the values of the parameters a _{g to} d _g according to the difference between them. good. This is different from a stepper that exposes a circuit pattern (base pattern) on the wafer W and a stepper that exposes the circuit pattern by overlaying a reticle pattern of the next layer. Are effective when are different from each other.

Here, one or both of the six parameters of the g-EGA and the six parameters of the t-EGA are used in the main exposure (Equation 11).
The EGA operation unit 62 determines whether or not to select for each parameter according to a predetermined program, or the operator specifies via an input device. In the above embodiment, at least two sets of offset errors (Ogx, Ogy) and (Otx, Oty) may be used.

Further, in the above embodiment, the reticle R (FIG. 4)
When performing the overlay exposure of the four reticle marks, in particular with the reticle mark RMx ₁ and RMy _1, it will be exposed overlaps its wafer mark MX and MY for each shot region. Also, the reticle marks RMx ₂ , R
My ₂ may also be exposed overlaps the wafer mark adjacent shot areas in the shot area to be exposed. So 4
The reticle mark formation position on the reticle R may be shifted in advance as shown in FIG. 10 so that one reticle mark does not overlap with the wafer mark. by this,
Since the area corresponding to the wafer mark on the reticle R is a light-shielding portion, particularly when a positive resist is used, the resist layer on the wafer mark is not exposed to light due to the superposition exposure, and the mark of the mark due to etching or the like is removed. Destruction will be reduced. By adopting the above arrangement, for example, an alignment mark for a TTR type alignment sensor can be formed on the reticle R separately from the reticle mark. The four reticle marks shown in FIGS. 4 and 10 are preferably arranged symmetrically with respect to the center of the reticle. In particular, considering Abbe error, the marks should be formed symmetrically with respect to the center of the reticle as shown in FIG. Is desirable.

In the above embodiment, the latent image is formed in the unexposed area using the two reticle marks RMx ₁ and RMy ₁ . If the unexposed area is small and the number of virtual shots is small, the shot area located on the outermost periphery of the wafer W is designated as a virtual shot as described above, and the shot is discarded for measurement. By doing so, the number of virtual shots can be apparently increased.
At this time, if the reticle R shown in FIG. 10 is used, the specified shot area does not need to be discarded and the yield can be prevented from lowering. Hereinafter, a brief description will be given with reference to FIG.

FIG. 11 is a partially enlarged view of the wafer shown in FIG. 11, when the shot area SA ₃₁ designated as the measurement shot calculated array coordinate values in g-EGA (Fgx, Fgy) after positioning the shot area SA ₃₁ to the exposure position according to the reticle R as shown in FIG. 10
When only the reticle marks RMx ₁ and RMy ₁ are exposed to the exposure light, the latent image marks LX ₃₁ and LY ₃₁ are formed on the wafer W. As is clear from FIG. 11, the latent image marks LX ₃₁ and LY ₃₁ are the wafer marks MX ₃₁ and MY ₃₁
, And this latent image mark is
On that can be used in -EGA, it is understood that it is not necessary to shot discarded shot area SA _31.

Here, if the formation positions of reticle marks RMx ₁ and RMy ₁ on reticle R are sufficiently shifted from the center of each side (the area corresponding to the wafer mark), the latent image on wafer W can be obtained. Marks LX ₃₁ , LY ₃₁ and wafer mark M
X ₃₁ and MY ₃₁ are formed sufficiently separated from each other.
Therefore, the latent image mark and the wafer mark can be separated from each other and detected independently by the alignment sensor, and a pseudo mark due to the latent image is not formed near the wafer mark in the next layer. Also, the reticle mark RM
Even when the latent image is formed in the shot area SA ₃₂ using x ₂ and RMy ₂ , the latent image marks LX ₃₂ and LY ₃₂ are also aligned with the wafer mark MX ₂₁ in the shot area SA ₂₁ adjacent to the shot area SA _32. wafer mark MY of the shot area SA ₃₁ and
_No overlap with each of ₃₁ .

As is clear from the above, t
-The virtual shots used in the EGA may not be set at all in the unexposed area, and all the virtual shots may be selected from a plurality of shot areas on the wafer W. At this time, the number and position of the shot areas specified as the virtual shots may be determined independently of the number and position of the sample shots specified by g-EGA. This means that the unexposed area on the wafer is small, especially depending on the pattern size.
This is effective when many virtual shots cannot be set therein.

Further, in the above description, virtual shots within one unexposed area are set apart by the shot size so that shots do not overlap each other. As described above, all or some of the virtual shots are selected from a plurality of shot areas and used. However, even if some of the virtual shots overlap in the unexposed area, the corresponding latent image marks overlap or are not formed close to each other. And the number of virtual shots decreases, ie, t-
It is possible to prevent the calculation accuracy of the conversion matrices At and Ot in the EGA from being reduced.

Therefore, the unexposed area on the wafer is small.
Alternatively, when increasing the number of virtual shots, as shown in FIG. 12, even if some of the virtual shots overlap, the latent image marks are not formed so as to be entirely or partially overlapped or formed in close proximity. A plurality of virtual shots may be set in the same unexposed area. In Figure 12, we have set a virtual shot SA _L9 apart a predetermined distance (shot size value less than) in the Y direction with respect to the virtual shot SA _L6, the latent image mark LX _L6 virtual shot SA _L6,
LY _L6 and the latent image mark L of the virtual shot SA _L9
Each of X _L9 and LY _L9 is sufficiently separated so that any mark can be used in t-EGA. Similarly, the virtual shot SA _L10 is X with respect to the virtual shot SA _L8,
The latent image mark L is set to be shifted by a predetermined amount in the Y direction.
X _L10, LY _L10 and LX _L8, LY _L8 and are not to be formed in proximity electrode.

In the above embodiment, the reticle R (FIG. 4,
Although only two reticle marks RMx ₁ and RMy _{1 in} FIG. 10) are formed as latent image marks, three or four reticle marks may be used. in this case,
By measuring the positions of three or four latent image marks for each virtual shot in t-EGA, two reticle marks RMx ₁ , RM arranged opposite to each other on reticle R
The residual rotation error θ _R of the reticle R with respect to the rectangular coordinate system XY can be obtained from the positions of the latent image marks x ₂ , RMy ₁ , and RMy ₂ . The residual rotation error θ _R may be corrected by fine movement of the reticle stage RS, but the array coordinates (Fx, Fx,
When calculating Fy), the residual rotation error θ _g of the wafer W (array coordinate system of shot areas) with respect to the orthogonal coordinate system XY,
Or corrected by using as a target value of theta _t. That is, in Equation 11, the residual rotation error of the wafer is represented by (θ
_g + θ _R ) as array coordinate values (Fx, F
y) may be calculated. Note that the residual rotation error θ
_{As R,} it is desirable to use a value obtained by averaging the measured values in each of the plurality of virtual shots. Further, when three or more reticle marks are formed as a latent image for each virtual shot, the latent image mark and the wafer mark in the shot area may overlap. Therefore, the reticle R shown in FIG. It is desirable to shift the position toward the outer periphery by at least one of the X and Y directions by a predetermined amount.

Here, in the above embodiment, since a virtual shot is set in an unexposed area on the wafer, only one or two latent image marks may be formed in one virtual shot. In such a case, all or a part of the virtual shots may be selected from a plurality of shot areas on the wafer W (FIG. 11). At this time, only the virtual shot used for calculating the residual rotation error θ _R may be selected from a plurality of shot areas.

Alternatively, as shown in FIG. 13, after forming latent images (marks LX _L8 , LY _L8 ) of two reticle marks RMx ₁ , RMy ₁ in virtual shot SA _L8 , the remaining 2
The wafer stage WS is stepped by at least one of the X and Y directions (only the Y direction in the figure) by a predetermined amount so that at least one of the two reticle marks RMx ₂ and RMy ₂ is formed as a latent image on the wafer W. , At least one of reticle marks RMx ₂ , RMy ₂ is a latent image (mark LX
_L11, formed as LY _L11). In other words, a virtual shot X, the virtual shot SA _L11 a predetermined distance in at least one of the Y direction set for the SA _L8, the reticle mark RMx ₂ in the shot SA _L11, RMy ₂
Are formed as latent images (marks LX _L11 , LY _L11 ). As a result, even for a virtual shot in which three or more reticle marks cannot be formed at a time as a latent image, it is possible to calculate the residual rotation error θ _R of the reticle R by applying the shift exposure as described above. Become.

In the shift exposure as described above, the wafer W
The latent image marks LX _L11 and LY _L11 are
Since it approaches X _L8 and LY _L8 (the interval becomes shorter than the shot size), the measurement accuracy of the residual rotation error θ _R of the reticle R may decrease. Here, the latent image mark L
If only one of X _L11 and LY _L11 is formed on the wafer, the residual rotation error θ _R of the reticle R can be calculated. Therefore, it is possible to determine the stepping amount of the wafer stage WS so that the above-mentioned interval is not shortened as much as possible. desirable.
Conversely, if there is a space in the unexposed area, the measurement accuracy can be improved by making the above-mentioned interval larger than the shot size. However, in any of the shift exposures, the wafer stage WS is stepped, so that the stepping error becomes an error factor of the measurement value (mark position), and the measurement accuracy is reduced. Therefore, while the latent image forming areas in the unexposed area are different from each other, the above-described staggered exposure is repeated a plurality of times, and the residual rotation error of the reticle R is determined from a plurality of measured values by, for example, an averaging process. It is desirable to do.

Next, an alignment method according to the second embodiment of the present invention will be described. In this embodiment, the projection optical system 13
Is different from the first embodiment in that the projection magnification is measured using a latent image. Here, only the difference will be described. FIG. 14A shows a schematic configuration of a reticle R used in the present embodiment, and is different from FIG. 10 in that two reticle marks are formed in each of four transparent windows in a light-shielding band LSB. FIG. 14 (b) two reticle marks RMax formed within the transparent window, shows an example of a RMay, here in each other are perpendicular to the array direction of the reticle mark RMx ₁ shown FIG. 4 (b) 2 Assume that a pair is formed. 2
When there is no need to distinguish between two reticle marks RMax and RMay, they will be simply referred to as reticle mark RMa.

The system controller 65 first performs the steps from loading of the reticle R (FIG. 14) to g-EGA, just like the first embodiment. G-EG of the present embodiment
In A, the transformation matrices Ag and Og and the virtual shot S in FIG.
A _{L1 to} SA _L8 and array coordinate values (Fgx, Fgy) of the virtual shot SA _{L12 in} FIG. 15 are calculated. FIG. 15 is a partially enlarged view of the wafer W shown in FIG.

Further, the system controller 65 drives the reticle blind 6 so that only the two sets of reticle marks RMa and RMb are irradiated with the exposure light IL, and then steps the wafer stage WS in accordance with the previous arrangement coordinate values, and Reticle mark RMa, RM for each shot
The latent image b is formed. At this time, as shown in FIG. 15, in the virtual shot SA _L7 , four sets of reticle marks RM are set.
a to RMd are irradiated with exposure light IL to expose the latent image (L
Ma ₇ ~LMd ₇₎ to form a. Further, in each of the three virtual shots SA _L6 , SA _L8 , and SA _L12 , after forming two sets of reticle marks RMa and RMb as latent images, the reticle is irradiated so that only the reticle marks RMc and RMd are exposed to exposure light. The blind 6 is driven to perform a shift exposure similar to that described with reference to FIG. 12, that is, the wafer stage WS
Is stepped by a predetermined amount in the X and Y directions to form a latent image.

For example, as shown in FIG. 16, in the virtual shot SA _L6 , two sets of reticle marks RMax and RMay and RMbx and RMby latent image marks LMax and LMay (FIG. 15)
LMbx and LMby (corresponding to the latent image LMa ₆ in FIG. 15)
After forming the support) to the latent image LMb ₆ in the reticle mark RMc, the exposure light IL only RMd drives the reticle blind 6 is irradiated. Next, virtual shot S
After the wafer stage WS is stepped so that A _L6 ′ is set, two sets of reticle marks RMcx and RMc are set.
y and latent image marks LMcx and LMcy of RMdx and RMdy
(Corresponding to the latent image LMc ₆ in FIG. 15), and LMdx, LMdy
Forming a (corresponding to the latent image LMd ₆ in FIG. 15). A latent image is formed by the same operation in each of the virtual shots SA _L8 and SA _L12 .

Next, the system controller 65
The wafer stage WS is moved according to the column coordinate values (Fgx, Fgy).
Stepping is performed, and wafers are set using two sets of LSA systems.
C. Measure the positions of all latent image marks formed on W
You. The EGA operation unit 62 has four virtual shots S
A_L6~ SA_L8, SA_L12Based on location information at each
The projection magnification M of the projection optical system 13 and the residual number of reticle R
Rolling error θ_RIs calculated. For example, virtual shot SA
_L6In FIG. 16, the latent image marks LMay, LMcy, or
The projection magnification is calculated from the position of LMby and LMdy in the Y direction,
X direction of image mark LMax, LMcx, or LMbx, LMdx
The residual rotation error is obtained from the direction. The remaining three virtual
For shots, the projection magnification and residual rotation error are calculated in the same manner.
Put out. Here, the residual rotation error θ_RTo the projection magnification M
Since a calculation error may occur, the residual rotation error θ _RThrow according to
The shadow magnification M is corrected in advance. From the above calculated values,
For example, by the averaging process, the projection magnification M and the residual rotation error θ
_RTo determine. This is due to the stepping error when forming the latent image.
Is effective in averaging. Thereafter, the first embodiment and
In exactly the same way, using the position of the latent image mark for each virtual shot
T-EGA to calculate the transformation matrices At and Ot.

Further, EGA operation unit 62 calculates gE
The system offsets (Ogx-Otx) and (Ogy-Oty) are determined using the calculation results of GA and t-EGA, and the array coordinates ( Fx, Fy) are calculated. Further, a magnification error (amount of change in size) of the shot area on the wafer W is calculated based on the scaling error (Rgx, Rgy) or (Rtx, Rty), and this information and the previous projection magnification M are calculated by the system. Send to controller 65. System controller 65
Adjusts the projection magnification of the projection optical system 13 using the imaging characteristic correction unit 19 so that the magnification error between the projected image of the reticle pattern and the shot area becomes zero.

Incidentally, the residual rotation error θ _R of the reticle _R
Is relative rotation of the reticle R and the wafer W (particularly, rotation of the wafer W), or the residual rotation error θ _g ,
array coordinate values for each shot area by changing the values of the parameters a _g to d _g in Equation 11 according to the difference between θ _t (Fx, F
By correcting y), a decrease in alignment accuracy due to the residual rotation error θ _R is prevented. Also, orthogonality ω _g , ω _t
Also, the array coordinate values (Fx, Fy) for each shot area are corrected by changing the values of the parameters a _{g to} d _{g in} Expression 11 using at least one of the above.

After that, the system controller 65
The wafer stage WS is stepped according to the calculation result (Fx, Fy) in the GA calculation unit 62, and the reticle pattern is superposed and exposed for each shot area. As a result, the system offset, the residual rotation error of the reticle R, and the magnification error of the projection optical system 13 (ie, the magnification error between the projected image of the reticle pattern and the shot area)
This can prevent a decrease in alignment accuracy due to the above, and perform high-accuracy overlay exposure.

As described above, in this embodiment, the EGA method can be applied to calculate not only the system offset, but also the residual rotation error of the reticle R and the projection magnification of the projection optical system. Becomes possible.
By the way, in the present embodiment, four virtual shots SA _{L6 to} S _L
Although the residual rotation error θ _R and the projection magnification M were calculated using the latent image positions at each of A _L8 and SA _L12 (FIG. 15), at least among the four sets of reticle marks RMa to RMd, One virtual shot in which latent images of three sets of reticle marks can be formed (for example, virtual shot SA _{L7 in} FIG. 15)
May be used only. However, since the unexposed area is small, if the latent images of at least three sets of reticle marks cannot be formed unless the above shift exposure is adopted, in order to remove the influence of the stepping error, a plurality of virtual shots are used. It is desirable to use measured values. Further, in this embodiment, the latent image is formed while irradiating only two sets of reticle marks with exposure light. However, except for the virtual shot for performing the shift exposure as described above, all four sets of reticle marks are exposed. Irradiation with light may form a latent image.

In the above embodiment, the latent image is formed after performing g-EGA, and the system offset and the system offset are determined by t-EGA.
Although the residual rotation error and the projection magnification are calculated, the residual rotation error may be calculated and corrected prior to g-EGA. For example, from among four sets of reticle marks RMa to RMd, at least one virtual shot capable of forming a latent image of at least three sets of reticle marks is selected, and a latent image is formed using the selected virtual shot.
The virtual shot to be selected may be in an unexposed area or may be selected from a plurality of shot areas on the wafer W. Thereafter, a residual rotation error (and a projection magnification) is calculated from the position of the previously formed latent image mark.
The rotation is corrected, for example, by rotating the wafer W so that the relative rotation error with respect to the above becomes zero. After the above operation is completed,
G-EGA, latent image formation, and t-EGA are executed to determine the system offset in exactly the same manner as in the first embodiment.
Array coordinate values (Fx, Fy) of shot areas on wafer W
Should be determined.

Next, an alignment method according to a third embodiment of the present invention will be described. In the present embodiment, an alignment method which is an improvement of the EGA method adopted in the first and second embodiments (disclosed in Japanese Patent Application No. Hei 4-10091 previously filed by the present applicant, hereinafter referred to as a P-EGA method). ) Is used. In the P-EGA method, data (coordinate position) is weighted for each sample shot in accordance with the distance between a shot area to be aligned with a predetermined reference position and each of a plurality of sample shots. The conversion matrix is determined by performing a statistical operation (least square method) using the data thus obtained, and then the array coordinate values of the shot area are calculated based on the conversion matrix. Therefore, before describing the alignment method of the present embodiment, FIG.
The P-EGA method will be briefly described with reference to FIG.

As shown in FIG. 17, in the P-EGA system,
I-th shot area E on wafer W to be aligned
Si and m (nine in the figure) sample shots SA
According to the distance L _K1 ~L _K9 between each of ₁ -SA _9, 9
Weight W _is for each of the coordinate positions of the sample shots
give. Further, similarly to the conventional EGA method, the sum of squares Ei of the residual is evaluated by the following equation, and the transformation matrices A and O (parameters a to f) are determined so that the following equation is minimized. In the P-EGA method, the number and position of the sample shots used for each shot area are the same, but the weight given to the alignment data (coordinate position) changes for each shot area. Therefore, in the P-EGA method, the parameters a to f are determined for each shot area, and the coordinate position is calculated.

[0103]

(Equation 12)

[0104] Here, in each shot area to be aligned, for changing the weight W _IS for each sample shot, the weighting W _IS as follows, i th shot area ESi and s-th to be aligned As a function of the distance L _{ks from} the sample shot SA _s . Where U
Is a parameter for changing the degree of weighting.

[0105]

(Equation 13)

[0106] As is clear from equation 13, as the sample shot distance L _ks is short of the shot area ESi be aligned, the weighting W _IS increases given to the alignment data (coordinate position). Here, when the value of the parameter U is infinity in Expression 13, the result of the statistical calculation processing is equal to the result obtained by the conventional EGA method. On the other hand, if all shot areas to be exposed are sample shots and the value of parameter U is zero, D / D
It is equal to the result obtained with the method. That is, in this embodiment, by setting the parameter U to an appropriate value, E
An intermediate effect between the GA system and the D / D system can be obtained. Particularly for a wafer having a large non-linear component, by setting the value of the parameter U small, it is possible to obtain an effect (alignment accuracy) substantially equal to that of the D / D method, and it is possible to excellently remove an alignment error due to the non-linear component. It is possible to do. When the measurement reproducibility of the alignment sensor is poor, by setting a large value of the parameter U, it is possible to obtain substantially the same effect as in the EGA method, and it is possible to reduce the alignment error by the averaging effect. Becomes That is, by appropriately changing the value of the parameter U, the effect can be changed from the EGA system to the D / D system. For various layers, depending on, for example, the magnitude of the non-linear component, the quality of measurement reproducibility of the alignment sensor, etc. Thus, the alignment can be flexibly changed, and alignment can be performed under optimum conditions for each layer.

Next, the positioning method according to the present embodiment will be described. Here, only the differences from the first embodiment will be described. Now, the system controller 65 first performs from loading of the reticle R (FIG. 4) to g-EGA, just like in the first embodiment. G-EGA of this embodiment
After measuring the coordinate positions of m sample shots, a conversion matrix (parameters a to f) is determined for each virtual shot for forming a latent image, and the array coordinate values are calculated.
Further, the system controller 65 steps the wafer stage WS in accordance with the arrangement coordinate values of the plurality of virtual shots, and forms two sets of reticle marks RMx ₁ and RMy ₁ latent images for each virtual shot.

After that, the system controller 65
The positions of all the latent image marks formed on the wafer W are measured using the set of LSA systems, and t-EGA is performed using the positions of the latent image marks for each virtual shot, just as in the first embodiment. , And transformation matrices At and Ot. TE of the present embodiment
The GA is exactly the same as that of the first embodiment, that is, a set of transformation matrices At and Ot are calculated using the conventional EGA calculation. Next, the EGA calculation unit 62 calculates the coordinate positions of all the sample shots measured by g-EGA and t-EG.
The conversion matrices A and O are determined for each shot area on the wafer W using the conversion matrix Ot calculated in A, that is, the offset errors (Otx and Oty), and their array coordinate values (Fx,
Fy) is calculated. That is, in the present embodiment, a conversion matrix is determined for each shot area using the coordinate positions of the sample shots weighted as described above, and then based on the determined conversion matrix and offset errors (Otx, Oty). The array coordinate values are calculated. As a result,
When the wafer stage WS is stepped according to the calculation result (Fx, Fy) in the EGA calculation unit 62, the projected image of the reticle pattern and the shot area are accurately overlapped and exposed.

As described above, in this embodiment, even when the P-EGA method is used in the main exposure, the system offset can be calculated by applying the EGA method, and high-speed and high-accuracy overlay exposure can be performed. By the way, in this embodiment, only the measurement and correction of the system offset have been described. For example, by using the reticle R shown in FIG.
Just as in the second embodiment, the residual rotation error of the reticle R and the projection magnification of the projection optical system can be measured, and the correction may be performed in the same manner as described above. In the present embodiment, the array coordinate values of the shot areas are calculated after t-EGA. However, the array coordinate values of all shot areas are calculated together with a plurality of virtual shots in g-EGA. When positioning the wafer stage WS according to the array coordinate values, the wafer stage WS may be shifted by an offset error calculated by t-EGA. Furthermore, P-
In the EGA method, the alignment data is weighted in accordance with the distance between the shot area and each of the plurality of sample shots. For example, the distance between the i-th shot area and the wafer center, and the The alignment data may be weighted according to the distance from the wafer center. At this time, the weight given to the sample shot in which the difference between the two distances is zero or the smallest is the largest.

In each of the embodiments described above, two sets of wafer marks are provided for each shot area. However, three or more sets of wafer marks may be provided. It may be arranged at four corners. At this time, also in the reticle R, four sets of reticle marks are formed at four corners of the pattern area, and g-g is formed using the four sets of wafer marks by the same operation as in the above embodiment.
Perform EGA. Here, when the coordinate position is determined for each sample shot by g-EGA, the coordinate position is determined (with an average value) using each position of the wafer mark formed at the four corners of each shot. Subsequently, after forming latent images of four sets of reticle marks for each of a plurality of virtual shots based on the result of g-EGA, the coordinate position of each of the virtual shots is formed at the four corners at t-EGA. The system offset is determined using each position of one latent image (for example, by averaging), and the system offset is determined based on the results of g-EGA and t-EGA as in the above-described embodiments. As described above, if the coordinate positions of the sample shots and the virtual shots are determined by using the marks formed at the four corners, the imaging characteristics (such as distortion) of the projection optical system and the expansion / contraction (scaling) of the wafer ), Etc., the maximum positional deviation amount in each shot can be significantly reduced, and the calculation accuracy of the system offset, that is, the alignment accuracy can be further improved. Furthermore, g-EGA and t-
In each of the EGA and each of the EGAs, it is desirable to obtain a rotation error and a magnification (including a distortion amount) of each shot from four sets of mark positions, whereby a rotation error (a chip rotation) and a magnification error (a distortion) for each shot area, and It is also possible to correct the imaging characteristics (projection magnification, distortion, etc.) of the projection optical system, manufacturing errors of the reticle R, and the like.
Note that the formation positions of the four sets of wafer marks and reticle marks need not be at the four corners, but since the positional deviation is the largest at the four outermost corners, it is actually necessary to provide the four corners at the four corners. This is advantageous in terms of measurement accuracy. Further, five or more sets of wafer marks or reticle marks may be provided for each shot, and the measurement accuracy and the like can be further improved. It is preferable to use the marks shown in FIG. 14B as the wafer mark and the reticle mark.

In each of the above embodiments, if the reticle R shown in FIG. 14 is used, the residual rotation error of the reticle R and the projection magnification of the projection optical system 13 can be measured in addition to the system offset. At this time, if the number of latent image marks formed by one virtual shot, that is, the number of reticle marks is increased, even the distortion of the projection optical system 13 can be measured, and these can be corrected by the imaging characteristic correction unit 19. . The imaging characteristic correction unit 19 drives at least one of a plurality of lens elements constituting the projection optical system 13.
A method in which the space between two lens elements is sealed and the pressure in the sealed space may be controlled, and the projection magnification is finely adjusted by the method, and the distortion is finely adjusted by driving the lens element or the reticle. You may do it.

[0112]

As described above, according to the present invention, since the system offset is calculated by using the latent image and the EGA method is applied, it is possible to perform high-speed and high-accuracy offset measurement. can get. Further, if the latent image mark is formed at an arbitrary position on the substrate, there is an advantage that there are few restrictions on the place when forming the latent image. In particular, if a latent image is formed in an unexposed area near the outer periphery of the substrate, the latent image is not arranged near the alignment mark (base mark) attached to the exposed area, and the latent image is formed on the base by the latent image. A decrease in alignment accuracy in the next layer due to a pseudo alignment mark that can be formed can be prevented. Furthermore, it is also possible to detect the wafer mark and the latent image mark with separate alignment sensors. In other words, since it is possible to select and use the optimal alignment sensor for detecting each mark, it is possible to improve the measurement accuracy of the system offset particularly. Becomes possible.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a projection exposure apparatus suitable for applying a positioning method according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a specific configuration of a TTL type alignment sensor in FIG. 1;

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

FIG. 4 is a diagram showing a specific configuration of a reticle used in the first embodiment.

FIG. 5 is a diagram showing a state of one shot area on a wafer.

FIG. 6 is a diagram showing an example of a specific configuration of a reference member.

FIG. 7 is a view showing a state of shot areas arranged on a wafer.

FIG. 8 is a diagram for explaining a positioning method according to the first embodiment.

FIG. 9 is a diagram illustrating a state where the LIA system is used for latent image detection.

FIG. 10 is a diagram showing a modification of the reticle used in the first embodiment.

FIG. 11 is a partially enlarged view of the wafer shown in FIG. 8;

FIG. 12 is a partially enlarged view of the wafer shown in FIG. 8;

FIG. 13 is a partially enlarged view of the wafer shown in FIG. 8;

FIG. 14 is a diagram showing a specific configuration of a reticle suitable for a positioning method according to a second embodiment of the present invention.

FIG. 15 is a partially enlarged view of the wafer shown in FIG. 8;

FIG. 16 is a view showing one virtual shot in FIG. 15;

FIG. 17 is a diagram for explaining the principle of a positioning method according to a third embodiment of the present invention.

[Explanation of symbols]

 62 EGA operation unit 63 Storage unit 65 System controller

Continuation of front page (56) References JP-A-4-115518 (JP, A) JP-A-62-291133 (JP, A) JP-A-63-299122 (JP, A) JP-A-60-160613 (JP) JP-A-3-96219 (JP, A) JP-A-3-154326 (JP, A) US Pat. No. 5,124,927 (US, A) (58) Fields studied (Int. Cl. ⁷ , DB name) H01L 21/027

Claims (19)

(57) [Claims] 1. A method for aligning each of a plurality of exposure regions arranged on a substrate with respect to a predetermined exposure position in a stationary coordinate system that defines a movement position of the substrate, the method comprising: Measuring position information of at least three sample regions in the exposure region in the stationary coordinate system;
The arrangement of the plurality of exposure regions, which is calculated by performing statistical operation on the measured position information of the plurality of sample regions.
A first step of obtaining position information of at least a plurality of desired exposure areas in the stationary coordinate system by using a plurality of first parameters related to the regularity of a column; and a step of obtaining the position information of the desired exposure areas obtained in the first step. A second step of exposing a specific mark to each of the plurality of desired exposure regions positioned at the exposure position based on position information; and the stationary coordinate system of at least three specific marks of the plurality of specific marks. the position information in internal measures, by statistically calculating the position information of the plurality of specific marks which are the measurement used to calculate the position information within said plurality of particular marks of the stationary coordinate system, the Multiple second regarding the regularity of the arrangement of specific marks
A third step of calculating a parameter; and a fourth step of determining position information for aligning each of the plurality of exposure regions with the exposure position based on the first parameter and the second parameter. And a positioning method. 2. The method according to claim 1, wherein the fourth step uses a deviation between a parameter representing an array offset of the plurality of first parameters and a parameter representing an array offset of the plurality of second parameters. 2. The position information of each of the exposure areas is determined.
The alignment method described in 1. 3. The method according to claim 2, wherein, in the fourth step, the first parameter is used as a parameter other than the parameter representing the array offset among the plurality of parameters. 4. The method according to claim 4, wherein in the fourth step, a parameter representing a residual rotation error of the plurality of first parameters and a parameter representing a residual rotation error of the plurality of second parameters are determined based on a deviation between the parameter representing a residual rotation error of the plurality of second parameters. 4. The alignment method according to claim 3, wherein the position information of each of the plurality of exposure areas is determined. 5. The alignment method according to claim 1, wherein the desired exposure area includes an unexposed area near an outer periphery of the substrate. 6. The desired exposure area includes a specific exposure area in which pattern exposure is not performed in an exposure step up to a previous layer, among the plurality of exposure areas. The positioning method according to claim 5. 7. The alignment method according to claim 6, wherein the specific exposure region is an outermost exposure region among the plurality of exposure regions on the substrate. 8. In the second step, the specific mark is
8. The alignment method according to claim 5, wherein the pattern is formed in a region where a pattern and a mark are not formed on a base on the substrate. 9. In the first step, the position information of the sample area is measured using a predetermined measurement method, and in the third step, the position information of the specific mark is calculated using the predetermined measurement method. The positioning method according to claim 1, wherein measurement is performed. 10. In the first step, position information of the sample area is measured using a predetermined measurement method, and in the third step, the identification is performed using a measurement method different from the predetermined measurement method. The position alignment method according to any one of claims 1 to 8, wherein position information of the mark is measured. 11. The method according to claim 1, wherein in the first step, position information of the sample area is measured based on an image signal obtained by imaging a substrate mark formed for each of the sample areas on the substrate. The alignment method according to any one of claims 1 to 10, wherein 12. In the first step, position information of each of the plurality of measured sample areas is determined according to a distance between an exposure area to be aligned with the exposure position and each of the plurality of sample areas. 12. The method according to claim 1, wherein a plurality of the first parameters are calculated for each of the exposure areas by weighting and statistically calculating the weighted position information of the plurality of sample areas. Or the alignment method according to claim 1. 13. A resist is provided on the substrate. 13. The method according to claim 1, wherein in the second step, a latent image of the specific mark is formed on the resist by applying an exposure amount equal to or more than the saturation energy of the resist to the resist. Item 2. The alignment method according to Item 1. 14. The method according to claim 1, wherein
Based on the position information of each of the plurality of exposure regions determined by using the alignment method described in section, the plurality of exposure regions are sequentially aligned with the exposure position, and are aligned with the exposure position An exposure method, wherein a pattern formed on a mask is transferred onto an exposure area via a projection optical system. 15. The projection optical system using one of a parameter representing a scaling error of the plurality of first parameters and a parameter representing a scaling error of the plurality of second parameters. The exposure method according to claim 14, wherein the projection magnification is adjusted. 16. The exposure method according to claim 14, wherein in the first step, position information of the sample area is measured without passing through the projection optical system. 17. The exposure method according to claim 14, wherein, in the third step, position information of the specific mark is measured via the projection optical system. 18. calculating at least one of a projection magnification of the projection optical system and a residual rotation error of the mask based on position information of the plurality of specific marks measured in the third step. The exposure method according to any one of claims 14 to 17, wherein: 19. A device manufacturing method, comprising: exposing the substrate with the device pattern by using the exposure method according to claim 14. Description: