TWI377598B - - Google Patents

Description

1377598 IX. Description of the Invention: [Technical Field] The present invention relates to a position detecting method, an exposure method, a position detecting device, an exposure device, and a manufacturing method of a member, and further details t, relating to detecting an object a position detecting method for configuring position information of a plurality of divided areas, using an exposure method of the position detecting method, a position detecting device for detecting position information of a plurality of divided areas arranged on the object, and using the position detecting device for exposure A device, and a method of manufacturing an element using the exposure method and the exposure device. [Prior Art] In recent years, the process of components such as semiconductor components has been performed using an exposure apparatus such as a step & repeat method or a step & scan method, a wafer probe, or a thunder. Shooting repair devices, etc. These devices shall be provided with a plurality of illumination areas on the substrate j, and a reference point in the stationary coordinate system (ie, the orthogonal coordinate system defined by the 'α laser interferometer) for specifying the substrate movement position. (For example, processing points of various devices) perform extremely precise alignment. ... especially for exposure devices, when the substrate (semiconductor wafer or glass plate) is aligned with the projection position of the reticle or reticle (hereinafter collectively referred to as "the reticle") In order to prevent a decrease in yield due to defects in the wafer at the manufacturing stage, it is desirable to maintain the accuracy of the alignment with high precision and stability at all times. ° The exposure process is to superimpose and transfer 10 or more circuit patterns on the wafer (the yoke is given to u + (the reticle pattern). If the overlap accuracy between the layers is poor, the characteristics of 1377598: The wafer will not be able to meet the characteristics of the defect. In the worst case, the wafer becomes a defective product and the yield is reduced. Therefore, in the exposure process, a plurality of illumination areas on the wafer are pre-labeled and detected. The 4 marks (coordinate values) of the mark on the pedestal W. Then, based on the information of the mark and the position information of the known reticle (predetermined in advance), the i illumination areas on the wafer are aligned to the mark. Wafer pattern wafer alignment. Wafer alignment can be broadly divided into two ways 'one is to detect the alignment marks on each of the illuminated areas on the wafer for positional alignment between the wafers (d/D die by -die) alignment mode. Another type is to detect the alignment marks of several illumination areas on the wafer to determine the alignment regularity of the illumination area, and to perform the full wafer pair alignment of each illumination area. Quasi-method witnessing in component manufacturing The combination of consideration and productivity is mainly based on full-wafer alignment. In particular, now, in the full-wafer alignment mode, especially the statistically precise specification of the entire array of exposed areas on the wafer The wafer alignment (EGA) method is mainstream (for example, refer to Patent Literature 1, Patent Document 2, etc.). The EGA method first performs a plurality of exposure exposure areas on the wafer (hereinafter, also referred to as "sampled illumination (sample sh 〇t)") the position coordinates on the pedestal coordinate system. The statistical operation processing (for example, the least square method) on the regression analysis is used, and the real side value and the arrangement of the area to be irradiated on the wafer are specified. The matching coordinates of the position coordinates obtained when converting the design coordinates of the design to the position coordinates on the carrier coordinate system are as small as possible on the coordinate system (3 or more, generally 7 to 15 or so) The method of 1377598 is used to determine the error parameters such as scaling (sca丨ing), rotation, and offset used in the conversion, and the regression mode according to the value of the obtained error parameter is The position coordinates of each of the irradiation regions in the stage coordinates are calculated. In order to meet the height requirement of the overlapping accuracy of the irradiation region due to the recent miniaturization of the element pattern, various improvements have been proposed in the EGA method. For example, as the above error parameter In the predetermined statistical mode, there is a method of using a regression mode in which the internal components of the alignment coordinate and the rotation of the relative alignment coordinate system in the irradiation region are also considered as parameters (see Patent Document 3), and the irradiation region on the wafer. The high-order component of the arrangement is also a method of considering the regression mode as a parameter (see Patent Document 4). That is, in order to improve the superposition accuracy of the irradiation region, the regression mode in the EGA method increases the parameter degree of freedom. Very complicated. However, in the regression analysis of the EGA method, the remainder of the measurement data from the regression mode (learning error) can be reduced indefinitely by increasing the degree of parameter freedom. However, in general, the measurement data contains the impurity sfl component caused by the measurement error. If the above regression analysis excessively increases the parameter freedom, the noise component will also be matched and become over-matched (over-learning). The situation arises in which a mode different from the true mode is selected as the most similar mode. In the EGA mode, the method of predicting the position of all the illumination areas from a limited sampling 'when an excessively large regression mode with unnecessary parameter degrees of freedom is selected' will likely cause the expected position and actuality of the illumination area other than the sampled illumination. The position produces a large deviation. According to the above description, it should not be minimized in the EGA mode, but the above-mentioned learning error should be considered, and the mode error (generalized 1377598 error) should be considered in all the illumination areas. In order to reduce this generalization error, the mode must be appropriately set. The parameter degree of freedom and the number of samples. In the EGA mode, the greater the degree of parameter freedom of the regression mode, the more the number of sampled shots used to match the regression pattern with good precision is increased. The measurement time becomes longer, resulting in a decrease in productivity. [Patent Document 1] Japanese Laid-Open Patent Publication No. JP-A No. Hei. No. Hei. No. Hei. No. Hei. No. Hei. SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and a first object thereof is to provide a position detection capable of detecting position information of a plurality of division regions on an object with good precision or in a short time with good precision. method. Further, a second object of the present invention is to provide an exposure method capable of forming a pattern with good precision or in a short time and with good precision in a plurality of division regions on an object. Further, a third object of the present invention is to provide a position detecting device capable of detecting positional information of a plurality of division regions on an object with good precision or in a short time and with good precision. Further, a fourth object of the present invention is to provide an exposure apparatus capable of forming a pattern on a plurality of divisional regions on an object with good precision or in a short time and with good precision. Provided is a method for manufacturing a component capable of producing a component of 1377598. The invention of claim 1 is a position detecting method for detecting position information of a plurality of divisional regions (SAp) disposed on an object (w), characterized by 'including: a third process, which is based on an established coordinate system The measured value of the position information of at least a part of the plurality of divided areas determines a pattern related to the arrangement of the plurality of area difficult areas; and the second process calculates the plurality of divided areas according to the determined mode Location information. According to the invention, in the first process, based on the measured value of the position information of at least a part of the cut area, I determines a mode related to the arrangement of the plurality of divided areas, and the second process calculates the plural according to the determined mode. Location information of the zoning area. That is, it is possible to set the information according to the actual zone area 4, and to determine the mode close to the true mode, so that the position information of the plurality of zone regions can be calculated with good precision according to the mode. At this time, as in the position detecting method of claim 2, the third process may include determining a process for measuring the location area of the location information based on the location resource m of the at least a part of the zone. At this time, as in the position detecting method of claim 3, the first process determines the degree of freedom of the parameter of the mode based on the measured value of the position information of the at least part of the zoned area and the parameter according to the mode of the determining The degree of freedom' determines the zone of the additional measurement. The invention of claim 4 is a position detecting method for detecting a plurality of divisional regions (SPa) disposed on the object (w)

The levy consists of: the first process, which uses the measured values of the position information of several zoning areas in the zonal area of the complex coordinate system, and the plurality of zonings with different degrees of freedom of the parameters respectively. The regression pattern of the region 'calculates the value of the predetermined evaluation specification of the model, and based on the calculation result, selects the mode determined to be the best in the evaluation specification as the regression mode related to the plurality of division regions; and the second process, The position information of the plurality of division regions is calculated according to the regression mode of the selection. According to the invention, in the first process, according to the predetermined evaluation specifications related to the plurality of regression modes in which the parameter degrees of freedom are different, the regression mode having a small generalization error associated with the plurality of division regions on the object is selected as the most The right mode. According to this, since the mode closest to the true mode is selected, the position information of the plurality of division regions can be calculated with good precision in the second process. For example, the position detecting method of the item 5 may further include a third process, and the process reflects the parameter value of the regression mode selected as the optimal mode in the evaluation specification to the coefficient used to determine the regression mode. Related pre-knowledge; forming a plurality of the objects in the plurality of zoning regions, sequentially changing the object to be the position information detection target, and repeating the first process, the second process, and the third process. In the present specification, the term "r-related knowledge relating to the coefficient of the predetermined regression mode" refers to a coefficient value including a regression model obtained in the past, and information on the system and the number obtained beforehand. X, the regression model obtained in the past, the coefficient value, including the regression analysis of the measured position information of each zone on the object, the regression analysis according to the simulation, and other means obtained by various means. 1377598 At this time, as in the position detecting method of claim 6, the number of samples of the real side value of the 香#香和amp; field^ is updated in the respective zoning areas, so that the evaluation is the best as the evaluation specification. The mode is chosen to be a sufficient number of correctness levels. The position detection method of the confession mode is as follows: [If the position of the request item 7 = the law: the first process, the parameter degree of freedom can be the first mode of the predetermined size: 2, the parameter degree of freedom is less than (4) the 2nd of the 1 regression mode In the regression, choose the best mode of the evaluation specification, the oral return mode test method 'the i-th process, you can check the degree of freedom of the Dan dream number with the parameter from the second item / position to the i = ΓΓ price of the established size Compared with the value of the specification, the parameter degree of freedom is smaller than the first, second mode... The value of the evaluation specification of the regression mode is relatively good: the value of the evaluation specification in the = (four) 2 regression mode as the best mode mode of the evaluation specification Compared with the second regression = the formula is good, and the generality of the i-th regression mode is better than the value; ^ select the first regression mode as the best mode of the evaluation specification; good first (four) The value of the evaluation norm is better than the established value of the second regression model. The degree of freedom of the parameter is greater than the predetermined value, and the 6-leaf system is the best mode of the evaluation specification. The third regression mode of the response type is as follows. In the position detection method of claim 8, the ι process can be selected as the zoning of the measurement target in the case where the regression mode is selected as the best mode of the evaluation specification. The area '俾 俾 取样 该 S S S S S S S S S S S S S S S S S S S S S S S S S S S 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 〜 Parameter value. The third regression mode is different. At this time, the request item regression mode and the value of the third evaluation specification are the most optimal mode. In the position detection method of 10, when a plurality of the third regression modes are selected, a good third regression mode can be selected as the position detection method for the evaluation of any of the above items 7 to 10. The coefficient of the regression mode predicted to have a large variation in the value between the objects P ^ MM i > can be used as a parameter in the 1, 2, and 3 regression modes; the other aspect will be predicted in the object. The coefficient of the regression mode with a small change in value (4) The third regression mode command is used as a parameter, and on the other hand, the second regression mode is used as a constant determined according to the prior knowledge; The coefficient predicted to be substantially the same regression mode is a constant determined based on the prior knowledge in the second regression mode, and is used as a parameter in the third regression mode. At this time, in the position detecting method of the request item U, when the number of processing units of the processing unit that is the object of the position detection is less than the Jf shape of the number of untwisted numbers, it is predicted to be between the processing units. The coefficient having a large change in value is used as a parameter in the first and third regression modes, and the second regression mode is a constant determined based on the prior knowledge, and the number of processes exceeds a predetermined number. In this case, the coefficient is used as the constant determined by the prior knowledge in the second port return mode, and the other is used as the parameter in the third regression mode. 12 1377598 The invention of claim 13 is a method for position detection of a state of the art, which is to detect the information of a plurality of zoning areas arranged on a plurality of objects, and sequentially detect each object, wherein: When the coefficient of the mode associated with the plurality of division regions is estimated for each object, the process of estimating the coefficient of the target object is estimated based on the estimation result of the object estimated by the coefficient of the mode. According to the invention, since the coefficient of the mode associated with the plurality of division regions arranged on the object is estimated based on the estimation result of the coefficient estimation of the mode in the past, the zone can be detected in a short time with good precision. Location information. The invention of claim 14 is a position detecting method for sequentially detecting the position information of a plurality of divided regions arranged on a plurality of objects, and wherein the method includes: the third process is performed according to each time. The pre-knowledge related to the coefficient of the back-mode associated with the plurality of zoning regions obtained by sequentially detecting the position information of the object, and the plurality of zonings in the plurality of zoning regions as the object of detecting the position information in the predetermined coordinate system The location information of the region is measured and the degree of freedom of the parameter of the regression mode is optimized according to the established evaluation specification of the regression mode; the second process is based on a regression mode having the degree of freedom of the optimized parameter. The position information of the plurality of division areas of the object to be detected by the position information is calculated; and the third process determines the object to be the object of the next position information detection based on the degree of freedom of the parameter to be optimized. The number of samples of the measured values of the location information of each zone; the plurality of objects forming the plurality of zones are formed, The order change is the object of the position information detection. The first process, the second process, and the 31st object according to the invention are used to sequentially change the object to be detected as the object to be detected. The first process, the second process, and the third process are repeatedly implemented. And in the first t帛+, the prior knowledge related to the coefficient of the regression mode related to the plurality of division regions obtained by sequentially detecting the position information of the object in sequence: and the predetermined coordinate system is used as the object of the position information detection The measured values of the position information of several zoning areas in a plurality of zoning areas are based on the established evaluation specifications of the regression mode, so that the degree of freedom of the parameters of the regression mode is maximized in the third process, and the degree of freedom according to the optimized parameters To determine the number of samples of the measured value of the position information of each of the division areas of the object to be detected by the next position information. In this way, since it is possible to dynamically change the parameter to be optimized and dynamically change the effective number of samples of the positional information of each zone, it is possible to increase the productivity while maintaining high-precision exposure. In the position detecting method according to any one of the above items 4 to 14, the position detecting method of the item 15 may be obtained according to the predetermined coordinate system obtained from the regression mode. The location information of each zone and the remainder of the measured value of the location information are estimated. In this case, in the position detecting method of the request item 16, the parameter of each of the regression modes may include a parameter indicating the deviation of the predetermined coordinate system from the alignment coordinate system defined by the arrangement of the plurality of division regions. The coefficient, a coefficient showing a design value deviating from a component of each of the zone regions, and at least a portion of a coefficient of a higher-order component relating to the arrangement of the plurality of zone regions. ^ 1 川如Request item P... In the detection method, there is compensation Γ, detection method, the evaluation specification can be set as the criterion of adding the degree of freedom of the household (7), and the compensation system is used for the parameter of the corresponding mode. In the New Moon, the so-called "Penalty" is used to indicate the composition of the direction of the regression mode. Therefore, if the degree of freedom of force is set to increase or decrease the "compensation", the mode can be corrected; the deviation of 疋 (the tendency to select a mode with a large degree of parameter freedom) is

: It is caused by the learning error, and you can choose to make the generalized error Xiaoji pepper as the most suitable mode. At this time, the position detection method of the request 18 can be used as a shirt. Any one of the pool information criterion and the Bayesian information criterion; the value pattern is used as a good mode, or the position detection method T is requested to set the compensation value to be increased relative to the regression mode parameter. The expected value of the change in the degree of prominence and the product of the degree of freedom of the regression mode parameter are used as a good mode with a mode having a large value.

The invention of claim 20 is an exposure method for sequentially exposing a plurality of zoning regions disposed on an object to form a predetermined pattern in each zoning region, characterized by comprising: any one of claims 1 to a a position detecting method of the item to detect a position information of the plurality of divided areas; and a process of moving the object to expose the respective divided areas according to the detected position information. At this time, since the position detection method of any one of the request items 1 to 19 is detected, α detects the position information of the plurality of division areas with high precision, so that high-precision overlapping exposure can be realized, or in a short time 15 1377598 and high precision Overlapping exposure. "Monthly Item 21 is a component manufacturer&> comprising a lithography process, characterized in that the lithography process uses an exposure method requesting $ 。. At this time, since the exposure method of requesting $ 20 is used Exposure, and the ability to achieve the southern precision of the warm umbrella. 璺 exposure, or in a short time and high-precision overlap exposure, so can improve the productivity of components with high integrated body. μ Item 22 of the b month, a position detecting device for detecting position information of a plurality of zone regions (SAp) disposed on the object (), wherein the measuring device (AS) is configured to measure the plurality of zone regions in the predetermined coordinate system Very C & - position information of the Q-zone; selection device (2〇) ' is the measured value of the position information of each zone area measured by the measurement, and several different degrees of parameter degrees are respectively related to the complex number The regression pattern of the zoning area ^ calculates the value of the established evaluation specification of the mode, and selects the best mode in the specification according to the regression pattern of the plurality of zoning regions; and the computing device ( 2〇) Calculating the position information of the plurality of divisional regions according to the selected returning chess style. According to the invention, the selection device is configured to perform a regression mode related to the plurality of divisional regions according to the degree of parameter freedom. The logarithmic approximation of the complement is used to select the best mode. According to this, since the mode that is closest to the pin mode is selected, it can be used to calculate the device and count the number with good precision. Position information of the zone; at this time, the position detecting device of claim 23 may be equipped with a mode=memory device for storing prior knowledge related to the coefficient of the read formula; And updating the device, the parameter of the regression mode selected as the best mode in the evaluation specification is reflected to the prior knowledge stored by the memory device; the selection is stored in the prior knowledge stored by the memory device, and the rotation is set The mode /, - is stored as the position detecting device of the requester 4, the memory device can be enemies:: the device for updating the measured value of the position information of each zone area measured by the device And updating, in the evaluation specification, the number of samples of the location values of the respective zone regions stored in the memory device as the best mode selected as the best mode. The invention of claim 25 is a position detecting device. The positional information of the plurality of divisional regions arranged on the plurality of objects is sequentially detected for each object. The characteristic device is that: the optimization device is based on the positional information of each of the objects in sequence. The pre-knowledge related to the coefficient of the regression mode associated with the plurality of zoning regions, and the information of the plurality of zoning regions in the plurality of zoning regions in the predetermined coordinate system as the object of the location information detection, according to the regression 帛a predetermined evaluation specification for optimizing the degree of freedom of the parameter of the regression mode; and the calculating means calculates the object to be the object of the position information detection based on the regression mode having the degree of freedom of the optimized parameter Position information of a plurality of zoning areas; and determining means are determined according to the degree of freedom of the optimized parameter The next number of samples of the position of each of the divisional areas of the object of the detection target location information is found. According to the invention, the 'optimization device' is based on the prior knowledge related to the coefficient of the regression mode associated with the plurality of division regions obtained by sequentially detecting the position information of the object, and the predetermined coordinate system as the position information 1377598 The measured position information of a plurality of the zoning areas in the plurality of zoning areas of the object is detected, and the degree of freedom of the parameters of the regression mode is optimized by the degree of compensation (parameter degree of freedom according to the regression mode). The determining means determines the number of samples of the measured position information of each of the divided areas of the object to be detected by the next position information based on the degree of freedom of the parameter to be optimized. When the device is used to process a plurality of objects in sequence, since the effective number of samples of the positional information of each zone region can be increased or decreased according to the parameter degree of freedom optimized for each object, high-precision exposure can be maintained. In the case of productivity. The invention of claim 26 is an exposure apparatus for sequentially exposing a plurality of divisional regions disposed on an object to form a predetermined pattern in each of the divisional regions, characterized by comprising: position detecting means, requesting item 22 to And the transfer device moves the object based on the detected position information to expose the respective zone regions. According to the position detecting device of any one of the claims 22 to 25, the position information of the plurality of divided regions is detected with accuracy, so that high-precision exposure or short-time and high-precision overlapping exposure can be realized. The invention of claim 27 is a method for manufacturing a component, comprising a lithography process, characterized in that the lithography process uses the exposure device 1 of claim 26 "At this time, exposure is performed by using the exposure device of claim 20 The high-precision exposure or the short-time and high-precision overlap exposure can be realized, so that the productivity of the high-accumulation component can be improved. [Embodiment] The illumination system 10' is illuminated by the illumination light IL with substantially uniform illumination. 18 1377598 Illumination is drawn with a slit-shaped illumination area defined by the reticle on the reticle R of the circuit pattern, etc. (the elongated rectangular illumination area in the X-axis direction is used here as the illumination light IL) KrF is a vacuum ultraviolet light such as far ultraviolet light, ArF excimer laser light (wavelength i93nm), or F2 laser light (wavelength 157nm) such as non-sub-field laser light (wavelength 248nm). The bright line (g line, twist line, etc.) of the ultraviolet light of the mercury lamp is used as the illumination light IL. On the aforementioned reticle stage rST, for example, the reticle R is fixed by vacuum suction. The reticle stage RST By line A reticle stage driving unit (not shown), which is a driving source, such as a motor or a voice coil motor, is immersed in a χ γ plane perpendicular to the optical axis of the illumination system 10 (which coincides with the optical axis 后 of the projection optical system PL to be described later). Driven, and can be driven at a specified scanning speed in a given scanning direction (here, in Figure 1 and in Figure 1.) The position of the reticle stage RST in the moving surface of the stage is ray laser interference The meter 16 (hereinafter, simply referred to as "the reticle interferometer") is detected by the moving mirror 15, for example, at a resolution of 〇5 to 1 nm. Here, actually, on the reticle stage RST a moving mirror having a reflecting surface orthogonal to the γ-axis direction and a moving mirror having a reflecting surface orthogonal to the X-axis direction are provided, and the moving mirror is provided with a reticle γ interferometer and a reticle X interference The apparatus 'is representatively shown in FIG. 1 as the moving mirror 15 and the reticle interferometer 16. Further, for example, the end surface of the reticle stage RST is mirror-finished to form a reflecting surface (corresponding to the moving mirror 15) Reflective surface) is also available. In addition, it can also replace the reticle stage RST At least one corner mirror (for example, a retroreflector) is used to detect the position of the scanning direction (the Y-axis direction in the present embodiment) and the surface of the reverse 1377598 extending in the x-axis direction. 'At least one of the reticle gamma interferometer and the reticle X interferometer', for example, a reticle gamma interferometer has a 2-axis interferometer with 2 length measuring axes. According to the measured value of the reticle gamma interferometer, In addition to the Y position of the reticle stage RST, the amount of rotation (the amount of deflection) in the θ z direction (the direction of rotation about the Z axis) can also be measured. The position information of the reticle stage RST from the reticle interferometer 16 (Rotation information including a deflection amount or the like) is supplied to the stage control device 19 and supplied to the main control device 20 through the stage control device 19. The stage control device 19 performs drive control of the reticle stage RST via the reticle stage drive unit (not shown) based on the position information from the stage control unit 19 in response to an instruction from the main unit 2〇. Above the reticle R, a pair of reticle alignment detecting systems 22 are disposed at a predetermined distance in the X-axis direction (although the reticle alignment detecting system 22 on the inside of the drawing is not shown in Fig. 1). Each of the reticle alignment detection systems 22, although not shown here, includes an illumination system that uses the illumination light of the same wavelength as the illumination light IL, and an indication of the monthly inspection/target object, and the A detection system that captures a mark image of the detection object. The detection system includes an imaging optical system and a photographic element, and the photographic results of the detection system (i.e., the detection results of the markings of the reticle alignment detection system 22) are supplied to the main control system 20. At this time, the deflecting mirror system for guiding the detection light from the reticle R to the reticle alignment detecting system 22 is configured to be freely movable, and when the exposure program starts, according to the main control device 2 In response to the command, the deflecting mirror is separated from the optical path of the illumination light IL by the unillustrated driving device and the reticle alignment detecting system 22, respectively. 20 1377598 The projection optical system PL is disposed below the reticle stage RST in Fig. 1, and its optical axis AX direction is the z-axis direction. As the projection optical system PL, a refractive optical system having a telecentricity on both sides and having a predetermined reduction ratio (for example, [/$, or 1/4) is used. Therefore, when the illumination region of the reticle R is irradiated with the illumination light IL from the illumination system, that is, the reduction line (partial inverted image) of the illumination region portion of the circuit pattern of the reticle R is projected through the projection optical system PL. The projection area in the field of view of the projection optical system conjugated to the illumination area on the wafer W is transferred to the photoresist layer on the surface of the wafer. The wafer stage WST is disposed below the drawing i of the projection optical system PL, and is disposed on a susceptor (not shown). The wafer holder WST is loaded on the wafer stage WST. The wafer W is fixed to the wafer holder 25 by, for example, vacuum suction or the like. The wafer stage WST can be driven by the wafer stage driving unit 24 in the X, γ, ζ, θ ζ directions (rotation directions around the z axis), the θ χ direction (rotation direction around the X axis), and A single stage in the direction of 6 degrees of freedom in the θγ direction (the direction of rotation about the γ axis). Further, in the remaining θζ direction, the wafer stage WST (specifically, the wafer holder 25) can be configured to be rotatable, and the deflection error of the wafer stage WST can be rotated by the reticle stage RST side. It can be corrected. The side surface of the wafer stage WST is fixed to reflect the laser beam from the wafer laser interferometer (hereinafter, simply referred to as "wafer interferometer") 18, by means of a mirror arranged on the outside. The circular interferometer 丨8 detects the X direction, the Y direction, and the θ ζ direction (the rotation direction around the Z axis) of the wafer stage WST at any time, for example, with a resolution of 〜·5 to 1 nm. Here, actually, The wafer stage WST is provided with an X-moving mirror (having a reflecting surface that is orthogonal to the x-direction in 21 1377598) and a gamma moving mirror (a reflecting surface having a direction orthogonal to the x-direction), and the wafer interferometer There is also an X-axis interferometer and a x-axis interferometer for illuminating the laser beam to measure the X-ray axis of the wafer stage WST, the Λ-axis direction, and the Υ-axis interferometer. In this embodiment, the 'X-axis and The γ-axis interferometer is composed of a multi-axis interferometer with a plurality of length measuring axes. In addition to the χ position of the wafer stage WST, it can also measure the rotation (yawing (02 rotation around the axis of rotation)) and the amount of vertical movement ( Pitching (the θ χ rotation around the X axis), and the amount of lateral motion (rolling ( 0y rotation of the x-axis rotation)) _ Although a plurality of wafer interferometers and moving mirrors are provided as described above, a representative display moving mirror 17 and a wafer interferometer 18 are shown in Fig. 1. For example, the end surface of the wafer stage WST may be mirror-finished to form a reflecting surface (corresponding to the reflecting surface of the moving mirror). Further, the reference label board FM is fixed in the vicinity of the wafer w on the wafer stage WST. The reference mark plate FM surface is set to be the same height as the wafer surface, and the surface is formed with at least a pair of reticle alignment reference marks and a reference mark for alignment line measurement of the alignment detecting system AS. The system AS is disposed in an off-axis mode alignment sensing on the side of the projection optical system pL. As the alignment detecting system AS, an image processing type FIA (Field Image Alignment)-based sensor is used, for example, The target mark illumination does not cause the detection beam of the wide-frequency detection of the wafer i < silk sense & and the image of the object mark imaged on the light-receiving surface by the reflected light from the target mark using a photographic element (CCD) or the like, and not shown Indicator The image is output and the image is output. In addition, it is of course not limited to the fia system, 22 1377598 is also 0 °. 蜀 or 适田 combination use the same mark on the object mark ((3) ^ coffee) detection light debt test from 4 object mark occurs Alignment sensor that detects scattered light or diffracted light, or interferes with two diffracted lights (for example, the same number of times) from the object mark. The photographic result of the alignment detection system as is not shown. The alignment signal processing system outputs to the main control unit 2. The control system is mainly composed of the main control unit 20 of the _1 and the stage control unit 19 of the subordinates. The main control device 20 includes a so-called microcomputer (or workstation) composed of a CPU (Central Processing Unit), a ROM (Read Only Memory), and a ram (Random Access Memory), and integrates the entire control device. The main control device 2G' is externally connected with an input device such as a keyboard, a mouse, etc., a display device such as a CRT display (or a liquid crystal display), and a CD (C〇mpact Disc), and a Ningca Disc). The drive device 46 of the information recording medium such as MO (Megneto-0pticle disc) or 叩 (7) (4) (Dis), and the memory device composed of the hard disk are set in the information recording medium (hereinafter, assumed to be cd) of the drive device 46, A program for processing a wafer alignment and exposure operation as shown in the flowchart below (hereinafter, for convenience, a "specific program"), other programs, and a database attached to the programs are recorded. The main control device 20 performs processing according to the specific program described above, for example, controlling synchronous scanning of the reticle R and the wafer W, stepping of the wafer w, etc., so that the exposure operation can be performed correctly. For example, when performing scanning exposure, the main control device 20 transmits a measurement value of 23 1377598 according to the reticle > 1 jammer 16 and the wafer interferometer 分别8 through a reticle stage driving unit (not shown) and a wafer carrier. station The moving portion 24 controls the position and speed of the reticle stage RST and the wafer stage WST, respectively, so that the wafer W and the reticle R can pass through the reticle stage RST in the + γ direction (or a Y direction). The scanning synchronization of the speed VR=V, through the wafer stage WST, the projection area of the vehicle and the illumination area are in a γ direction (or + γ direction) at a speed Vw=cold·v (/5 is a reticle) The board R scans the projection magnification of the wafer w. Further, during stepping, the main control unit 20 controls the position of the wafer stage WST through the wafer stage driving unit 24 based on the measured value of the wafer interferometer 18. Further, the exposure apparatus 1 of the present embodiment includes a multi-point focus detection system of an oblique incident type, and the multi-point focus detection system is composed of an illumination system and a light receiving system (not shown), and the illumination system is for projection optics. The optimal imaging surface of the system PL· and the optical axis AX direction illuminate the imaging beam for forming a plurality of slit images from the oblique direction, and the light receiving system transmits the imaging beam to the surface of the wafer w through the slits respectively. Reflected beam. As this multi-point focus detection system, for example The same configuration as that disclosed in Japanese Laid-Open Patent Publication No. Hei 6-283403, etc. is used, and the output of the multi-point focus detection system is supplied to the main control system 20. The main control system 20 is based on the multipoint. The wafer position information of the focus detection system is driven by the stage control device 19 and the turn table driving unit 24 to drive the wafer stage WST in the Z direction and the tilt direction. Next, the present embodiment is configured as described above. The operation of the exposure apparatus 曝光 the exposure processing of the second layer after the second layer is performed. The processing steps of the CPU in the main control unit 24 " 598 20 will be described based on the flowcharts of FIGS. 3 to 5 . (Based on the wafer # of Figure 2 above and the implementation of the above specific program). As a premise, it is assumed that a specific program and other programs set in the CD-R of the drive device 46 are installed in the memory device 47. Moreover, the program for the alignment of the reticle and the processing of the reference line f is taken from the memory device 47 by the CPU inside the main control device 20 to the memory of the main control rupture. After the second layer, the layers are On the wafer of the exposure target, as shown in FIG. 2, a plurality of (for example, N) irradiation regions SAP (P=1' 2, 3...N) in the processing steps up to the front layer are interposed. A street line having a width of 1 〇〇 (four) between adjacent irradiation regions is arranged in a matrix, and a wafer alignment mark for two-dimensional position detection is formed at four corners (spaced lines) of each irradiation region SAP. (wafer mark) ΜΡ, κ (Κ = 1, 2, 3, 4) The axis of the alignment coordinate system defined by the arrangement of the irradiated area SAP is the ^ axis (axis substantially parallel to the x-axis), no axis (An axis substantially parallel to the x-axis), assuming an average of the α-axis and the 乂-axis, and the yS-axis and the Υ-axis, the average position of the wafer marks Μρ丨,,, μρ,3, μρ,4 The value is the same as the gamma coordinates of the illumination area sAp (the center '^). That is, 'design, can be marked by each wafer ΜΡ, Κ The X position and the γ position are used to find the position coordinates of the center of the irradiation area sa〆. In this case, 'the wafer mark Μρκ, for example, the U-axis with the X-axis and the axis of the X-axis. The cross mark formed by the line of extension. As such mark 'in addition to the cross mark, it may be a box-shaped rod to eat, or an L/s pattern in which the direction of the α-axis direction is arranged _ &: like 25 1377598 pattern) And a combination of the s pattern in which the direction of the axis is not aligned. In addition, the information related to the irradiation area on the wafer W (the number of irradiations, the size of the irradiation, the arrangement, the arrangement of the alignment marks, the type, etc.) The shot map data is downloaded from the main computer of the lithography system to the memory device 47. As shown in Fig. 3, first, in step 3〇1, the reticle feeder is not shown. Loading the reticle onto the reticle stage RST. After the loading of the reticle is completed, in steps 303-305, the main control unit 2 (more correctly, the 'system CPU) is based on the aforementioned reticle pair Quasi- and baseline measurement processing program 'in the following manner The reticle alignment and the reference line measurement are performed. That is, the main control device 20 positions the fiducial mark plate FM on the wafer σ WST through the wafer stage driving unit 24 at a predetermined position directly below the hole of the projection optical system (below For convenience of the "reference position", the reticle alignment detection system 22 is used to detect a pair of first fiducial marks on the fiducial mark plate FM and a pair of reticle lines on the corresponding reticle R The position of the sheet is aligned with the mark. Then, the main control device 2 aligns the detection result of the reticle alignment detection system 22 with the measurement value of the interferometer 丨6, 丨8 at the time of detection to the memory β. Control device 2: The reticle stage is said to be in the γ-axis direction and to the opposite distance, respectively, using one of the aforementioned pair of reticle alignment detection systems 来 to detect another pair of i-th on the reference mark board FM The relative position of the fiducial mark to the alignment mark of the other pair of reticle on the corresponding reticle R. Then, the main control unit 20 stores the detection result of the reticle alignment detection system 22 and the measurement value of the detection 26: the instrument 16, 18 to the memory. Then, the summer relationship with the reticle alignment mark can be further re-paired with the i-th reference ^ and the corresponding reticle R on the 'b-step-measurement f-reference mark board TM of the upper b. Next, the main control device 20 uses the relative positional relationship between the at least two pairs of fiducial marks obtained in this way and the corresponding reticle alignment marks, and the measured values of the respective sizing instruments 16, 18 The reticle stage pedestal system defined by the long axis of the interferometer 16 and the wafer pedestal system defined by the length measuring axis of the interferometer 18 (hereinafter, simply referred to as "carrier pedestal system") are obtained. Relative positional relationship. According to this, the alignment of the reticle is ended. Next, in step 305, the measurement of the baseline is performed. Specifically, the wafer carrier WST is returned to the aforementioned reference position, and the design value of the reference line is moved from the reference position by a predetermined amount, for example, within the XY S, and the alignment system AS is used to detect the reference mark plate FM. 2 The main control device S 20 ′ is based on the information on the relative positional relationship between the detection center and the second reference mark of the alignment detection system as obtained at this time, and one of the first measurements when the wafer stage wst is previously positioned at the reference position. The reference mark and the corresponding positional relationship of the corresponding reticle alignment mark and the measured value of the wafer interferometer 18 at each measurement are used to calculate the reference line of the alignment inspection AS, that is, the reticle The distance between the projection center of the sheet pattern and the detection center (index center) of the alignment detection system AS (positional relationship). When the series of preparation operations are completed, the main control unit 20 takes out the program of the reticle alignment and the reference line measurement processing from the memory, and loads the specific program from the memory device 47 to the memory. Thereafter, wafer loading, wafer alignment (here, the alignment of the EGA method) and exposure of the respective irradiation areas SAP on the wafer are performed according to the 7 1377 program. Next, in step 307, the wafer is loaded onto the wafer tool 25 on the wafer stage WST through a wafer feeder (not shown). Here, in the present embodiment, the wafer carrier WST is subjected to a so-called adjustment of the rotation deviation and the deviation of the wafer w with high precision by a pre-alignment device (not shown) before loading the wafer. Pre-alignment, which is a coordinate system defined by the arrangement of the stage coordinate system (XY coordinate system) that defines the movement position of the wafer stage WST and the arrangement of the irradiation area on the wafer W (Fig. 2 The coordinates below (abbreviated as "wafer coordinate system") are consistent to some extent, and do not require a search aiignment of the mouth of the wafer w after the loading. Next, in the subroutine 309, wafer alignment processing is performed. This wafer quasi-processing system pushes the arrangement of the irradiation areas on the wafer W in the stage coordinate system, and calculates the arrangement, that is, calculates the center position of the total irradiation area. This wafer alignment process is left for later. Next, in step 311, the counter j displaying the irradiation area arrangement number is set to 1, and the first irradiation area is used as the exposure target area. Next, the 'step 313' moves the wafer stage through the stage control device 19 and the wafer stage drive unit 24 based on the arrangement coordinates (the center position of each irradiation area) of the exposure target area calculated in step 526 of FIG. 5 which will be described later. WST' moves the reticle by moving the wafer w to an acceleration start position for exposing the exposure target area on the wafer W, and passing through the stage control device 19 and the reticle stage driving unit (not shown). The stage RST is placed such that the reticle 28 1377598 R is at the acceleration start position. In step 315, the relative scanning of the reticle stage RST and the wafer stage is started. When the two stages reach the respective target scanning ambiguity to reach the isochronous synchronization state, the scanning area is started by illuminating the pattern area of the reticle R from the illumination light from the illumination system. Then, different areas of the pattern area of the reticle R are sequentially illuminated by the illumination light IL, and the illumination of the entire area of the pattern area ends, that is, the scanning exposure is ended. According to this, the pattern of the reticle R is reduced and transferred to the exposure target area on the wafer through the projection optical system PL. In step 317, it is judged whether or not all of the illumination areas have been exposed by referring to the counter value j. Here, since j = 1, that is, only the first irradiation area is exposed", the determination in step 317 is NO, and the process proceeds to step 319. In step 31, the value of the counter is increased (+ i ), and the next illumination area is the exposure target area, and the flow returns to step 313. After the determination in step 317 is affirmative, the processing and judgment of step 313 - step 315 - step 317 - step 319 are repeated. After the transfer of the pattern of all the irradiation areas on the circle W is completed, the step 317 is affirmative, and the process proceeds to step 321. In step 321 , the wafer W is unloaded to the wafer feeder (not shown): After removing the crystal grain W from the wafer holder 25, it is transported to a coating and developing apparatus (not shown) which is connected to the exposure apparatus by a wafer transfer system which is not shown. In the next (four) 323 +, determine whether the exposure of all wafers in the batch has ended at 29 1377598. If an affirmative judgment is obtained, the exposure processing is ended, and if the determination is negative, the flow returns to step 307. Here, since only the exposure to the first wafer w in the batch is ended, the determination is NO, and the process returns to step 307. Then 'until step 323 of FIG. 3, an affirmative judgment is obtained, and each wafer job processing object in the batch is sequentially implemented (4) 3〇7 (wafer loading sub-process f 309 (wafer alignment)-step 311 to step 319 ( Step 321 (wafer unloading) - step 323 (determining the end of the batch) processing. When step 323 obtains a positive judgment, the series of processing ends. That is, in the above exposure processing, each crystal in the batch is processed. The circle is sequentially loaded and loaded; the wafer on the D WST is processed as a processing object, and wafer alignment processing and exposure processing are performed. "Wafer alignment processing" Next, the wafer alignment processing of the subroutine 309 will be described. This wafer alignment processing system adopts the EGA method. Here, the ega method will be described first. The actual formation position of the wafer mark I formed on the wafer is not deviated from the design position (ie, irradiation). The reason why the formation position of the area % deviates from the design position) is that the stage coordinate system (X, Y) for specifying the moving position of the wafer stage WSt and the unconformity with the wafer coordinate system (α, β) generate this. The main reasons for unconformity are as follows: The rotation of the circle W. This is expressed by the retention error of the wafer coordinate system (α, β) to the carrier coordinate system (X, Υ). Degree: This is due to the movement of the wafer stage WST Incorrectly orthogonal, Β, the orthogonality of the stage coordinate system (X, γ) is represented by the orthogonality error W in the X-axis direction and the γ-axis direction. 30 1377598 . In the wafer coordinate system (α, Linear expansion and contraction in the direction and the p-axis direction. This is the overall expansion and contraction caused by the machining process, etc. This amount of expansion and contraction is represented by wafer calibration ~ and sY in the α-axis direction (four) axis direction. 'The wafer calibration Sx is the ratio of the measured value of the distance between the two points in the α direction on the wafer W to the design value, and the rounding scale Sy in the γ axis direction is between the two points in the p direction. The ratio of the measured value to the designed value of the distance D. The offset of the wafer coordinate system (α' β) relative to the carrier coordinate system (χ, γ): this is due to the slight offset of the wafer w relative to the wafer stage WST The result is a parallel component (offset) 〇χ in the X-axis direction and the γ-axis direction, and 〇 γ is used to indicate the error main cause of the field plus the 误差~D. To the design position (DX, DY) of the wafer coordinate system (α, β), in fact, it is predicted that the position (Εχ, Ε γ) on the stage coordinate system (χ, γ) obtained by the following equation can be obtained. [Formula 1] rEX, 'cos© -sin Θ, '1 -tan『, (DX\ 'Oc, sin0 cos© y , 〇i > w+ .(1)

In general, the orthogonality error w and the residual rotational error 被 are regarded as trace amounts, and the transfer position (DX, DY) of the design is determined by expressing the trigonometric function of the above formula 一次 by a similar approximation. Relationship with predicted transfer position (EX, EY). [Formula 2]

V EX, , Sx , Θ Θ -Sx (order + Θ)) (DX, DY,

(2) 31 1377598 The rotation component of the α-axis to the χ axis of the wafer coordinate system is r. The rotational component of the β-axis to the Y-axis of the wafer coordinate system is RX and y. When 旋转 directly corresponds to the rotation component Rx of the X-axis and the error W of the positive six production corresponds to Ry_Rx, the above equation (2) can be converted into the following equation. , [Formula 3]

Sx

\°y) -(3) Further 'Substitute Sx, Sy for 1 + Sx, J + Sy, Sx

Ry=Sy · Rx and 0, can convert the above formula (3) into the following [Equation 4]

Ί + & (〇χ\ ...(4) The above formula (4) 模式 EGA mode is the most general linear mode formula, in this formula, wafer calibration Sx, Sy, wafer rotation Rx, Ry, , offset 〇x, 〇y is the coefficient of the formula (so-called 6 EGA parameters). The main reason for the deviation between the actual transfer position and the design position is the deviation of the above-mentioned stage coordinate system from the wafer coordinate system. In addition to the cause of the error, there are other various causes of error. For example, it also includes the error component of the illumination area itself (for example, the projection magnification error including the projection optical system pL (which causes the change in the size of the illumination area), and the reticle The rotation error of the pattern itself on the stage in the scanning direction, etc. Hereinafter, simply referred to as "intra-emitter component"), the influence of the error component may not be neglected. The composition of the above-mentioned irradiation may be formed at the four corners of the irradiation area SAP. B曰circle '圯MP κ position is observed. Suppose that the center cP in the irradiation area sAp 32 1377598 is the internal coordinate system of the origin, and the wafer mark is placed on the design of the coordinate system (wafer) When the predicted transfer position of the mark mx 曰, K iiiyp, K) and the round mark MP K is (mxP, K', myPK') due to the influence of the internal component of the above irradiation, the relationship can be defined as follows. 5] (\ + sx ~ry^\ 'like, \ rx l^-sy) , myP^ ...(5)

Here, sx, sy means the irradiation internal coordinate system to the stage coordinate system calibration component (shot scaling component), and rx, ry is the rotation component (irradiation rotation component) of the irradiation internal coordinate system to the carrier coordinate system. According to the above description, the predicted positions (MXP K', MYP κ) of the wafer marks MP and K of the sample irradiation SAp of the stage coordinate system are expressed by the following equation. [Equation 6]

Rl + Sx t Rx 1 + handle claw

(6) where '(DXP, DYP) is the design coordinate of the center CP of the irradiation area SAP. In the EGA method, the design position (DXp, DYp) of the irradiation area is substituted into the formula (DX, DY) of the above formula (4), and the design position (EX, EY) of the irradiation area is obtained, and the position is regarded as The center position of the irradiation area SAp of the stage coordinate system. For this reason, the values of the coefficients Sx, Sy, Rx, Ry, Ox, and Oy in the mode of the above formula (4) may be obtained. However, since the values of the coefficients Sx, Sy, Rx, Ry, Ox, and Oy are calculated based on the actual positions 33 1377598 of the wafer marks MP and K at a plurality of points, the above-described internal components of the irradiation must be considered. Therefore, in the present embodiment, in order to determine the values of the coefficients, a plurality of (for example, "solid" wafer marks Mi attached to the plurality of irradiation regions 3Αί(ϋ 2, 3'·.·, η) are actually measured. K is at the position of the stage coordinate system (MXi K, Μγ; κ), using the sampled illumination s Α (designed position (DXi, called, and h measured wafer marks in the design position of the illumination internal coordinate system (mXiK , myiK), based on the model formula of the irradiation component shown in the above formula (6), and performing a statistical process (for example, the least squares method) to obtain the above formula (6) in which the evaluation function value of the following formula is the smallest. In the equations, the values of the coefficients Sx, Sy, RX, Ry, 0x, 〇y, % % %. In the following equation, the measured positions of the wafers to be measured are set to (MXg, MXg). ) (g = 1, 2, ..., h), the coefficient obtained, the values of Ry, 〇χ, 〇y, sx, sy, rx, ry are substituted into the above equation (6) to be measured. The predicted positions of the wafer marks Mi, k are set to (MXg, Μχ [Expression 7] g, g, respectively.

That is, the evaluation function is the square of the remainder of the predicted position (MXg, Μ, ,) of each wafer mark Mi, k to be measured by the above formula (6) and the position (MXg, nickname). And the number of i Q samples obtained by dividing by the wafer mark h (ie, the average of the sum of squares) is 'here' (four). The i-series is sequentially assigned (for example, the measurement order) to the sample number, and the g is the actual measured wafer mark. number. Further, in the EGA method, although the number of samples h of the wafer mark I is larger, and 34 1377598 is preferable from the viewpoint of alignment accuracy, on the contrary, the number of samples of the wafer mark h is from the viewpoint of efficiency. The less the better. Furthermore, the number of samples h of the wafer mark Mu is limited by the number of unknowns (parameters) in the EGA formula. For example, if one of the above equations (6) is used

When Sy' Rx' Ry' 〇x' 〇y, sx' sy, Γχ, ry) is all set to an unknown number (parameter), since the number of parameters of each axis is 5, the crystal for 2-dimensional position detection is used. The number of samples h of the circle mark Mik needs to be at least 6 (=5+) or more (if it is a wafer mark for one-dimensional position detection, it takes 6 samples or more per axis), and the coefficient of the above formula (6) is, for example, If sx, sy, Γχ, and 7 are constant, the number of parameters of each axis of the formula is 3, and the number of samples of the wafer mark ^ for 2-dimensional position detection is 4 (= 3 + 1). In the ship mode, the more the parameter of the model is increased, that is, the more the number of samples of the wafer mark h must be increased, so it is disadvantageous in the efficiency of the τ 丨 丨 α α α α α α α α α α α α α α α α α Parameter degree of freedom and number of samples of wafer mark Μ·) == Decrease the alignment accuracy, reduce the parameters of the above formula (4) as much as possible: the parameters of the paradigm, and reduce the number of samples of the wafer mark I as much as possible. The circle mark M... the sample number is finally obtained, for example, in a manner of not selecting the three marks arranged on the straight line. The number of samples of the measurement mark that is valid for the derivation of the reference number. " In addition, in order to implement the optimization of the parameter number of the model expression shown in the above formula (6), in the present embodiment, the coefficient is taken from the experience. Each of the coefficient values is classified into a class in the equation (6), and the coefficients are classified in advance. 丨) Specifically, in the following embodiment, the coefficient of the above-described formula is classified as 35 1377598 as follows. Each wafer is estimated, and according to the current sampled irradiation (the measured value of the wafer target, the coefficient that is necessarily changed, that is, the coefficient as a parameter; b, the estimation of each wafer is used to determine that it is based on this The measured value of the measured value of the wafer mark, or the coefficient obtained from the knowledge (constant); the competition c, the shirt for each batch, according to the sampling exposure (the wafer rod :i magnitude, must be changed as a parameter [but in the batch, the parameters obtained by the pre-knowledge (constant) are used; d. Presumption is performed for each batch, which is used to judge the irradiation according to this sampling. Wafer mark The measured value is the estimated parameter, or the coefficient obtained as the value (constant) obtained from the pre-knowledge, but in the batch, the parameter obtained by the pre-knowledge value (constant) is used; e, in principle, the value is fixed to The constant, which is estimated as a parameter as needed, and the coefficient of the changed value; :, the so-called prior knowledge refers to the model of the above formula (6), each coefficient == knowledge is broad, is the simulation, or the actual empirical formula obtained in the past The value obtained from the pre-knowledge refers to the value of the coefficient that is derived from the past: the value of each coefficient is predicted. Moreover, the preset value of each coefficient in the prior knowledge, for example, may be remembered in memory. This: The value of the prior knowledge of the previous batch can be used. It is also possible to use the material value (the change in the value of each coefficient of the observable model, since a tendency to observe a certain degree is not observed separately) The 36 1377598 coefficients are classified. In order to show the classification. For example, since the offset component 0x' 〇y and the rotational component Rx, Ry of the wafer are uneven due to the pre-alignment result of the wafer being loaded before the sunburst, the WST is generated, so each - The wafer is presumed to be more appropriate, and the coefficient classified as a. is better. 'Because of the wafer calibration Sx, Sy depends on the characteristics and process of the exposure device more than the dependence on the wafer, so the variation between the wafers in the same batch exposed by the same exposure device in the same process It won't be too big, but since the value between the wafers in the same batch may change somewhat, the coefficient of the knife is b. Further, when the values of the coefficients (sx, sy, rx, ry) of the components in the irradiation are basically regarded as not to be changed, the coefficient classified as C· is preferable. Further, the classification of each coefficient of the above-described EGA model (Equation 6) is not limited to this classification example, and various classification modes can be employed. For example, the coefficient (sx, Sy, rx, ry) of the component within the illumination can be classified as a coefficient of d. However, here, it is assumed that the classification of the above examples is carried out, and the following description is made. - Again, the designation of this exemplary coefficient classification has been performed by the operator through an input device not shown, and the specified content has been stored in the memory device 47. Hereinafter, the circular alignment processing will be described along the flowchart of the subroutine 3〇9 shown in Figs. 4 and 5 . X ’ is as described above. This subroutine 3G9 performs the wafer once when replacing the wafer of each processing target. Therefore,

In the following description, the processing flow when the first H τ <the wafer wafer in the batch is processed as the processing object will be described first. Then, the processing flow when the second and second chips are used as the processing target will be described in order. When the wafer is to be processed, the first example of the processing of the first piece of the 37 丄J / / is described. In FIG. 4, in the step 402, it is determined whether the s yen of the processing object loaded on the wafer cutting table u, ^..u is the wafer within the m (for example, 5) chip before the batch. w. If the right judgment is affirmative, the process proceeds to step 404, and if it is negative, it proceeds to step 418. Here, since the first wafer w in one of the batches of b β . ^ on the sundial stage WST is loaded, the determination is affirmative and proceeds to step 4〇4. In step 4〇4, the sampling information of the Japanese yen w, 兮b m ΛΙ, >& ^ ...,,. In the EGA process, the month J is selected from the irradiation area A on the wafer w to select a plurality of irradiation areas as the sampling irradiation %, from the wafer mark Μ attached to the sampling irradiation %; u, 屯, 丨The person ^丨, , k Ψ measures the position of the h wafer marks (MXg, MYg) (g L 2, ~, h). The so-called sampling information refers to the information on the quantity and configuration of the measured wafer marks. In the present embodiment, although the number of samples is optimized along with the optimization of the model parameters, at this stage, since the wafer w in the front of the batch is the processing target, the model is not optimal. The number and arrangement of the wafer marks Mi>k of the measurement object at this point in time must be such that when the coefficients of the model of the above formula (6) are all 4 as parameters, The number and configuration of the number of valid samples required for the parameter. Specifically, since the maximum parameter degree of freedom of the exemplary expression of equation (6) is 5 ′ for each axis, the number of samples here must be 6 or more. Here, the number and arrangement of the wafer marks Mi k satisfying such an effective number of samples are assumed to have been stored in the memory device 47. Here, the predetermined sampling information is obtained from the memory device 47 'held in the memory for reference. The information. Also, here, as the preset sampling information, the number of samples is specified as h = 3 〇. In the next step 406, the wafer stage 38 1377598 WST is moved in the XY plane so that the sampling information is designated as the irradiation area SA of the sample illumination; h wafers in the wafer mark Mi k attached thereto The mark can sequentially move in the detection field of the inspection system AS according to the measurement order specified by the sampling information, and take the wafer mark with the alignment detection system AS.

Mi, k' simultaneously finds the position of the wafer stage wsT at this time from the measured value of the wafer interferometer 18. Since the position information of the wafer mark I in the photographing field of view is sent from the alignment detecting system As, the position information of the wafer mark, the measured value of the wafer interferometer 18, and the reference line can be obtained. The measured value of the position of the wafer mark Mi, k in the stage coordinate system. The measured values of the obtained wafer marks Mi, k are held in the memory. In this way, the measured value (MXg, MYg> of the position of the wafer mark ～ attached to the irradiation area designated as the sampling information is obtained. The next step g 4G8 'is classified as the coefficient of e as a matter of fact. After knowing the constant of the decision, the value of the system and the number classified as a. to d. is calculated based on the 1 = value of the position of the obtained wafer mark. First, refer to the ί:: hidden device 47 In the case of prior knowledge, the classification is classified as e." Here, 'because it is still based on the first wafer in the batch', it is classified as the value of the pre-knowledge of the coefficient. It is 〇. That is! Here is the coefficient of the model of Hi(4), set U =

After ry=〇'' according to the memory, the measured value of A (MMg, MY), the circle of the umbrella, the position of the §Mi, the position of the k, the evaluation function is the most: the method is also used to find the above formula (7) The value of 0y shown. Also the corpse satisfies the following parameters sx, Sy, Rx, Ry, 〇x,

[Formula 8J 39 1377598

dE__ dE _dE dE dE

dE dSx dSy dRx d^~dO^ = ^; = 0 - (8) Next, in step 410, the determined EGA parameter is set to the coefficient of 6) above, and the measured wafer is marked~ The δ position (mXi, k, myi, k) of the inner coordinate system is substituted (mxP, "c, myp, k), and the design position (DXi, DY) of the center q of the sampled illumination SAi is substituted (DXp, DYp). ), the position of the expected wafer mark Mi > k (MXik', MYj, k,) is obtained. Then, the position (MW, MYi, k') to be paid is taken as (MXg, MYg,), and the measured value (MXi k, MY; k) of the wafer mark position is substituted as (MXg, MYg) into the above formula. (7) Find the value (residual). Further, in step 412, the above-mentioned mode (the statistical mode in which the coefficient classified as a. to d. is a parameter and the coefficient classified as e. is based on the value (constant) of prior knowledge Μ (first regression mode) is calculated. )) AIC (Akaike's Information Criterion). Next, simply explain this AIC. When the mode μ is obtained by measurement, as the amount of the logarithm of the normal distribution when the error indicating the mode is the same as the true mode, KuUbaek_

Leibler message volume (KLI). The number of KLI data (herein referred to as the number h of sampled illumination) is the AIC which is corrected by the deviation of the number of parameters in the performance mode. The statistical mode AIC is expressed by the following formula. [Equation 9] AIC(m) = -2 · log z, + 2 · i/ (9) where 'logL is the maximum logarithm of the correlation with the statistical mode M, d is the parameter of the system mode ( The number of unknowns and L are approximate. For example, 1377598 uses L=(1/E) according to the evaluation function E shown in the above formula (7). That is, since the value of the evaluation function (remainder) in this mode is smaller, and the degree of approximation is "larger", the validity of the mode is higher, and the value of the result AIC is smaller. In addition, the item of 2d is the compensation item with the increase of the relative parameter, and the larger the number of parameters, the larger the value of the item. Also, gp, the increase in the degree of freedom of the meal, if the heart is more than the increase in the degree of accuracy, the mode with a large degree of parameter freedom will not be selected. In addition, although AIC is an abbreviation for the Akaike Information Standard, it is also used as a function name of the information amount criterion in the above formula (9). That is, if AIC is used as an indicator of the degree of correctness of the model, the tendency to select a mode with a large degree of freedom of parameter is suppressed, and the material that is the number of lines is not selected to be reduced, and the error of the generalized error can be selected. mode. As described above, in step 412, the value of the EIC (G) value is calculated by substituting the value of the EtlGgL value of the evaluation function (residue) and the number of unknown parameters d in the above equation (7). In addition, in the case of μ, +•, and the above-mentioned coefficients, since ji, y, Rx, Ry, Ox, 0y are used as parameters, sx, sy, rx are used as the parameter number of m according to prior knowledge. 6β 下-Step 414, the coefficient of the material pattern M, U2_(four) a. C. according to the above formula (4) is calculated as the parameter, and the value of the pre-knowledge (constant, pull) is used as the coefficient. According to the value of the matter (constant), the evaluation function obtains the pre-existing knowledge of the § "body and ly'ry of the above formula (6) from the memory device 47, based on the acquired 5 sx, sy, ΓΧ, 于·^, the value of the pre-knowledge decision of the service, w; (6). Then, according to the above steps 4 » 篁 篁 之 h h wafer mark 41

I jyQ records the measured value of the position, which is calculated as the minimum or full value. The value of the coefficient of the knife neck ac is 0. Substituting the actual value of the position of each wafer mark Mi, k into the above equation (6), and obtaining the evaluation function of the above-mentioned gate is minimized by a thousand method. The sub-type break is classified as a. c is your number (parameter) Rx, Ry, 〇χ, 〇 y • (10) dRx dRy dOx ~ inside shot (four) 4 () 6 measured wafer mark % in Zhao =: The design position of the coordinate system (, ~: p'k yw in the sampled irradiation SA), the position (1) of the DY } φ ^6, and the undesired wafer mark of the DYi) (Dx DYP) (The measured value of V, ', k, MYi, k'), the measured value of the marked position 2, MYi, k') as (MXg, MV), the measured value of the wafer ( Mxi,k, MYik) as (Mxg, MYg) = = (7) 'No value (residual). In the above values, the above statistical mode M is calculated, and the AIC is... "In the example, since the system is Rx , Ry, 〇x, 〇yh number n ^, &quot;, (d) is based on the value of prior knowledge by the number) Therefore modulo 4 M, the number of parameters d is 4. (often the dog tr· owed, in Figure 5胄5G2, comparing the AIC(M) obtained in step 412 With the mode M, $ value obtained in step 416, it is judged whether or not the piano, and <AIC(M)e '' proceeds to step 504, and if the determination is negative, then proceeds to 510. Again, this The wafer w is the object of the front side (the first piece) of the batch, and the coefficient of the value is determined by the prior knowledge, which is the same as the coefficient of the model formula shown by the formula (6) related to the wafer 42 1377598. The probability is not necessarily large β. Therefore, here, the value of AIC (M,) is assumed to be greater than the value of AIC (M), and it is judged as negative. Then, the process proceeds to step 510 to perform the following description. In step 510, The value of the evaluation function (residual) E in the statistical mode 与 is compared with the predetermined value to determine whether E is smaller than the predetermined value. If the determination is affirmative, the process proceeds to step 512. If the determination is negative, the process proceeds to step 516. Here, since the value (predetermined value) of the prior knowledge classified as the coefficient of 6 does not coincide with the wafer preceding the batch of the processing object, the value of the evaluation function (residual) E of the mode m is larger than the predetermined value. The determination is negative, and the process proceeds to step 516 to perform the following description. The step after 5 1 6 is a process of taking all the coefficients of the mode shown so far as parameters. First, in step 516, referring to the sampling information stored in the memory, it is determined whether additional sampling illumination is required. ° If the determination is affirmative, proceed to step 518. If the determination is negative, proceed to step 522. Here, as previously described, since the storage in step 404 is preset in the sampling information of the memory. The number of valid samples (h 2 3 〇) 'sufficiently greater than the coefficient that will be classified as a.~e• as the parameter, the degree of freedom of the parameter (5 per axis), therefore, because no special sampling is required The additional measurement of the illumination, so the determination is negative, and proceeds to step 522. In step 522, after all the coefficients to be classified as a•~e. are taken as parameters, the h wafer mark positions measured in step 406 are determined. The measured value 'calculates the value of the coefficient classified as a. to e. That is, the measured value of the position of each wafer mark is substituted into the above formula (6), and the minimum value of 43 1377598 is determined by the least square method, that is, the value of sy, ΓΧ, ry is satisfied. The following equations give the evaluation function E number Sx, Sy, Rx, Ry, 〇x, 〇y, sx, [Expression 11] = -= ~-_a£ dE dE dE dE λγ eSx of the above equation (7). The value of the coefficient (parameter) that is obtained by the kiss dRx (10) is classified as a·~e., and is used in the exposure processing described later. Therefore, the memory is classified as b. e. The parameter value, as a brother and sister, the knowledge storage device 47. Next, in step 526, the step 524 is obtained. The coefficient value of the memory is set to the value of each line of the above formula (4), and the design position (DXp, DYp) of the full = region SAP is substituted into (10) DY of the above formula (4), respectively, to calculate the total irradiation area SAp. Position of the center cp (reference position) (EXP, EXP) 'The calculation result is stored in the memory device 47. After the step 5M is completed, the subroutine 309 is ended. Next, an example of the processing flow of the second wafer w in a batch as the processing target subroutine 309 will be described. First, in step 402 of Fig. 4, it is judged whether or not the wafer W to be processed is a wafer of m or less (e.g., 5) in the lot. Here, since it is only the second wafer W, the determination is affirmative and proceeds to step 404. In step 406, the sampling information stored in the memory is referred to, and the number of sampling illuminations and the configuration contained in the sampling information are taken. The sampling information stored in the memory (the number of samples and the configuration of the wafer mark for sampling the irradiation, the number of valid samples) is in the previous subroutine 309 'because the sampling information stored in the memory device 47 has not been used yet. Update, therefore, here is still ^/7598 with reference to the preset sampling information. In step 408, according to the obtained wafer mark μ, only the measured value is set, and the same as the previous subroutine, the coefficient classified as \d. is used as a parameter, and will be divided into four (4) e materials. The modulus of the constant determined by = is classified as a coefficient: the value of ^, which is obtained by using the above formula (8) as the least squares method of the conditional expression, and in the step, the remainder E of the time is calculated. At the time point of the step of the nasal discharge AIC (M) e, as the coefficient classified as e: the number of the first (four) value stored in step 524 of the sub-program 309 of the number I (refer to FIG. 5) The job value of the wafer is stored in the memory device 47. Here, the coefficient value of the second knowledge can still be used as the pureness of this time. In the eight types of Γ-step 414, the statistical mode of the EGA is used as the mode m of the model expression of the formula (6) when the coefficient is classified as bd, and the value of the knowledge (constant) is The least-flat method is used to calculate the value classified as ... and the evaluation function (residual) E, and the value in the mode M is calculated. At this point, as the second: 524fl pre-existing knowledge, in the previous sub-program 3G9 step 5: (refer to Figure 5) stored in the memory device 47 &lt; b, d.: that is, in - _ s. . &lt; coefficient value, is stored in the memory device = processing coefficient value obtained, d., e · coefficient value is expected this ^ ^ as the prior knowledge b ·, # is the current b., d ., the coefficient of e· is used. - The person's step in Figure 5 compares the AIC(M) value of the mode obtained by the step 412 45 1377598 and the mode M of the M which is requested by the step 416 of this time. The value of (M') is judged whether or not AIC(M,) &lt;AIC(M). Here, 'AIC (M') 2 AIC (M) is judged as negative, and the process proceeds to step 510 to carry out the following description. Step 5 10, judging whether the evaluation function (residual) E in the mode IV is less than a predetermined value, here, because in the processing of the s曰 circle w of the head of the previous subroutine 3〇9, in the batch Then, each coefficient which is classified as e. which is predicted to be almost worthless is obtained, and the coefficient is used as the mode m. Therefore, the value of the remainder E is smaller than the predetermined i, and a positive judgment is obtained, and the process proceeds to step 512. Carry out the following instructions. Next, the step 512' selects the mode M, and the coefficient of the mode M is held in the memory for performing the wafer W exposure process for the current time. The next step is to determine whether the wafer w of the processing object is

Within the first m (= 5) slices. Here, since the wafer W of this time is the second piece of the lot, it is judged as positive, and the process proceeds from step 524 to step 562. In step 524, the coefficient values of the mode 迸芏, __•~e. are stored in the memory device 47. In step 526, the heart position is stored and stored in the memory device 47. In the step _ &amp; 4, the program ^ step... ends, that is, the end of the sub-step, and the processing flow of the third program 309 in a batch is explained. First, when 0aB1 is used as the processing target, the determination is affirmative, and whether or not the wafer w of the processing target is the step 402 of the batch is determined to be "there is only the third wafer w, -5". The wafer is within the slice to step 404. 46 Step 404' refers to the number of samples on the memory and the g-sampling of the wafer mark of the sampled illumination (taken .m } is also updated in the previous subroutine 309, so it is still referred to here. Preset sampling information. The mode in which the coefficient of e. is determined to be constant.) The pre-root knowledge will be classified into actual measured values, and the position of the wafer mark of the classified one will be calculated, and the remainder E 〆 will be calculated. The value of 'in the step here, can also be calculated according to the prior knowledge 7 412 (5). The knowledge stored in the memory device 47 is classified as the coefficient value of e. For example, the average value of the values of the former & 妁佶, 田你生 A, and then 刖 is still classified as e The coefficient value. Secondly, 'Step 414' is the statistical mode of the job, which is classified as b, d... by the classification of the parameters as the value of the prior knowledge, ', M, and the model of the most () The model thousand method calculates the value of the coefficient valence function (residual) E of the classified 4 a., c., and at step 416: the value is sweated at this time point, the value of the previous time, # (). Step 524 of updating H sub-private 3〇9 in Figure 5) (refer to the pre-knowledge of the coefficients of b, / which are classified as:,::= beforehand, here, H The average value of the coefficient of H 1•, (4) is the current b., the step of 2 2 5 is required, and 502 is compared with the step of this step - the mode obtained... the value of 22 C (M), And here, as determined by step 416 of this time, it is classified as jd==M,) &lt;AIC(M). • The coefficient of the factor just the value of knowledge, convergence 47 1377598 The coefficient of the true mode, the judgment is affirmative, and proceed to step 5〇4 to make the following explanation. In the next step 504, in order to perform the current wafer exposure processing, the coefficient of the mode is held in the memory. 〃-人, in step 5 〇 6, to determine whether the wafer w of the processing object is within the m (= 5) slice of the batch. Here, since the wafer w of the current time is the third slice of the batch, the determination is affirmative, and the process proceeds to step 524. In step 524, the mode M, the b. to e2 coefficient values are stored in the memory device 47 as prior knowledge. At step 526', all the areas of the illumination area are calculated and stored in the memory device 47, after which the processing of the subroutine 3〇9 is ended. : After 'in the implementation of subroutine 3〇9, until the wafer to be processed is the m+1th (=6)th wafer of the batch, that is, until the judgment of the step becomes "No" The processing of steps 4〇4 to 416 is performed, and step 502 of FIG. 5 is determined. When the arrangement of the irradiation areas of the target wafer 1 is processed to show the tendency to align with the irradiation area of the wafer w to be processed in the past, | AIC(M') &lt;AIC(M)' is judged , in m4, the mode m, the coefficient is kept on the memory, the implementation of step 6 (conditional judgment is of course affirmative), in the subsequent step 524, the coefficient (parameter) value of a, c., and according to b , d as the pre-existing knowledge is stored in the note 7, the creation: the coefficient value of knowledge 'end subroutine 3. In the second step, the second step is performed in step 526, and the thinning is performed... The arrangement of the irradiation areas of the wafer W is obtained by the tilting C (MgAIC (M) of the irradiation area of the processing target wafer. Negative judgment, go to the step. 48 ^377598

Moreover, in step 510, the remainder of the mode M, E., ΆίΓ π j &amp; when the value is 疋, the mode 选择 is selected as the best mode, and in step 512. The coefficient value of the mode 保持 is maintained, and the pre-knowledge U is operated in step 524. . , Λ/ί towel%. I have remembered to set 47, but when the remainder ε of the mode 大于 is greater than the predetermined value, the processing of step 516 (negative judgment) - step 522 will be divided into the same cylinder, the knife m horse a The coefficient value of .e is re-determined as a parameter and held in the memory. In step 524, the coefficient value is stored in the memory device 47 as prior knowledge. 'As mentioned above', in the previous embodiment, before a batch of films, the focus is placed on the process of pattern optimization, the value of the knowledge before the coefficient of the mode, and the coefficient of the true mode of the batch In the vicinity of the value, the EGA is performed while maintaining the effective number of samples at a preset number (example #h=3〇). In this process, the coefficient value of the illumination which is classified as the value of e which hardly changes is determined by the processing of the flute, the circle of the ^ 隹, and the first slice. In the second piece of wafer processing, the prior knowledge of the coefficient values classified as b·~d. is converged to the coefficient value of the true mode, and the mode M having a small parameter degree of freedom is selected. Further, if the arrangement of the wafers after the fourth sheet is substantially the same as the arrangement of the third wafer, the subroutine 3G9 in the processing of the wafer always selects the mode M, but the fourth When there are some differences between the arrangement state of the wafers after j and the arrangement of the first to third wafers, the mode M will be selected, and the coefficient of the bde which becomes the parameter mode will be regarded as new prior knowledge. The accumulation is stored to the stealing device 47. Further, when the arrangement state of the wafers after the fourth sheet is significantly different from the arrangement state of the third wafer, the coefficient "b" is newly obtained as the new prior knowledge to the memory device 47. 49 i377598 Then, when the wafer to be processed is the sixth slice in the batch, in the subroutine 309, the determination in step 402 is negative, and the process proceeds to step 418. At step 418, reference is made to the sampling information of the wafer. Since the sampling information stored in the memory is still in a preset state, the preset sampling information is obtained. In the next step 420, in the same manner as in the step 4G6, the position of the h wafer marks as the measurement target in the wafer mark included in the sample area specified as the sampled illumination in the sample information is sequentially measured. . In the case of the human V step 422 t 'the mode M is taken as the parameter classified as &amp;., b: the number, and the coefficient classified as e., d., e is used as the constant determined by the prior t The mode calculates the coefficient value classified as a and b according to the position of the obtained wafer mark = using the least square method, and the circle is obtained by repeating the above-mentioned formula (4) to obtain the crystal. 3 = Measured position 'Substitute the predicted position and the measured value of the wafer mark position into the above formula - "Four () function E ' to find the value (residual). Step by step, in step 426, the coefficient of the upper b. is calculated as a parameter, and the coefficient of the fabric is classified as a·, and the coefficient of the fabric of the prior knowledge is used as the AIC(M) of the system 4 model μ). In the next step 428, EGA is the coefficient of a. of the above formula (6) as a parameter, (as a mode based on the value of prior knowledge), the parameter 2 is the most: the coefficient value is required to calculate the evaluation function in mode M ( The remainder is the value of E. The method plus the value ΤΙ in step 43G, the above statistical mode M is calculated, and the Aic() M, 50 is changed: ' In the same way, when there are estimates for each batch, and the number of free households is classified as the coefficient of c., d., the parameter degrees of freedom of the mode M, M, are in the mth piece before the batch and then Each of the slices... =, after the ... slice is set to the freedom of each mode: Next, in step 50 of Figure 5, the AIC(M) value and the step of the mode are determined. The mode M obtained by gamma, the AIC(m,,) pair = comparison: whether or not, &lt;AIC(M). Here, if the arrangement of the processing target circle and the arrangement of the wafers processed in the past are substantially the same as AIC(M)&lt;AIC(M), an affirmative determination is obtained, and the process proceeds to step 5〇4. Select mode M and keep its coefficients on the memory. At the next time, 506, because it is the sixth wafer, it is negative and proceeds to step 508 ° step 5〇8, which updates the sample 2 on the memory. At this stage, the reliability of the prior knowledge classified as the coefficient of b. to e is increased, and it can be judged that the mode M has been reached, and the selected state, so the number of valid samples (h = 3〇) can be When the predetermined larger number of samples is reduced to, for example, only the coefficient classified as a. as a parameter, the number of high-precision alignment precision can be sufficiently ensured. For example, at this time, if the coefficient of (4) a. is set to the offset of the wafer as "x, 〇y, the rotational component Rx, Ry, the parameters of each axis are free; &amp; 1 m is 2' Set effective sampling for each axis

The number is 3 for the # p A of the 2D mark and 6 for the 1D mark. However, here, the arrangement of the wafers after the flutes may be different from the arrangement of the wafers in the past. This embodiment is based on the case where AIC(M) is also calculated. AIC (M). Therefore, in this case, a certain margin is considered, and the number of effective samples is increased by 1 ’ in the mode Μ parameter 3 of each axis, and each axis is set to 4 or more (for example, 16). Here, the number and arrangement of the wafer marks to be measured for each of the effective number of samples 1 to 3 are presumed to be the most approximate, and are stored in the memory device 47, when the number of valid samples is updated. The information on the degree of freedom (the number of wafers and the configuration) is read from the memory and the event to the memory, and the information on the memory is updated to the information. The wafer W after the (m + 1)th sheet is subjected to the processing of steps 418 to 430 for the wafer W to be processed, and the step 502 of Fig. 5 is determined. Then, 'when the arrangement of the irradiation areas of the processing target wafers shows the same tendency as the arrangement of the irradiation areas of the wafer W to be processed in the past, 'AIC(M,) &lt;AIC(M)' is judged. And step 5〇4 (selection of mode M')—step 5〇6 (conditional judgment is negative)~step 5〇8' further sample information so as not to be less than the parameter μ parameter degree of freedom + 1 Limit, reduce the number of valid samples. Then, in step 524, the coefficient values based on the prior knowledge b. to d. are stored in the memory device 47 as prior knowledge. On the other hand, when the arrangement of the irradiation areas of the processing target wafer W is different from the arrangement of the irradiation areas of the processing target wafer w in the past, ^AIC(M) is judged as negative, and the process proceeds to step 51. Step Η. In the case where the remainder of the mode 小于 is less than the predetermined value, the mode μ is selected as the optimal mode, and in step 508, the sampling information is updated to increase the number of valid samples slightly. In step 524, the mode is classified as The coefficient (parameter) value of a., b. is stored as a new prior knowledge with other coefficient (constant) values to the C memory device 47 (the value of the coefficient classified as e is originally based on the value of the knowledge before 52 1377598) 〇 — — — 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于 大于As the mode of the bright circle of this time, the coefficient is kept in the memory, and will be:; the coefficient value is stored as a new prior knowledge to the memory device: a ^ outside the 仃 仃 乡 乡 516 ～ 522 522 processing memory In the number of samples, the number of samples in the number of samples is the number of samples, which is the number of samples (only a., b. is the number of samples, so it must be, there are: the degree of freedom has a small margin ., "The number of valid samples is updated to correspond to the a.~e. When the number is a parameter, the τ step 516 is affirmed by the number of 1, and in the step sample m, ... in step 518, the memory on the memory is freely produced corresponding to the coefficient of 3•~ The number of valid samples of the parameter at the time of the parameter (3)) (6 or more) (for example, the same as the preset in the next step 520, the 媸 媸 s 加 加 取样 sampling &amp; (4) # updated sampling information 'measurement The measured position of the position of the round mark is used to determine the number of crystals measured by the classified data. The second part of the rounded mark is classified into the system of the item, and the knowledge of the item is further classified. Save to the memory of the king., ', · the processing of the product. ^ After the device 47', the end of the subroutine 3〇9 line containing two of the wafers in the batch for processing, sequentially into the second step St 3 〇7~step cycle processing, step 323 obtains a positive judgment. During this period, 53 1377598 program 309, depending on the arrangement state of the processing target wafer, appropriate, ^ per processing 'moving $' mode Parameter freedom and effective number of samples are optimized 6 and described in the present embodiment The processing flow of the first, second, and first wafers for the processing object (fourth) is only _example: when the wafer is well-arranged based on prior knowledge, the first wafer is processed. In the case of the object ', the positive judgment can be obtained in step 502 and step 510. In the case of selecting the mode M, M', and when the second wafer is to be processed, the determination in step 502 is affirmative, or If the determination in step 51 is negative, or when the third wafer is to be processed, the determination in step 502 is negative, or the determination in step 5 1 is affirmative. In any case, in this embodiment, In the process of processing the wafers in sequence, the pre-existing knowledge is accumulated to optimize the degree of freedom of the parameter parameters. Further, in the present embodiment, 'm = 5 ' is set, but this value can be arbitrarily set. Preferably, m is set to be sufficient to obtain the prior knowledge of the coefficient of the pattern M of the batch (for example, if it is assumed that the prior knowledge can be fully obtained at one time), m = 1. Further, in the present embodiment, 'the number of valid samples is not optimized until the mth slice in the batch', but the judgment of step 506 can be omitted, and effective sampling can be performed within the mth slice in the batch. The number is updated. Further, in the present embodiment, in the subroutine 309, it is determined whether or not the first regression mode or all the coefficients of the above equation (6) are selected by determining whether or not the remainder of the mode as the first regression mode is smaller than a predetermined value. As the third regression mode of the parameter, the AIC of the third regression mode of 54 1377598 is calculated together, and the AIC of the mode is compared with the AIC of the mode, and the value is smaller. However, at this time, since it is necessary to set the number of effective samples of the circle mark to the degree of freedom of the third regression mode having a large degree of parameter freedom, the number of parameters tends to increase, so that the above program 3 is still performed. 0 9 is preferred. Further, in the present embodiment, the average value of the past values of the coefficient obtained as the prior knowledge is set as the representative value of the coefficient, but the coefficient value at the time of processing the previous (or previous) wafer processing is represented. Values can also be used directly. When the coefficient value when processing a plurality of wafers is reflected to the current coefficient value, the coefficient values can be weighted separately. For example, the value of the most prior knowledge can be weighted the most, or the value of the prior knowledge of the remaining small wafer can be maximized. As apparent from the above description, in the present embodiment, the memory device is constituted by the memory device 47 of the main control device and the memory. Further, in the present embodiment, the main control device 20 corresponds to the selection device of the position detecting device of the present invention, the calculation device, the updating device, the optimization device, and the determining device. That is, the processing by the CPU of the main control device 20 to step 430 (FIG. 4) and step 502, step 504, step 51, step 、, step 522 (FIG. 5) realizes the selection device. Function, the function of the calculating device is realized by the processing of step 526 (Fig. 5), and the function of updating the device is realized by the steps of %%', step 508, step 5U~step "o, step 524 (picture)" And performing the functions of the optimization device by the processes of step 4〇2 to step 43〇 (FIG. 4), steps 5〇2 to 504, step 51〇, step 512, step 522, and step (FIG. 5), By step 5〇6, step 5M, 55

According to the AIC of several correct modes with different degrees, according to the AIC of the parameter regression mode (mode Μ, M', etc.), the arrangement of the plurality of illumination areas SAP on the wafer W is selected to select the most correct regression mode. . According to this, since the mode closest to the true mode can be selected, the position information of the center position of the plurality of irradiation areas SAP is calculated with good precision. Further, according to the present embodiment, when a plurality of wafers W are sequentially detected, the mode parameters and the number of effective samples of the subroutine 3〇9 are successively optimized for each wafer. Further, in the subroutine 309, according to the coefficient of the regression mode (the above equation (6), the regression pattern of the plurality of irradiation regions SAp obtained by sequentially detecting the position information of the irradiation region SAP of the wafer w) The pre-existing knowledge, and the measured position value (h) of the wafer mark Mi k attached to the plurality of irradiation areas SAi in the plurality of irradiation areas SAP of the wafer W as the object of the position information detection in the stage coordinates 'The parameter freedom of the regression mode is optimized according to the AIC of the parameter degree of freedom corresponding to the regression mode. And determining the effective number of samples β attached to each of the illuminations of the wafer w to be the next position information detection target according to the optimized parameter degree of freedom, such as the measured value of the measured value of the bb circle of the SAP field. Since the wafer mark Mi, k bit 56 1377598 can be used to increase or decrease the effective number of samples according to the optimized parameter degree of freedom, it is possible to detect each irradiation area SAP in a short time and with good precision. The base position. Further, according to the exposure apparatus 100 of the present embodiment, since the reference position of the plurality of irradiation areas SAP in the original step is detected with high precision, it is possible to realize superimposed exposure with short time and high precision. Further, in the above embodiment, AIC' is used for comparison of modes having different degrees of parameter freedom, but other criteria may be used. For example, Bayesian Information Criterion (BIC) may be used. The BIC of the above statistical model is expressed by the following formula. [Expression 12] ' -2 * logZ + d - logw - (12) As shown in this equation, the second term of the compensation term of BIC is different from Aic. That is, the second item of AIC is 2d, whereas 眺 is the effective number of samples of d·-n. Therefore, compared with the Xiao AIC, the BIC is set to have a large compensation, and is set to a tendency to reduce the statistical mode in which the number of selection parameters is large. In addition, both 'AIC and BIC are derived as a generalization error or asymptotic logarithm of the pattern to the parent group when the number of samples is sufficiently large in the regular mode. Since the linear regression model is a regular pattern, it is appropriate to use these evaluation specifications. However, in the above embodiment, since the throughput is small, the number of samples used for the analysis of 2 is small. Therefore, the application of the above-mentioned gradual theory is not necessarily the second. It is preferable to define the logarithmic generality of the complement of the above-mentioned AIC and mc as follows, and select the best mode. [Equation 13] 57 1377598 PL = logL-P (m〇delnyd tune here, the logarithm of the compensation of the PL system)

The reason for the opposite of BIC is that the symbol of the item is AIC ’, and the formula i is used to evaluate the larger PL. The second item of PL·, the total A right, eight is good, the other is the same as d丨, ,, and AIC^BIC, and the right is cold (model, n) = 2 ίΛ红... 2, PL is For AIC, if = logn, it is BIC. P, 丨, η), the value of the point can be determined by simulation, etc., and can also be set as dependent on the number of shots (ie, type (ie, 昶1) and take ^(also η η). As a function of this function, the period # value in the remainder of the parameter increase amount can be used. For example, when comparing the first regression mode and the third regression mode, take the animal as 6 The coefficient is the expected value of the difference between the mode based on the value of the knowledge and the remainder of the parameter that is classified as the number of passes, and the value is 10,000. Thus, the most suitable mode In the remainder of the pattern before and after

If the end is small, the expected value of the above difference may be extremely large on the negative side, and the value of ^ K becomes large, and the mode is easy to be selected. In addition, although the linear regression mode is the positive shell I, this is a mode in which the self-sufficiency mode is evaluated, and the order of the parameters to be added (priority order) is determined in advance. Although there is a problem that the number of samples is small, in order to ensure the validity of the use of the AIC and the BIC, the above-described embodiment order determines the order of the parameters to be added in advance. When the parameter to be added can be freely changed, the regression mode is non-regular, and the Aic MG loses its validity. However, when using the logarithm of the logarithm with compensation, if the use of the non-regularity is used, In particular, in the above embodiment, the remainder is expressed by the price function E of 58 1377598 in the above equation (7), that is, by dividing the square of the error by the number of samples, but The present invention is not limited to this. For example, the remainder is an unbiased estimate of the remainder of the parameter deviation obtained by subtracting the parameter degrees of freedom from the number of samples, or the non-error square itself (since the square of the error itself increases or decreases according to the number of samples) Inappropriate), suitable for any of the remainder. Further, in the above-described embodiment, although the mode parameters in one batch are optimized, it is explained that, in the plural batches of the same process, it is also possible to know that there is a model coefficient. For example, when the processing is performed through the same batch, the prior knowledge of the parameters associated with the batch previously processed is stored in the memory device 47' for prior knowledge of the parameters of the batch. Further, in the above embodiment, as a reference to the mode (4), the optimum mode is selected by comparing the mode whose parameter degree of freedom is smaller than the mode and the degree of determination of the three modes larger than the mode. :: It is not limited to this. It is also possible to compare more than four modes. For example, = above mode calculates the value of AIC, and selects the mode with the smallest value of 5τ, 隹 can week - &lt; chess type, when In the same manner, the parameter free 旰 and the mode smaller than the real 2 are set to the mode M, and the == mode is set to the mode M' and the mode, and the optimal IS complex mode of the mode is plural, so in the step 51. Judgment: Since the third regression mode of the third regression modulus finds the value of aic two, respectively, the complex can be used. Select the model with the smallest value of 59 1377598 and the above embodiment is In addition to the 6-point EGA parameter, and the optimization of the model of the illumination component is considered, the present invention is not limited to this model. For example, it can be applied to not only the primary component of the irradiation region arrangement but also Considered more than 2 times at the same time The mode of the higher order component. This mode is a mode in which (DX. DY) of the above formula (4) is replaced by (dx, DY,) of the following formula. [Expression 14] rDX'^ rDX + α, · (DX)2 + α2 (DX DY)2 + α3 · (DY)2 + · · N, sink + A · (like O2 + A · (form · Dr) 2 + is . (train) 2 + ... ( 14) In the above mode, except for the above six EGA parameters, the coefficients αι,%, α3,...'β1} β2,β3,... are unknown parameters. Because of this parameter, it is generally considered that the variation within the batch is also It is small, so it is preferable to classify it as d · or e · When considering the above-mentioned high-order component, the mode containing the above-mentioned high-order component and the irradiation used in consideration of the above embodiment can be used. The composition of Buddhism's or the pattern containing both the high-order component and the irradiated component is divided into the third regression mode. In the above embodiment, the EGA regression mode is based on the above formula (6), but It may be a mode according to the above formula (4). Alternatively, it may be a mode according to the above formula (2) which uses the orthogonality error W and the residual rotation error Θ as coefficients. In this case, although the residual rotation error Θ is preferably classified as a., since the orthogonality error W is dependent on the exposure device to expose the layer of the original process, it is considered to be within the batch. The change is small, so it is more appropriate to classify it as C· or d· 60. When the value of the mode parameter with this orthogonality error... as the coefficient is significantly different, the measured value of this mark =: exposure : The device and the exposure device of the past wafer are different. The high-order component and the internal component of the irradiation may also be used as parameters. Thus, the number of passes = the coefficient of the off, unconditionally related to the change of other coefficients, the change of the coefficient, the non-* iceer, if the correlation is used, this optimizes the mode parameters. Efficient. Pity ^'47 In the above embodiment, 'the average value of the coefficient value stored as the prior knowledge, S is the value of the (four) number, but: the best: the number of the coefficient that can be obtained from the past The change of the parameter values: (most... smoothing by low-pass processing, etc.), etc., to predict the coefficient value of the wafer to be processed at this time. In the embodiment, the coefficient of the GA mode is used for all h wafer marks, but for the h wafer marks, the t position can be regarded as deviating from the illumination arrangement pattern, and the mark can be used. The measured values are excluded from the results. However, after exclusion, the effective number of wafer marks = the number of parameters must be greater than the parameter's degree of freedom. As a method of excluding the actual measurement value, various methods can be used, and the method of the above (4) can also be applied. Furthermore, 'the measurement value is not excluded at all, and the probability distribution function showing the constant value distribution of each measurement value or the mixed distribution pattern showing the linear combination of the probability distribution of the degree distribution function is considered. The normality of the measured value, the degree of abnormality, and the above optimization may be performed. 13775卯 In the above embodiment, VIII B. You are very embarrassed by the above criteria (with the compensation of the double degree) to judge the choice mode and the description of the mode and the mode. According to the judgment of this criterion, although it can be said that it is the most important point in the standard, it is the "sexual judgment" of the 2, but the wafer is correct to the accuracy of the alignment. The reduced remainder is the final objective M. In the case of the second, for example, even if the modulus value is selected, /P ^ is set at the value of the evaluation function £ in the mode M. When the value of E is greater than the value, Whoever ^ .Λ , Έ. 进步 Improve the additional measurement of the wafer irradiation by sampling and irradiation, re-execute the mode Μ and Μ, _, or can be set with the mode Μ and mode Μ, a ^ _ _ , Β _ The new mode of 'in the new mode is based on the evaluation of the logarithm of the compensation logarithm and the remainder. According to the second paragraph: in the embodiment, the mode is selected from the mode of the parameter degrees of freedom, and the optimal mode is selected according to the logarithm of the compensation of each chess type, and is not limited thereto. The coefficients of the mode—and the value obtained, the value of the coefficient of the model is also the value of the pre-knowledge of the coefficient of the formula = two and according to the modulo., AIC (M) and AIC (M,) The ratio of the ratio of the weighted average is taken as the value of the coefficient. that is,

This ^' can be regarded as a value that will be corrected based on the value of the prior knowledge and based on the measured value of the wafer mark irradiated by the sample. In addition, ^艮 is determined by adding the $(four)W remainder E to the heart, and calculating the 桓h in the mode of the μ-, -- numerical formula, and the pattern Μ, the AIC (M, h = ' Wide (5) to compare, choose the best mode among the three modes: the number of each of the 1st, when the above weighting is taken, 'in response to the change of the parameters of each wafer', it is better to use the compensation variation of the parameter increase. The criterion of small 62 1377598. In this regard, the use of AIC is better than the use of bic. In the above embodiment, the coefficient of the mode of the EGA mode is classified as a. to e., but is not limited thereto. For example, It can be classified into various factors such as the coefficient of estimation and change of each process, the coefficient of estimation and change of the product type, or the coefficient of periodic estimation and change. In addition, priority is given to the parameters classified into the same norm t. In this order, the parameter degrees of freedom may be slowly increased or decreased. In addition, the coefficients of the EGA mode norm are not specifically classified, and all coefficients are given priority order, and the order is 2 rows of parameter degrees of freedom. Of course, it is also possible to increase or decrease. In the case of a position, if the coefficient also includes a high-order component, it is preferable that the higher the number of times, the higher the priority is (the setting is easier to use the prior knowledge). Moreover, the present invention is effective in the parameter degree of freedom of the mode to be evaluated. The number of samples is used to measure the wafer mark of the sampled illumination. In the above embodiment, the sampling of the effective number of samples corresponding to each degree (4) is obtained for a plurality of modes in which the degrees of freedom of the parameters are different: the circle 5 Number and configuration, save it to the memory device 47, u* _L. n± «Select the most correct mode, that is, read from the memory device 47, the parameter degree of freedom satisfying the ^ ^ # A ^ ^ / mode Corresponding to the appropriate sampling quantity of the appropriate number of samples and the number of configuration-related μ $ 日 day mark and configuration related information 'will read the content to update the sampling information on the memory. In the above, according to the method of forming and configuring the related wafer mark information according to the established effective number of samples, various methods can be employed. In the same manner, as shown in FIG. 2, the above embodiment No In the circular 'storage I 4 wafer mark, the group circle mark when h wafer marks are selected is 4 horse 4NCh. Therefore, the expectation value and the mark of the overlap error of each of the estimated total irradiation areas are 63. Disperse, the combination of the estimated value is (small) as the number and arrangement of the sampled irradiation in the effective number. &amp;, the expected value of the stacking error in the total irradiation area and the dispersion of the specimen, The highest probability estimation value of the coefficient of the selected mode, but the highest approximate value of each coefficient of the mode can be obtained by using the design position of the wafer mark and empirically determined. The standard deviation of the error distribution between the design and the measured position is determined by a so-called normal equation which is formed by the conditional expression in which the evaluation function is the smallest under the least squares method. In the above embodiment, the number and arrangement of the sampled illumination are obtained by the above method before the exposure operation. However, in the subroutine 309 in the exposure operation, the highest estimation method is performed, and the sampling is performed immediately. The number of illuminations and the configuration can be optimized. Further, in the above-described embodiment, the alignment sensor AS is an FIA-type alignment sensor, and as described above, the laser beam is irradiated onto the dot-line alignment mark on the wafer w, and the mark is used. An alignment sensor that detects the LSA (Laser Step Alignment) mode of the mark position by diffracting or scattering light, and an alignment sensor that appropriately combines the alignment sensor with the FIA method described above is also applicable. In the present invention. Further, for example, the alignment sensor that irradiates the detected light to the detected surface, and the two diffracted lights (for example, the same number of times) generated by the mark interfere with each other to detect the alignment sensor alone or in combination with the FIA method described above. An alignment sensor that is suitably combined, such as the LSA method, is of course also applicable to the present invention. Further, the alignment sensor may be an on-axis (e.g., 64 1377598 TTL (Thr〇Ugh The Lens) method, etc.). Furthermore, the alignment detecting system is not limited to the detection of the alignment mark substantially stationary in the detection field of the alignment system, or the detection light irradiated by the alignment detection system may be opposite to the alignment mark. The manner of movement (for example, the aforementioned LSA system, and the homodyne (H〇m〇dyne) LIA system, etc.) ^ When using the method of moving the detection light and the alignment mark relatively, it is preferable to enable the relative movement direction. The direction of movement of the wafer stage WST when detecting each of the alignment marks is the same direction. Further, in the above embodiment, the present invention has been described in the case of a scanning type exposure apparatus of a step-and-scan type, but the scope of application of the present invention is of course limited thereto. That is, it is also very suitable for use in Step &amp; Repeat mode, Step & Stitch mode, Mirror projection, Aligner, and optical copying machine (ph〇 T〇repeator) and so on. Further, the projection optical system PL may be either a refractive system, a refractive system, or a reflective system. Further, the light source of the exposure apparatus used in the present invention is a KrF quasi-molecular laser and an ArF excimer laser &amp; laser, but may be a pulsed laser light source of other vacuum ultraviolet regions. In addition, as the illumination light for exposure, a single-wavelength laser which is an infrared region or a visible region from the DFB semiconductor laser or the laser laser oscillation may be used, for example, doped 铒 (or 钵 纪 两者The fiber amplifier is amplified and converted to the higher harmonics of the ultraviolet light using a nonlinear optical crystallographic wavelength. In addition, an illumination optical system, a projection optical system, and an alignment detection system AS composed of a plurality of mirrors are assembled to the exposure apparatus body, and optical adjustment is performed 65, and a reticle stage and a wafer carrier composed of a plurality of mechanical parts are mounted. The stage is attached to the main body of the exposure apparatus, and the line and the piping are connected, and further comprehensive adjustment (electrical adjustment, operation confirmation, etc.) is performed, that is, the exposure apparatus of the above embodiment can be manufactured. Further, the manufacture of the exposure apparatus is preferably carried out in a clean room in which the monthly b is managed in m•degrees and cleanliness. Further, the present invention can be applied not only to an exposure apparatus for manufacturing a semiconductor element, but also to an exposure apparatus for transferring a component pattern onto a glass substrate, which is used in the manufacture of a display including a liquid crystal display element, and the like. An exposure device for transferring a component pattern used on a ceramic wafer, and an exposure device for manufacturing a photographic element (CCD, etc.), an organic EL, a micromachine, and a DNA wafer. Further, not only a micro device for manufacturing a semiconductor element or the like, but also a circuit pattern or a photomask used for manufacturing a photo-exposure device, an EUV exposure device, a calender exposure device, and an electron beam exposure device, and transferring the circuit pattern The present invention is also applicable to an exposure apparatus such as a glass substrate or a tantalum wafer. Here, an exposure apparatus using DUV (deep ultraviolet ray) or vuv (vacuum ultraviolet ray) light or the like generally uses a transmission type reticle, and as a reticle substrate, quartz glass, fluorine-doped quartz glass, fluorite is used. , or crystal, etc. Further, in the X-ray exposure apparatus or the electron beam exposure apparatus of the proximity type, a transmissive mask (stencil mask, membrane mask) is used as the mask substrate. Round and so on. Further, the position detecting method of the present invention is not limited to the exposure apparatus, and any one of a plurality of types of marks which are required to be formed from an object, and a plurality of means for detecting and detecting them can be applied. <<Element Manufacturing Method>> Next, an embodiment of a manufacturing method of the element 66 1377598 using the above-described exposure apparatus 1 in the lithography process will be described. Jgj 6 ' is a flowchart showing a manufacturing example of a display element (a semiconductor wafer such as 1C or LSI, a board, a CCD, a 菹-deposited film, a micro-machine, etc.). As shown in Fig. 6, the first student is in the first step. In step 601 (design step), the component (for example, the circuit design of the +conductor component) is performed, and the function is performed by the xenon lamp for real pattern design. . Next, in step 602 (mask manufacturing step), a reticle with a designed circuit pattern is formed. On the other hand, in the step (manufacturing step), a wafer is produced using a material such as tantalum. ~ In the step 604 (substrate processing step), using the mask and wafer prepared in step 6〇1: Π3, as described later, the actual circuit is formed on the wafer by lithography on. Then, in step 605 (component, mounting step), the component group is processed using the wafer processed in step 604. In this step 605, a process such as a cutting process, a bonding process, and a packaging process (chip package) are included as needed. Finally, in step 606 (inspection step), the inspection of the action confirmation test, the endurance test, and the like of the component made in step 6〇5 is performed. After such a process, the components are completed and can be shipped. . Fig. 7 is a detailed flow of the above-described step 604 when a semiconductor element is shown, an example of which. In Fig. 7, in step 611 (oxidation step), the wafer surface oxygen is formed in the step 612 (CVD step) to form an insulating film on the surface of the wafer. In Ray = 613 (electrode forming step), a sum is formed on the wafer by steaming. In step 614 (ion implantation step), ions are implanted into the wafer. 1 Upper = Each of steps 611 to 614 constitutes a pre-processing of each stage of the wafer processing, and the process is selected and implemented in accordance with the required processing in each stage. In each stage of the wafer fabrication process, when the above pre-treatment process is completed, that is, 67 1377598, the post-treatment process is carried out in the following manner. The post-treatment process, t first, in step 615 (resistance forming step), applying a sensitizer to the wafer. Next, in step 6, the exposure step, the circuit pattern of the photomask is transferred onto the wafer by the lithography system (exposure device) and the exposure method described above. Next, in step 617 (development step), the exposed wafer is developed, and in step 618 (step (4)), the exposed portion (4) remaining outside the photoresist portion is removed by (d). Then, in (4) 619 (Kawasaki's step-by-step), remove the unnecessary photoresist. Multiple circuit patterns are formed on the wafer by repeating these pre-processing and post-processing processes. According to the element manufacturing method of the present embodiment described above, since the exposure apparatus 1 to which the position detecting method of the above-described embodiment is applied is used in the exposure process (step 616), high-precision exposure can be realized. As a result, it is possible to manufacture an element having a higher degree of integration. The position detecting method and apparatus of the present invention are very suitable for detecting positional information of positional alignment of an object. Further, the exposure mounting method of the present invention is not suitable for the lithography process for manufacturing a semiconductor element, a liquid crystal display element or the like. Furthermore, the component manufacturing method of the present invention is very suitable for the production of microcomponents. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a view schematically showing the configuration of an exposure apparatus according to a first embodiment of the present invention. Fig. 2 is a view showing an arrangement example of an irradiation region on a wafer and a wafer mark attached thereto. Fig. 3 is a flow chart showing the processing steps of (10) when the exposure apparatus according to the embodiment of the present invention performs the 68 exposure processing and the main fourth diagram. Figure 5 is a flow chart of the alignment process (hereafter. Fig. 6 is a flow chart of the process of the circle alignment process (the flow chart of the small state thereof), the implementation of the component manufacturing method of the present invention Figure 7 shows the steps of Figure 6 [Main component symbol description] 10 Lighting system 15, 17 Moving mirror 16 Screen interferometer 18 Wafer interferometer 19 Stage control device 20 Main control device (Selection device 22 24 Optimisation device, decision loading. Marking line alignment detection system Wafer stage drive unit 25 Wafer holder 46 Drive unit 47 Memory unit 100 Exposure unit AS Alignment detection system (Measurement FM reference mark board) IL Illumination% ' 69 1377598 Μρ,κ Wafer Marking PL Projection Optical System R Marking Sheet RST Marking Line Stage SAP Irradiation Area W Wafer (Object) WST Wafer Stage

70

Claims (1)

1377598 ψ年】5Selection ΜReplacement page amount No j Patent application scope: Several bar mV detection method' is used to detect the position information of multiple &amp; · The first course is based on the established coordinates of at least one shirt. 〜 疋 疋 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 该 刀 刀 刀 刀 刀 刀 刀 刀 刀 刀The measured value is used to determine the mode associated with the configuration of the plurality of zones; and the second process is to calculate the location information of the plurality of zones according to the mode of the F+ v 疋 疋; The two-first process includes determining, according to the position L of the at least-partially-divided area, a degree of freedom for determining a parameter of the location information of the position information and a mode, and according to the parameter of the determined mode, To determine the process of adding the measured zone area. The utility model is characterized by detecting a complex configuration on an object, comprising: 2. a position detection method, position information of a plurality of division areas, and a first process, using a plurality of divisions in the plurality of division areas in the established coordinate system The actual position of the location information of the region is calculated for each of the plurality of zoning regions with different degrees of freedom of the parameter, and the value of the predetermined evaluation specification of the model is calculated, and the calculation result is used as the plurality of zoning regions. In the regression mode, the mode determined to be the best in the evaluation specification is selected; and the second system is based on the regression mode of the selection to calculate the position information of the plurality of division regions. 3 · As in the position detection method of the second application patent scope, wherein the step-by-step process includes the 帛3 process, the process is reflected in the evaluation specification as the best mode 71 ^ 77598 100 May 2 correction replacement page a prior knowledge related to a coefficient of a parameter value pattern for specifying a regression mode selected by the regression equation; forming a plurality of the objects in the plurality of division regions, and sequentially changing the object as the position information detection object The first process, the second process, and the third process. For example, the position detecting method of the third item of the patent scope, wherein the % system updates the number of samples of the real side value of the position information of each of the divisional areas, so that it is used to evaluate the best mode as the evaluation specification. A sufficient number of correctness levels for the selected regression model. For example, the third process of the T-slip patent scope is the first process of the ith parameter model whose parameter degree of freedom is a predetermined size, and the degree of freedom of the i-th regression mode is smaller than the second regression mode of the i-th regression mode, and the 1 free f is greater than the first regression. In the third regression mode of the mode, select the best mode of the price specification. x ° • The mouth position requires the position detection method of item 3 of the patent scope, the first process, in comparison with the value of the parameter from ', ~ the evaluation specification, the degree of freedom of the early parameter of the formula i is smaller than the first regression In the case where the value of the flat-type specification of the second regression mode is relatively good, 俜^ and gong are the best modes of the evaluation specification; the formula of the first regression mode is good, and the i 锜 锜 ... 6 6 6 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 亥 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第 第The value of the specification is better than the second formula, and the number is! The second regression model of the backs is like the chess type. The general value is worse than the established value. 72 1377598 1〇〇 May 2D曰 Correction replacement page „ „ j j, by 朁 朁 : : : : : : : : : : : : : : : : The third regression model of the 1st regression model is the best model for the S Hai Ping price specification. 7·If the position detection method of the sixth application of the patent scope is the first process, the third regression mode is ', When VA ^ ^ 坷 式 is selected as the most common mode of the evaluation specification, it is further secretive, and the zoning area of the / / J 延 罝 罝 俾 俾 俾 俾 俾 俾 俾 俾 俾The number of samples of the measured value only becomes a sufficient number for the degree of correctness of the third regression mode. = The measured value of the positional information of the added zoned area is measured and the third regression mode is extracted. Parameter value: , : · If the position detection method of item 7 of the application scope is in the case of the second item, and the third regression mode has a plurality of cases, the evaluation rule is selected. Xi ^ ^ as "... for a good 3rd regression model, come Floor price for the best mode of specification. , 9: If the position detection method of item 5 of the patent application scope is predicted to be the value of the value between the objects, the regression pattern of the first, second, and third styles (4) The number; the coefficient 'is a small number of changes in the value to be predicted between objects, and is used as a parameter in the third regression mode, and the other is based on the prior knowledge. The constant will be predicted in all objects Simplified, Μ '. ^ value is roughly the same regression coefficient, in the hi, the return mode as a constant according to the formula, the other side, in the .... In the third round of knowledge, it is decided to do 〇 如 如 如 如 如 如 如 如 申请 申请 申请 申请 申请 位置 。 。 。 。 。 。 。 。 。 。 。 。 In the method of detecting a position information, the processing unit of the plurality of objects of the object 73 ^//^98 ^//^98 100 May 2D 曰 correction replacement page number is not satisfied with the predetermined number, The coefficient that is predicted to be a large change in the value between the processing units is used as the number in the third regression mode, and the other is the constant determined according to the second regression mode; When the number of processes exceeds a predetermined number, the coefficient is used as a constant determined based on the prior knowledge in the second regression mode, and the third regression mode command is used as a parameter. η· The position detection method of the second item of the patent scope is as follows, wherein the probabilities of the regression modes, the injury 6 6 6 according to the self-conversion mode, the predetermined seat 4 shows the area of the area The position information of the North Coast and the remainder of the measured value of the location information are estimated. 12. The method of position detection according to item u of the patent scope, wherein the parameter "for the respective fork type" includes a deviation indicating an alignment coordinate system defined by the configuration of the predetermined coordinate system: the plurality of division regions i = a number, a number, a coefficient showing a design value deviating from the components of each zone: at least a part of the coefficient of the higher-order component associated with the configuration of the plurality of zone regions ❶ ι τ' = The position detection method of item 2, wherein the = quotation specification sets "the criterion for adding the probability of compensation, and the parameter degree of freedom of the corresponding mode." The faculty system will request the position detection method of the 13th patent (4), among which The criterion is taken as the Chichi Shen-Zhe. In the... and the Bayesian Information Criterion, the mode with a smaller value is used as a good model. 15·If the position detection method of the 13th article of the patent application, the basin 74 1377598 is a larger modulo 16 that varies the complement. The plurality of regions are characterized by the first and the prior knowledge, and the plurality of regions are set according to the maximum degree of freedom of the back degree with respect to the record y Α Λ ^x 杈 杈 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数 参数The position information of the area arranged on the object is detected in sequence, and the method includes: a process, which is based on the coefficient of the regression mode associated with the plurality of division areas according to the position information of the object in sequence. In the case of the non-fixed cobalt system, as the object of the position information detection: the measured value of the position information of several zoning areas in the J area, and the established evaluation specification of the mode, so that the parameters of the regression mode are good. *1 The second process calculates the position information of the plurality of division regions of the object to be detected by the position information based on the return V mode ' having the degree of freedom of the optimized parameter; and the third air path According to the degree of freedom of the parameter to be optimized, the number of samples of each zone of the object to be detected by the lower-person position information is set to the number of samples of the measured value; / A plurality of the objects in the plurality of divisional regions, and the object to be detected as the position information is sequentially changed, and the 帛1 process, the 6th second process, and the third process are repeated. 17 . The position detecting method of item 16, wherein the approximate degree of the regression mode of the singularity is based on the position information of the zoning area in the predetermined coordinate system obtained from the regression mode, and the measured value of the position information is 75 丄 J / The remainder of / is to be presumed. The 100-year May 曰 correction replacement page 1 8 · If the patent application is used as the position detection method of the I-th circumference of each regression mode, wherein the coordinates of the plurality of divisional regions Deviation from the coordinate system defined by the broken -4m]: The forming factor and the display deviate from the two zones! The number, and at least part of the coefficient showing the ten values of the plurality of districts. The higher-order component related to the configuration of the &amp; domain will be the position detection method of the 16th patent scope of the application in May, and the evaluation criterion of the bean is set to the criterion with the probability of compensation, and the compensation system corresponds to the parameter of the mode. Degree of freedom. And the supplemental system 2〇·If the position detection method of the 19th article of the patent application is applied, and the 1st word of the standard is used as the Akaike information standard and the Bayesian information standard, the mode with the smaller value is good. Mode. For example, in the position detecting method of claim 19, the '(4)4 compensation value is set as an expected value of the degree of change in the degree of freedom of the parameter of the regression mode parameter, and a product of the degree of freedom of the parameter of the regression mode'. The larger mode is a good mode. . 22 · An exposure method is to sequentially expose a plurality of zoning regions arranged on an object to form a predetermined pattern in each zoning region, and includes: The method for detecting the position information of the plurality of divisional areas by the position detection method of the first and second items of the first, second, and sixth items; and moving the object according to the detected position information to make the respective The process of zoning area exposure. 76 1377598 May 2014 Correction Replacement No. 23 23—The manufacturing method of the components, the page is changed to: ', including the lithography process, which is characterized by the use of the application of the feces in the lithography process. , soil 24 · a type of position detection device method. The location information of a plurality of zoning areas, the feature configured on the object to be detected is: a measuring device for measuring the position information of a plurality of zoning areas of the predetermined locomotive; the plurality of zonings within the 糸In the area, the device is selected to use the measured position of the measured sound, the target, and the location information of each of the divisions in the region, respectively, and the number of degrees of freedom of the parameter 夂 Λ 侗F 丨ra are respectively the same a regression model of £ crystal^, calculating the value of the (four) mode, selecting the best mode in the evaluation specification as the ^ according to the nose result; and calculating the device according to the selected regression mode Location information of the zoning area. Plural 25. If the patent application scope is in the 24th position, it further has: 八Τ 吕己忆装置, which is used to store the gold 楣仔, and to regulate the prior knowledge of the coefficients of the refinement mode of several zoning areas; And the parameter of the back (fourth) type selected as the best mode in the "fourth" evaluation specification is reflected to the selection device stored in the memory device, and the regression mode is set according to the prior knowledge stored by the memory device. . 'Do not know 26 · If the application scope is 帛25帛, the memory device stores the position of the division area measured by the measurement device 77 丄J/ 100 May 2〇曰Replacement page The number of samples of the measured value of the Gongxun; the new device of the λ - ^ , is a sample of the measured value of the position information of each Q-zone stored in the memory device in the regression mode selected in the evaluation specification = nucleus number. A plurality of position detecting devices are configured to detect the position information of the area in the plurality of objects, and to detect each object in sequence, and have: &quot; fetch: i" is based on each sequential detection The position information of the object is the complex number: the month-related knowledge of the coefficient of the regression mode related to the zone, and the measured value of the position information of several zone zones in the zone of the location information in the established coordinate system, ::= The established evaluation specification, such that the parameter of the regression mode is based on the 最佳 = having the optimized parameter, and the plurality of objects that are the object of the position information detection are calculated. The position information; and the determining means determine the number of samples of the information I measured as the object of the next position information detection according to the degree of freedom of the optimized parameter. . 28 · - Type of exposure device' exposes a plurality of zoning areas arranged on the object in order to form a zoning in each zoning area, and has: (4), its special position detection device, the patent application scope is 24 Or 27; and a gastric transfer device, according to the detected position information, moving the object 78 1377598 Aug. 100, 100 days to correct the replacement page to expose the respective zone regions. 29. A method of manufacturing a component, comprising a lithography process, characterized in that: the lithography process uses an exposure device of claim 28 of the patent application. 3 〇 — 种 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置 位置The position information of the plurality of division areas; the second process is to perform the predetermined processing operation after the first process; and the third process is to measure the number of the area of the object after the second process, not in the process笙, the position information of the area is set; the plurality of divisions measured in the process. The optimal process is based on the measurement result of the first process, and the processing action for exploring the best mode is performed as a disk ΛΑ 勹 勹 茨 物体 object The processing mode related to the plurality of zoning regions, and the necessity to further enter the location information of the zoning region for the best mode of the circumstance 4 别~3: The third process is based on the second process The processing judges that the position of the zoning area that needs to be further located is 3 1 · The position detection dream system r, set the sage to detect the plurality of objects arranged on the object The location information of the area is characterized in that: the device in which the device is located is a position information of a plurality of zone regions in which the plurality of zones on the object are measured; and the device is disposed in the &amp; After measuring the measurement of the Jinluoji J device, the predetermined processing is performed. The measurement device is used to measure the object after the processing operation performed by the processing device. In the plurality of zoning regions, the position information of the plurality of zoning regions that are not measured during the measuring operation; the 〇 processing device is the root of the % 置 农 农 农 1 1 1 1 1 1 1 1 1 1 In the preferred mode of determining the position of the zoning area with the plural (four) lines on the object, the processing action of the next day is performed; the measuring device is determined according to the needs of the processing dream The position of the further step-by-step (4) is processed, and the wave is sent. The location of the zone of Beixun is the guest, the schema: as shown in the next page 80