JP2012038794A - Detection condition optimization method, program creation method, exposure device, and mark detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optimize alignment mark detection conditions.SOLUTION: An alignment detection system is used to detect an alignment mark (EGA mark or search mark) formed on a wafer under a plurality of illumination conditions and a plurality of imaging conditions. Then analysis processing of the acquired detection signal is performed using a signal processing algorithm and a determination amount related to a waveform shape of the detection signal is calculated (steps 302 to 310). Next, reproducibility of the detection results of a plurality of the marks is evaluated based on the determination amount (step 312) and the plurality of the illumination conditions and the plurality of imaging conditions are optimized based on the analysis result (step 314). This enables the alignment mark detection conditions to be optimized so as to improve the reproducibility of the detection results.

Description

  The present invention relates to a detection condition optimization method, a program creation method, an exposure apparatus, and a mark detection apparatus, and more specifically, detection condition optimization for optimizing detection conditions for alignment marks formed on a substrate. The present invention relates to a method, a program creation method using the detection condition optimization method, and an exposure apparatus and a mark detection apparatus suitable for implementing the detection condition optimization method.

  Conventionally, in the lithography process in the manufacture of microdevices (such as electronic devices) such as semiconductor elements and liquid crystal display elements, step-and-repeat projection exposure apparatuses (so-called steppers), step-and-scan projection exposure apparatuses ( A so-called scanning stepper (also called a scanner)) is used relatively frequently.

  When performing overlay exposure in this type of projection exposure apparatus, a circuit pattern already formed in each shot area on a wafer (or glass plate or the like) as a substrate to be exposed and a reticle as a mask to be exposed from now on Fine alignment of the wafer, which is alignment with the pattern, must be performed with high accuracy. As a conventional high-precision fine alignment method, for example, as disclosed in Patent Document 1, coordinates of a fine alignment mark (wafer mark) attached to a predetermined number of shot areas (sample shots) on a wafer are disclosed. An enhanced global alignment (EGA) method is known in which the position is measured and the measurement results obtained are statistically processed to calculate the array coordinates of each shot area.

  When performing EGA fine alignment, search alignment is performed in advance so that the test mark is surely within the detection range of the alignment sensor. That is, by detecting the position of a predetermined search alignment mark (search mark) formed on the wafer, a rough shot arrangement of the wafer is obtained, and the wafer mark of each sample shot is determined based on this shot arrangement. It was positioned within the detection range of the alignment sensor.

  In the fine alignment described above, it is important to accurately measure the position information of the wafer mark (EGA mark). In search alignment, it is important to reliably detect a search mark. The mark structure depends on the semiconductor device being created by the projection exposure apparatus. Depending on the mark structure, a large signal waveform can be obtained and detection is easy, or conversely, only a small signal waveform can be obtained and detected. May be difficult. Also, especially in the case of fine alignment, even if the signal waveform can be detected, the signal waveform is small, so it is affected by the noise of the processing system and measurement reproducibility (position measurement accuracy) is insufficient. There is also. Especially in such a difficult situation, the detection conditions (measurement conditions) at the time of mark detection of the mark detection system (wafer alignment system (alignment sensor)) for detecting the test mark (EGA mark or search mark) are optimal. Is effective. Conventionally, at the time of optimization, when creating a process program (hereinafter also referred to as a recipe file) that defines the processing procedure of wafer alignment (search alignment or fine alignment), an engineer can optimize measurement conditions. It was done.

  However, with recent exposure apparatuses, the required alignment accuracy has become stricter. As a result, the types and sizes of alignment marks have been diversified, and the search marks have become smaller than before. For this reason, it is becoming more difficult for an engineer to optimize measurement conditions when creating a process program, and it is also difficult to optimize measurement conditions themselves. In addition, when manufacturing a semiconductor element or the like using a plurality of projection exposure apparatuses, it is also important that there is little difference in overlay accuracy between the plurality of projection exposure apparatuses. In this respect as well, it is becoming important to optimize measurement conditions and reduce error factors.

US Pat. No. 4,780,617

According to the first aspect of the present invention, there is provided a detection condition optimization method for optimizing the detection condition of a mark for alignment formed on a substrate, wherein a plurality of the marks formed on the substrate are Detecting under a plurality of illumination conditions and imaging conditions using a mark detection system; analyzing a detection signal from the mark detection system using a predetermined signal processing algorithm, and relating to a waveform shape of the detection signal Detecting a determination amount, evaluating reproducibility of detection results of the plurality of marks based on the determination amount, and optimizing the plurality of illumination conditions and imaging conditions based on the evaluation results A condition optimization method is provided.
In addition to such basic processing, the detection signal from the mark detection system may be subjected to a plurality of filter processing before the determination amount relating to the shape of the waveform is obtained, and the signal waveform may be deformed and applied.

  Here, the mark detection conditions are, of course, the illumination conditions (for example, illumination wavelength, illumination method, illumination numerical aperture, etc.) and imaging conditions (focus, imaging numerical aperture, aberration, etc.) of the mark detection system at the time of mark detection. In addition, various conditions associated with the processing of the detection result after the mark detection are also included. That is, the mark detection condition is a concept including various conditions related to mark detection until the result of position detection mark detection (position information) is obtained, and has substantially the same meaning as the mark measurement condition in a broad sense. . In this specification, the term “mark detection condition” or “detection condition” is used in this sense.

  According to this, it becomes possible to select (or determine) the optimal illumination condition and imaging condition that provide good measurement reproducibility based on the reproducibility evaluation result of the detection results of a plurality of marks. Therefore, it is possible to optimize mark detection conditions in a short time and / or efficiently.

  Here, the filtering process itself is not an essential condition, but a large improvement effect can be expected depending on the signal waveform.

  Here, the effect of changing the general illumination condition setting and changing the imaging condition setting will be described.

  For aberrations, it is not essential to change the normal conditions. For example, if the aberration changes when the illumination conditions are changed and it is effective to change a minute amount, it is necessary to change the illumination conditions. It is effective to change the aberration state. As described above, the mark structure depends on the semiconductor element or the like being created by the projection exposure apparatus. Depending on the mark structure, a large signal waveform can be obtained and detection is easy, or conversely, only a small signal waveform can be obtained. In some cases, detection is difficult. These conditions often have wavelength dependence because the structure of the semiconductor element often has a thin film layer or metal layer, and switching the illumination wavelength condition greatly improves the signal that was small at a certain wavelength. There is a possibility that. In addition to the line width of the line portion and the space portion, as elements constituting the alignment mark, the structure such as reflectivity and step differs depending on the manufacturing process of the semiconductor element, etc., so that generally the signal waveform is different. (Here, the step includes the phase difference component generated when the material constituting the mark is different between the line portion and the space portion and the optical path length of the light is different). In the FIA optical system, particularly in the bright field illumination system, the numerical aperture NA of the illumination light beam is set to be smaller than the numerical aperture NA of the imaging optical system (ratio of about 0.8). The setting of the optical system is called a partially coherent optical system. Depending on the structure of the alignment mark (the three elements described above), the image formation characteristic can be determined based on the defocus position (when the focus position is changed) rather than the initial focus position. However, there is a known feature that a larger signal waveform can be obtained. Therefore, in addition to the signal waveform at the initial focus position, the signal waveform at the defocus position is confirmed. If the signal waveform at the defocus position is better, it is desirable to change the focus setting. In addition, when different signal waveforms are obtained when the wavelength is switched as described above, the optimum focus setting position often changes depending on the wavelength. Therefore, when the wavelength is switched, the focus position is also changed. Checking the waveform is effective in improving measurement accuracy.

Moreover, the following effects can be expected when the bright field illumination is switched to the dark field illumination as the illumination condition change. First, in general, a dark field image is completely dark and invisible when there is no mark, and when the mark structure is close to a so-called phase pattern, the edge portion of the mark is an illuminated image. Therefore, an image in which the mark edge portion shines only when there is a mark from a completely dark state, and thus the obtained image is basically characterized by very good contrast. In the following contrast formula, the contrast increases because the minimum intensity is close to zero.
Contrast = (maximum intensity-minimum intensity) / (maximum intensity + minimum intensity)
On the other hand, however, dark field illumination tends to reduce the maximum intensity of the obtained image itself, and requires a light source with a bright luminance. Since the luminance of the light source has a limit determined by the light source itself, bright field illumination is more advantageous in that the signal intensity, that is, the maximum intensity is large.

  In addition to the line width of the line portion and the space portion, the structure of the mark, as well as the line width and the space portion, and the structure of the reflectivity, step, etc. differ depending on the manufacturing process of the semiconductor element, etc. (Here, the level difference is different between the line part and the space part, and the phase difference component generated by the difference in the optical path length of the light is included). However, when these mark structures change in general, The signal waveform is measured by being shifted laterally due to the interaction with the aberration remaining in the optical system. In general, this lateral displacement is often called TIS (Tool Induced Shift). As aberrations of the optical system, in addition to the coma aberration of Seidel 5 aberration, the illumination light beam is deviated from the direction perpendicular to the wafer surface (generally, the telecentricity error of the illumination light beam: here called illumination telecentric), imaging optics Since the aperture stop of the system is decentered, there is an error that the illumination light beam reflected by the wafer does not pass through the center of the aperture stop of the imaging optical system (herein, this error is called vignetting). The coma aberration and the vignetting error remain in a very small amount even in a very good optical system, which becomes a problem when a very small position measurement accuracy (for example, nm order) is required.

  As a method of adjusting these aberrations, there are a method of adjusting the lens of the imaging optical system by decentering it to generate a minute amount of coma aberration, and a method of adjusting the vignetting by decentering the aperture stop.

  In the case of recent state-of-the-art semiconductor exposure apparatuses, a technique is adopted in which the number of EGA measurement points is set to a large number (for example, 16 points) for one process wafer to improve the overlay accuracy. When there is an error in the mark structure of the alignment mark, each mark has a random displacement factor, and as a result, the measurement result of each EGA point varies. Here, such an error is referred to as a TIS random component in the sense of a position error caused randomly by each EGA mark in one wafer. If the TIS random component is large, even if the linear component is corrected and exposed by EGA measurement, a random component error (referred to as an EGA random error) resulting from the TIS random component remains, resulting in poor overlay accuracy.

  It is not only the TIS random component that induces EGA random errors. In general, an ideal fine measurement mark structure of the alignment mark is a line-symmetric structure with respect to the mark center line. For example, a plurality of line portions are line symmetric with respect to the mark center line. The reflectivity of the mark line part and the space part is axisymmetric with respect to the mark center line. The step structure of the mark (or the phase difference of the transparent layer) needs to be axisymmetric with respect to the mark center line. In recent years, the exposure line width has become very fine, and it is often necessary to precisely focus during projection exposure, so the wafer surface is often flattened by CMP (Chemical Mechanical Polishing) during the semiconductor manufacturing process, It has been pointed out that this step makes the alignment mark asymmetric with respect to the mark center line. It is also assumed that the degree of asymmetry changes between the wafer center and the periphery. The asymmetry of the mark structure induces measurement misalignment. In general, this lateral displacement is often called WIS (Wafer Induced Shift). When such asymmetrical variation of the mark structure occurs in the mark to be measured by EGA, a random position error is induced, and as a result, the EGA random component is deteriorated. In recent years, it has been considered that the generation amount of such mark asymmetry errors has also been reduced and improved. However, even a very good process wafer remains, and a very small amount of position measurement accuracy (for example, nm) It becomes a problem when an order is required. Here, such an error is referred to as a WIS random component in the sense of a position error caused randomly by each EGA mark in one wafer. If the WIS random component is large, even if the linear component is corrected and exposed by EGA measurement, a random component error (referred to as an EGA random error) resulting from the WIS random component remains, resulting in poor overlay accuracy.

Considering the above decision factors, determine which illumination wavelength setting is good, which focus setting is good, and whether bright field illumination or dark field illumination is advantageous based on the actual signal waveform Necessary, but the items that need to be determined here are the following three points.
1. Measurement reproducibility judgment 2. TIS random error factor determination Details of the WIS random error factor determination will be described later. With regard to measurement reproducibility judgment, after analyzing the noise characteristics of the image sensor used in the optical system and performing an optical teaching simulation in advance to obtain a conversion factor, it is characterized by using a predetermined signal processing algorithm from the signal waveform. Can be calculated numerically by extracting and converting.

  Optical teaching simulation is an A / D conversion by generating optical images of alignment marks with various assumed structures under various optical setting conditions (illumination conditions and / or imaging conditions), aberration settings, and focus setting conditions. This means the process of obtaining the signal waveform measured by the imaging device in a pseudo manner and calculating and evaluating various feature amounts and / or determination amounts in a pseudo manner using a signal processing algorithm similar to that of an actual apparatus. When obtaining a conversion coefficient for measurement reproducibility determination, a simulation is performed by adding noise obtained as a result of analyzing an optical image to an image sensor to the pseudo-generated signal waveform. Although details will be described in the embodiment, the conversion coefficient and the determination amount obtained in this way have a very good correlation with the measurement reproducibility.

  In addition, 2. In TIS random error factor determination, a plurality of marks in one wafer are measured, and a mark structure (line width, step, reflectance ratio between line portion and space portion) is determined from a signal waveform using a predetermined signal processing algorithm. The amount of error can be determined by extracting the corresponding feature and calculating the amount of variation within the wafer.

  Again, an optical image with aberration remaining in the optical system is generated in advance, an optical teaching simulation is performed, and the interaction between the aberration and the mark structure error is analyzed in advance. For example, a mark structure with a large difference in reflectance between the line part and the space part or a mark with a large contrast has a small interaction between the aberration and the mark structure error, and the mark with the largest edge slope at the initial focus position It turns out that the interaction between the aberration and the mark structure error tends to be small. Whether such a tendency is present or not depends on the mark structure determined by the manufacturing process of the semiconductor element or the like. From the result of the optical teaching simulation, it can be seen that the actual TIS random generation amount tends to be small in the mark in which the interaction between the aberration and the mark structure error tends to be small even if the mark structure variation in the wafer is large. In this way, if the simulation results are stored in a database, for example, one wafer is measured, the difference in reflectance between the line portion and the space portion and the contrast are determined, and the most suitable edge slope for measurement is obtained. If the difference between the increased focus position and the initial focus position is determined, the weight for the interaction between the aberration and the mark structure error can be determined for each mark error (line width, step, reflectance ratio between the line portion and the space portion). . Therefore, a TIS random error can be predicted by taking the sum of variations in the wafer of each mark error in terms of the weight for the interaction between the aberration and the mark structure error.

  3. In the WIS random error factor determination, a plurality of marks in one wafer are measured, the position is first measured to obtain a mark center line, and then, as a characteristic of the signal waveform, an extreme value of the waveform obtained by differentiating the signal waveform ( That is, the position, differential value, and signal strength of the differential maximum (minimum value representing the edge slope), the signal strength, and the like are obtained using a predetermined signal processing algorithm. Edge slope positions that are approximately line symmetric with respect to the mark center line are extracted as feature values of the pair, and the deviation from line symmetry for each pair, that is, the edge slope position, the line symmetric pair average value and the mark If the difference value from the center and the difference value of the asymmetric pair are simply obtained with respect to the differential value and the signal strength, it becomes a determination amount representing the degree of asymmetry. Among these determination amounts, for example, the WIS random error can be grasped by calculating the asymmetry of the edge slope position with respect to each mark in one wafer and seeing the variation of the asymmetry value. Moreover, when performing a filter process, the improvement possibility can be expected for the above three determination items.

  The present invention provides: However, TIS random error factor determination and WIS random error factor determination may be used in combination as appropriate to make a comprehensive determination.

  According to the second aspect of the present invention, there is provided a program for optimizing the detection conditions of the detection signals of the plurality of marks using the detection condition optimization method of the present invention and determining the alignment procedure of the substrate. A method for creating a program to be created is provided.

According to a third aspect of the present invention, there is provided an exposure apparatus that forms a plurality of patterns on the substrate in an overlapping manner, and a mark detection system that detects an alignment mark formed on the substrate; A detection signal obtained by detecting a plurality of the marks under a plurality of illumination conditions and imaging conditions using the mark detection system is analyzed using a predetermined signal processing algorithm, and the waveform of the detection signal is related to An optimization device that obtains a determination amount, evaluates the reproducibility of the detection results of the plurality of marks based on the determination amount, and optimizes the plurality of illumination conditions and imaging conditions based on the evaluation results; An exposure apparatus comprising the same is provided.
Here, before obtaining the determination amount related to the shape of the waveform, the detection signal from the mark detection system may be subjected to a plurality of filter processes to deform and apply the signal waveform.

  This makes it possible to optimize the mark detection conditions in a short time and / or efficiently.

According to the fourth aspect of the present invention, a mark detection system for detecting an alignment mark formed on the substrate; a plurality of the marks using the mark detection system, and a plurality of illumination conditions and imaging conditions A detection signal obtained by detecting the detection signal is analyzed using a predetermined signal processing algorithm, a determination amount related to the waveform shape of the detection signal is obtained, and the detection results of the plurality of marks are reproduced based on the determination amount And an optimization device that optimizes the plurality of illumination conditions and imaging conditions based on the evaluation results.
Here, before obtaining the determination amount related to the shape of the waveform, the detection signal from the mark detection system may be subjected to a plurality of filter processes to deform and apply the signal waveform.

  According to this, it becomes possible to select (or determine) the optimal illumination condition and imaging condition that provide good measurement reproducibility based on the reproducibility evaluation result of the detection results of a plurality of marks. Therefore, it is possible to optimize mark detection conditions in a short time and / or efficiently.

It is a figure which shows schematically the structure of the exposure apparatus of one Embodiment. It is a flowchart which shows the process algorithm of creation of the recipe file (process program) of alignment measurement (EGA measurement) performed by the main controller of FIG. FIG. 3A is a diagram showing the arrangement of a plurality of shot areas on the wafer and alignment marks attached to each shot area, and FIG. 3B is a diagram showing one shot area in FIG. FIG. It is a flowchart which shows the subroutine of step 108 (alignment mark detection process) of FIG. 5A shows an EGA mark, FIG. 5B shows a one-dimensional signal corresponding to the X component of the EGA mark, and FIG. 5C shows a one-dimensional signal corresponding to the Y component of the EGA mark. It is a figure for demonstrating normalization of a detection signal. It is a flowchart which shows the subroutine of step 110 (the feature-value extraction process with respect to a detection signal, and a measurement reproducibility calculation process) of FIG. 8A and 8B are diagrams showing typical gain functions of filter processing, and FIGS. 8C and 8D are diagrams showing detection signals before and after the filter processing, respectively. . 9A and 9B are diagrams (No. 1 and No. 2) showing the feature amount of the detection signal. FIGS. 10A and 10B are diagrams (No. 3 and No. 4) showing the feature amount of the detection signal. FIG. 11A and FIG. 11B are diagrams for explaining group processing of feature points. It is a figure for demonstrating the amplitude, contrast, and edge slope which are the determination amounts regarding a shape. It is a figure for demonstrating the procedure which comprehensively evaluates the determination amount regarding a shape and evaluates measurement reproducibility. FIGS. 14A and 14B are diagrams (No. 1 and No. 2) showing a correlation between a determination amount related to a shape and measurement reproducibility. It is a figure which shows the result (the 1) of optical teaching. It is a figure which shows the result (the 2) of optical teaching.

  Hereinafter, an embodiment of the present invention will be described with reference to FIGS.

  FIG. 1 shows a schematic configuration of an exposure apparatus 100 that can suitably implement the detection condition optimization method of the present invention. The exposure apparatus 100 is a step-and-scan projection exposure apparatus (so-called scanning stepper (also called a scanner)).

  The exposure apparatus 100 includes an illumination system IOP, a reticle stage RST that holds a reticle R, a projection unit PU that projects a pattern image formed on the reticle R onto a wafer W coated with a sensitive agent (resist), and a wafer W. A wafer stage WST that holds and moves in the XY plane, a drive system 22 that drives the wafer stage WST, and a control system thereof are provided. The control system is mainly composed of a main controller 28 including a microcomputer (or a workstation) that performs overall control of the entire apparatus.

  The illumination system IOP includes a light source composed of, for example, an ArF excimer laser (output wavelength 193 nm) (or KrF excimer laser (output wavelength 248 nm)), and an illumination optical system connected to the light source via a light transmission optical system. As an illumination optical system, as disclosed in, for example, US Patent Application Publication No. 2003/0025890, an illuminance uniformizing optical system including an optical integrator, a beam splitter, a reticle blind, and the like (all not shown) are used. In addition, the laser beam emitted from the light source is shaped, and this shaped laser beam (hereinafter also referred to as illumination light) IL is elongated in the X-axis direction (the direction perpendicular to the plane of FIG. 1) on the reticle R. The illumination area is illuminated with almost uniform illuminance.

  Reticle stage RST is arranged below illumination system IOP in FIG. Reticle R is placed on reticle stage RST. The reticle R is sucked and held on the reticle stage RST via a vacuum chuck or the like (not shown). Reticle stage RST can be finely driven in a horizontal plane (XY plane) by a reticle stage drive system (not shown), and is predetermined in the scanning direction (here, the Y-axis direction which is the horizontal direction in FIG. 1). Scanned in the stroke range. Position information of the reticle stage RST is measured by the laser interferometer 14 through the movable mirror 12 fixed to the reticle stage RST, and the measurement value of the laser interferometer 14 is supplied to the main controller 28. Instead of the movable mirror 12, the end surface of the reticle stage RST may be mirror-finished to form a reflective surface (corresponding to the reflective surface of the movable mirror 12).

  Projection unit PU is arranged below reticle stage RST in FIG. The projection unit PU includes a lens barrel 40 and a projection optical system PL including a plurality of optical elements held in the lens barrel 40 in a predetermined positional relationship. Here, as the projection optical system PL, a bilateral telecentric reduction system is used, and a refractive optical system including a plurality of lens elements (not shown) arranged in a direction parallel to the optical axis AXp (Z-axis direction) is used. It has been.

  The projection magnification of the projection optical system PL is ¼ as an example. For this reason, when the reticle R is illuminated with uniform illumination by the illumination light IL as described above, the pattern of the reticle R in the illumination area is reduced by the projection optical system PL, and is applied onto the resist-coated wafer W. The projected image forms a reduced image (partial image) of the pattern in a part of the exposed area (shot area) on the wafer W. At this time, the projection optical system PL forms a reduced image of the pattern in a part of the field of view (that is, an exposure area and a rectangular area conjugate with the illumination area with respect to the projection optical system PL).

  Wafer stage WST is driven with a predetermined stroke in the X-axis direction and Y-axis direction by a drive system 22 including a linear motor and the like, and also rotates in the Z-axis direction, the rotation direction around the X-axis (θx direction), and the rotation around the Y-axis. It is micro-driven in the rotational direction (θy direction) and the rotational direction around the Z axis (θz direction). Wafer W is held on wafer stage WST by vacuum suction or the like via a wafer holder (not shown).

  The position information of wafer stage WST is measured by a laser interferometer system (hereinafter abbreviated as “interferometer system”) 26 via a moving mirror 24 fixed to wafer stage WST. Is supplied to the main controller 28. Based on the measurement value of interferometer system 26, main controller 28 determines positional information (rotation information (yaw amount (rotation amount in θz direction), pitching amount (rotation amount in θx direction)) of wafer stage WST in the XY plane. , Rolling amount (rotation amount in the θy direction))) is measured. Note that the end surface of wafer stage WST may be mirror-finished to form a reflecting surface (corresponding to the reflecting surface of movable mirror 24). Further, instead of wafer stage WST, a first stage that moves in the X-axis direction, the Y-axis direction, and the θz direction, and a second stage that finely moves in the Z-axis direction, θx direction, and θy direction on the first stage. A stage provided may be used.

  The measurement value of the interferometer system 26 is supplied to the main controller 28, and the main controller 28 determines the position of the wafer stage WST in the XY plane (in the θz direction) via the drive system 22 based on the measurement value of the interferometer system 26. Control (including rotation).

  Further, the position and the tilt amount of the surface of the wafer W in the Z-axis direction can be determined by, for example, an oblique incidence type multi-point focus having a light transmission system 50a and a light reception system 50b disclosed in US Pat. No. 5,448,332. It is measured by a focus sensor AFS comprising a position detection system. The measurement value of the focus sensor AFS is also supplied to the main controller 28.

  On the wafer stage WST, a reference plate FP is fixed so that the surface thereof is the same height as the surface of the wafer W. On the surface of the reference plate FP, a reference mark used for so-called baseline measurement or the like of an alignment detection system AS described later is formed.

  An alignment detection system AS that detects the alignment mark formed on the wafer W and the reference mark is provided on the side surface of the lens barrel 40 of the projection unit PU. As an example of the alignment detection system AS, a kind of image-forming type alignment sensor that measures a mark position by illuminating a mark with broadband light such as a halogen lamp and processing the mark image. The FIA (Field Image Alignment) system is used.

  The detection signal of the alignment detection system AS is supplied to the alignment control device 16, and the alignment control device 16 performs A / D conversion on the detection signal and arithmetically processes the digitized waveform signal to detect the mark position. . Here, an image pickup device that receives an optical image of the alignment mark is used for the alignment detection system AS. However, in recent years, the image pickup device has been remarkably evolved, and A / D conversion is performed in the image pickup device (in the alignment detection system AS). The alignment control device 16 is supplied with a digital signal, and the alignment control device 16 often performs arithmetic processing on the already digitized waveform signal to detect the mark position. This result is supplied from the alignment controller 16 to the main controller 28.

  As the alignment detection system AS, in the present embodiment, a detection system that can change (or switch) at least the illumination condition and the imaging condition is employed. The illumination condition is changed by switching at least one of a wavelength selection filter (illumination wavelength), an illumination aperture stop (illumination numerical aperture), and an ND filter (neutral density filter). Illumination wavelengths include, for example, green (for example, wavelength 530-610 nm), orange (for example, wavelength 600-710 nm), red (for example, wavelength 700-800 nm), broad (for example, wavelength 530-800 nm), cyan (for example, 480-520 nm), And six wavelength bands of near infrared (770 to 850 nm) are prepared.

  As the illumination aperture stop, a plurality of circular stops (normal stop, small σ stop) having different diameters and a plurality of annular stops having different zone ratios are provided so as to be switchable. The plurality of annular stops includes an annular stop having a numerical aperture larger than that of the imaging optical system. In the present embodiment, for example, bright field illumination can be switched to dark field illumination by positioning an annular diaphragm having a numerical aperture larger than the numerical aperture of the imaging optical system in the illumination optical path. Even if dark field illumination is realized by constructing an imaging aperture stop with a ring-shaped light-shielding part so that it can be inserted into and removed from the imaging optical path, and using it together with an annular stop (illumination stop) good.

  Also, especially during fine measurement, the eccentricity of the illumination aperture stop with respect to the optical axis needs to be adjusted very precisely to prevent the light beam guided onto the wafer from falling (deteriorating telecentricity). Considering the positional deviation (eccentric deviation) between the circular light emitting unit during bright field illumination and the annular light emitting unit during dark field illumination, for example, as disclosed in JP 2007-42966 A, circular and circular A band-shaped double-circle aperture, and a light bundle that guides light with a light guide fiber and guides the light to the circular light emitting part, and a fiber bundle that guides the light to the annular light emitting part are configured as separate bundles, The incident light combined and guided to each fiber bundle may be switched.

  As the dimming device having the ND filter, for example, a device having a plurality of rotating plates is used. Here, each rotary plate is provided with a plurality of ND filters having different transmittances (light attenuation rates). The dimming device sets the transmittance (the dimming rate), that is, the illuminance, by controlling the rotation of the rotating plate and switching the ND filter. The imaging conditions include focus (focus offset), imaging numerical aperture, aberration, and the like. These conditions are adjusted, for example, by minutely driving the optical elements that constitute the optical system of the alignment detection system AS. The illumination condition and the imaging condition are changed by the alignment control device 16 based on an instruction from the main control device 28. In search alignment where the measurement accuracy is less than that of fine alignment, it is also possible to use an AGC (Auto Gain Control) function mounted in the alignment detection system AS. An alignment sensor having the same configuration as the alignment detection system AS (however, the number of wavelength selection filters is partially different) is disclosed in, for example, US Patent Application Publication No. 2008/0013073.

  Further, in the exposure apparatus 100 of this embodiment, although not shown, light having an exposure wavelength disclosed in, for example, US Pat. No. 5,646,413 is used above the reticle R. A pair of reticle alignment detection systems comprising a TTR (Through The Reticle) alignment system is provided, and detection signals of the reticle alignment detection system are supplied to the main controller 28 via the alignment controller 16.

  Next, creation of a recipe file (process program) for alignment measurement (EGA measurement) performed by the exposure apparatus 100 of the present embodiment will be described. Here, the processing algorithm executed by the main control device 28 (internal CPU) will be described with reference to other drawings as appropriate along the flowchart of FIG.

First, in step 102, a wafer is loaded onto a wafer holder (not shown) mounted on wafer stage WST using a wafer loader (not shown). Here, at least one reticle pattern, which is actually used for manufacturing a device, is transferred, and a wafer W on which a search alignment mark and a fine alignment mark transferred together with the reticle pattern are formed is loaded on a wafer holder. The As shown in FIG. 3 (A), in each shot area S _n on the wafer W, as an alignment mark, fine alignment mark (EGA mark) 32 and search alignment marks (search mark) 33 is formed ( for the fine alignment marks, to be precise, it is formed on the periphery of the scribe line of each shot area S _n). As shown in FIG. 3B, the EGA mark 32 is an uneven line-and-space pattern (multi-bar pattern) 32X formed at a predetermined pitch in the X-axis direction, the X-axis direction, and the Y-axis direction, respectively. ₁ , 32X ₂ , 32Y are included. Hereinafter, these three multibar patterns 32X ₁ , 32X ₂ , and 32Y are also referred to as a first X component, a second X component, and a Y component, respectively. The search mark 33 is a lattice-shaped pattern formed by superimposing multi patterns arranged at a predetermined pitch in the X-axis direction and multi patterns arranged at a predetermined pitch in the Y-axis direction. Normally, the search mark 33 is larger, but in FIGS. 3A and 3B, the size relationship and size of each mark are different from the actual one. The design parameters (shape, number, position, etc.) of the EGA mark 32 and the search mark 33 and the design parameters of the wafer W (size of the wafer W, layout of the partition area, etc.) are determined in advance. Is stored in a memory (not shown) prior to the optimization process (part of the recipe file creation procedure). In the following, both the EGA mark and the search mark are referred to as alignment marks, unless distinction is necessary.

  In the next step 104, search alignment of the wafer loaded on the wafer holder is performed. Specifically, for example, two search marks positioned in the peripheral portion substantially symmetrically with respect to the wafer center are detected using the alignment detection system AS. Detection of these search marks is performed while sequentially positioning wafer stage WST such that each search mark is positioned within the detection field of alignment detection system AS. Based on the detection result of the alignment detection system AS (relative positional relationship between the measurement coordinate origin of the alignment detection system AS and each search alignment mark) and the measurement value of the interferometer system 26 when each search alignment mark is detected, The position coordinate of the search alignment mark on the stage coordinate system is obtained. Thereafter, the wafer center deviation and the residual rotation error are calculated from the position coordinates of the two search alignment marks, and the wafer stage WST (or wafer holder) is slightly rotated so that at least the residual rotation error becomes substantially zero. This completes the wafer search alignment.

  In step 106, the light quantity (illuminance) is optimized. Specifically, the initial focus position (by using the alignment detection system AS while changing the illuminance (ND filter) in a broad illumination band (wavelength band in conventional illumination, for example, conventional illumination). The first sample mark of EGA (first mark of the first sample shot area) is detected at the alignment autofocus driving position), and the main controller 28 finds the illuminance at which the contrast of the obtained imaging signal is optimum. Thus, the illuminance, that is, the amount of detection light of the alignment detection system AS is optimized.

  In the next step 107, focus optimization processing (FFO: FIA focus optimization) is performed. As described above, the alignment mark structure depends on the semiconductor element or the like being created by the projection exposure apparatus. Depending on the mark structure, a large signal waveform can be obtained and detection is easy, or conversely small. In some cases, only a signal waveform can be obtained and detection is difficult. As an element constituting the alignment mark, in addition to the line width of the line portion and the space portion, the structure such as the reflectance and the step differs depending on the manufacturing process of the semiconductor element, etc. (The phase difference component generated when the material composing the mark is different between the line part and the space part and the optical path length of the light is different is also included). In the FIA optical system, particularly in the bright field illumination system, the numerical aperture NA of the illumination light beam is set to be smaller than the numerical aperture NA of the imaging optical system (ratio of about 0.8). The setting of the optical system is called a partially coherent optical system. Depending on the structure of the alignment mark (the three elements described above), the defocusing (when the focus position is changed) is more important than the initial focus position. However, there is a known feature that a larger signal waveform can be obtained.

  Further, as will be described in detail later, there is a characteristic that the measurement accuracy is better as the signal waveform is larger, the contrast is larger, and the edge slope (the angle component of the curve forming the waveform) is steeper.

  Focus optimization detects the alignment mark while moving the wafer stage in the Z direction, measures the edge slope maximum value (or contrast maximum value) at each focus position, obtains the curve distribution for the focus, This is a known process for detecting the focus position as the peak position. The focus position obtained as a result of the processing is stored as the optimum focus position.

  Similarly, the FFO measurement process is also performed in dark field illumination. In dark field illumination, in principle there is no difference between the initial focus position and the focus position detected by optimization, but the mark portion does not necessarily coincide with the wafer surface, depending on the alignment mark structure. Therefore, carry out. However, if it is known in advance that the error is small, it may not be performed.

  Thereafter, the process proceeds to a subroutine for alignment mark detection processing in step 108. Both the alignment mark detection process in this step 108 (subroutine process) and the detection signal feature amount extraction and determination process (subroutine process) using a predetermined signal processing algorithm in the next step 110 are performed. This corresponds to a process called an optical teaching process.

  In the subroutine of step 108, first, in step 202 of FIG. 4, the illumination wavelength of the alignment detection system AS is set or changed. The setting of the illumination wavelength here is to set (change) the wavelength bands sequentially in the predetermined order from the six wavelength bands of broad, green, orange, red, cyan, and near infrared. Done in As an example, it is assumed that broad is defined as the first. In this case, there is no need to do anything because broad is already set at the time of the light quantity optimization described above.

  In the next step 204, the illumination method and imaging conditions of the alignment detection system AS are set or changed. Here, the setting / change of the illumination method is realized by selecting and setting the illumination system aperture stop. Accordingly, setting / changing of bright field illumination, dark field illumination, etc. is included in addition to normal illumination and modified illumination. As described above, for example, when annular illumination is performed using an annular diaphragm having an illumination numerical aperture larger than the numerical aperture (NA) of the imaging optical system, dark field illumination is obtained.

  Further, as described above, the imaging conditions include the focus position, imaging numerical aperture, aberration, and the like of the imaging optical system of the alignment detection system AS. Therefore, although the imaging numerical aperture may be changed, in the following, for convenience of explanation, the setting / changing of the imaging conditions mainly refers to the setting / changing of the focus position of the imaging optical system.

  To set or change the illumination method and imaging conditions, set a combination of the illumination system aperture stop and dark field or bright field in a predetermined order, and further set a predetermined focus position for each combination in a predetermined order. This is done by setting in.

  Here, as an example, the main controller 28 switches between the bright field illumination and the dark field illumination by switching between the normal illumination ring and the annular illumination diaphragm. For the bright field illumination, the focus position is set to the optimum focus described above. The optimal focus position detected by the generalized FFO process, and one focus position on each of both positive and negative sides centered on the optimal focus position, the best focus position (so-called zero point), and both positive and negative sides centered on the best focus position For the dark field illumination, the optimum focus position detected by the above-described focus optimization FFO processing (if optimization is not required, the best focus position (so-called zero focus position) is set as it is. Point)), and the focus position of one point each on the positive and negative sides.

  Therefore, in the present embodiment, in this step 204, combinations of 25 illumination methods and imaging conditions are set for one illumination wavelength.

  In the next step 206, using the alignment detection system AS under the setting of the alignment optical conditions that are currently set, an alignment mark attached to a predetermined number, for example, 20 shot areas (here, EGA mark) is detected. Main controller 28 drives wafer stage WST in the XY directions to sequentially position and image a predetermined number of alignment marks attached on wafer W within the detection field of alignment detection system AS.

In the next step 208, the obtained imaging result is converted into a one-dimensional signal for each of the X and Y components of the alignment mark. For example, FIG. 5 to EGA mark 32 shown in (A), X component for each scan line _{L X1, L X2 32X 1,} 1 -dimensional signal and 32X _2, 1-dimensional signal of the Y component 32Y in the scanning line _{L Y} And are extracted. Alternatively, a one-dimensional signal obtained by adding the luminance values (pixel values) of the X components 32X ₁ and 32X ₂ for a plurality of scanning lines parallel to the scanning lines L _X1 and L _X2 or taking the average of the added values; the sum of the luminance values of the Y component 32Y for a plurality of scan lines parallel to the scan line L _Y, or a one-dimensional signal averaged for the added value, is extracted. In any case, a signal having a waveform as shown in FIG. 5B is obtained as a one-dimensional signal of the X components 32X ₁ and 32X ₂ , and as shown in FIG. 5C as a one-dimensional signal of the Y component 32Y. A signal with a simple waveform can be obtained.

In the next step 210, the obtained (detected) one-dimensional signal (hereinafter referred to as a detection signal) is converted into a detection light amount (amount adjusted by the AGC function in this embodiment) and a black level (signal level). The origin is normalized according to BL, and the detection signal (after normalization) is stored in a storage device (not shown) in association with the illumination condition and the imaging condition of the alignment detection system AS. For example, as shown in FIG. 6, the detection signal I is normalized as I ′ = (I−BL) / (I _Max −BL). Here, I _Max is the maximum gradation and corresponds to the limit signal intensity that can be detected by the alignment detection system AS. In order to improve detection accuracy, for example, when one alignment mark is imaged in a plurality of times, the detection signal is normalized in consideration of these.

  In the next step 212, it is determined whether or not the detection of the alignment mark at the illumination wavelength of the currently set alignment detection system AS has been completed for all combinations of the illumination methods and the imaging conditions that have been planned. If the determination is negative, the process returns to step 204 to change to the next combination of illumination method and imaging conditions, and the processing of steps 206 to 210 is performed on the changed illumination method and imaging conditions. Thereafter, the loop processing of steps 206, 208, 210, and 212 is repeated until the determination in step 212 is affirmed. If the determination in step 212 is affirmed, the process proceeds to step 214.

  In step 214, it is determined whether or not the detection of the alignment mark has been completed for all illumination wavelengths of the alignment detection system AS, in this embodiment six illumination wavelengths. If this determination is negative, the process returns to step 202 to change to the next illumination wavelength, and the loop processing from step 204 to step 212 is performed until the determination in step 212 is affirmed for the changed illumination wavelength. Repeat. Thereafter, the processes in and after step 202 are repeatedly performed until the determination in step 214 is affirmed. If the determination in step 214 is affirmed, the subroutine of the alignment mark detection process is terminated, and the process returns to step 110 of the main routine of FIG.

  Through the processing up to step 108 described above, detection signals for 27 × 6 = 162 kinds of illumination conditions and imaging conditions are obtained for one alignment mark.

  In step 110, that is, in the detection signal feature amount extraction and determination processing subroutine, first, in step 302 of FIG. 7, the alignment mark detection signal obtained by the alignment mark detection processing is analyzed using a predetermined signal processing algorithm. To set or change the analysis conditions. Here, the analysis conditions to be set or changed include a plurality of filter processes for shaping a detection signal such as noise removal, a combination of the plurality of filter processes, and respective filter characteristics of the plurality of filter processes. You can also The processing in step 302 is performed by sequentially setting (changing) these analysis conditions in a predetermined order. In addition, as a normal process, you may perform the analysis which does not perform the filter process of step 304. FIG. In this case, no filtering is set in step 302, and no processing is performed in step 304.

  Although not an indispensable item in this embodiment, when performing a filter process, it implements as follows.

  As the plurality of filter processes, for example, a low-pass filter, a high-pass filter, a band-pass filter having a gain function shown in FIG. 8A, a comb filter having a gain function shown in FIG. A process using is prepared. The combination of filter processes includes any combination of these filter processes. The filter characteristics include, for example, a cut-off frequency for a low-pass filter, a high-pass filter, and a band-pass filter, and a cut-off frequency, a bandwidth, and a band interval for a comb filter.

  As described above, the filter process is a so-called digital signal process, and the frequency component of the signal waveform (referred to as a spatial frequency in the case of an optical image, but the signal waveform process itself is the same as shown in FIG. (Represented as a frequency) is limited, and the frequency characteristic is changed. For example, in the case of a low-pass filter, the signal waveform has only a low frequency component, and a high frequency component is deleted.

  In general, an image sensor (generally referred to as a sensor that can obtain a two-dimensional image such as a CCD) has a number of pixels that are two-dimensionally aligned, samples an alignment mark image formed by an optical system at a predetermined pixel pitch, A two-dimensional image signal can be obtained. The intensity distribution of the optical alignment mark image can be obtained from the intensity distribution of the image obtained in this manner, and further, a process of changing the frequency characteristic by performing the filtering process can be applied.

  Although the comb filter shown in FIG. 8B is shown in FIG. 8B as an example, characteristics having eight comb shapes are shown. (Gain) can be changed individually, and various settings including zero height can be made. 8A and 8B are shown in a rectangular shape for simplification, but are not shown in the figure, but are generally optimized for signal processing and are more square than the rectangular shape. It is possible to adopt a more optimal shape, such as adopting a rounded shape.

  In the next step 304, the analysis conditions set in step 302 with respect to the detection signals of the alignment marks for all the detection conditions obtained by the alignment mark detection process, that is, the filter process having the set filter characteristics or a combination thereof The process using is performed. The processed detection signal is stored in a storage device (not shown) in association with the analysis condition.

  FIGS. 8C and 8D show examples of detection signals before and after application of processing using a combination of filter processing suitable for noise removal. As shown in FIG. 8D, main frequency components are extracted from the detection signal before processing shown in FIG. 8C by a band pass filter. Thus, the filtering process may be effective.

  In the next step 306, it is determined whether or not the detection signal filtering process has been completed under all scheduled analysis conditions. If this determination is negative, the process returns to step 302 to change to the next analysis condition, and in step 304, the detection signal is filtered under the changed analysis condition. Then, the loop processing of steps 302, 304, and 306 is repeated until the determination in step 306 is affirmed. When the determination in step 306 is affirmed, the process proceeds to the next step 308.

  In step 308, the feature amount of the detection signal is extracted using a predetermined signal processing algorithm. Here, an example of the feature amount will be described. Corresponding to each line pattern in the alignment mark, for example, one signal (referred to as a piece signal) having a waveform as shown in FIG. 9A is obtained. An inflection point of the signal waveform of the piece signal indicated by a broken line in FIG. 9A can be extracted as a feature point using a predetermined signal processing algorithm. Therefore, for each piece signal, the position of the inflection point of the signal waveform and the signal intensity and inclination (differential value) at the inflection point can be obtained as a feature amount by the signal processing algorithm. The inflection points are points where the waveform obtained by differentiating the signal waveform becomes maximum and minimum within the piece signal. In FIG. 9A, four inflection points are extracted. Similarly, an extreme point of the signal waveform of the piece signal indicated by a broken line in FIG. 9B can be extracted as a feature point by a signal processing algorithm. Therefore, for each piece signal, the position of the extreme value point of the signal waveform, the signal intensity at the extreme value point, and the differential value at the extreme value point can be obtained as the feature amount. The extreme point is a point where the signal waveform becomes maximum and minimum. Accordingly, the obtained differential value at the extreme point is almost zero, and it can be confirmed that the extreme point is obtained.

  Note that the position of the extreme point can also be obtained as the position of the midpoint between two inflection points that sandwich the extreme point. As a tendency of the optical image, it is assumed that the extreme point is not one point between two inflection points sandwiching the extreme point (when there is a small undulation component in the signal waveform between the two inflection points) However, in such a case, the present method is effective in assuming that the midpoint of two inflection points is an extreme point. The slope (differential value) of the signal waveform at the extreme point obtained at the same time can be effectively utilized because it takes a non-zero value in the case of this method. If the extreme point is difficult to determine due to noise or the like, the extreme point may be detected after moving average processing (smoothing) with a so-called predetermined width.

  Further, the maximum value and the minimum value of the piece signal indicated by the broken line in FIG. 10A can be extracted as the feature amount. Unlike the feature value at the extreme point described above, it is possible to simply extract the waveform maximum value and the minimum value of the piece signal as the feature value of the piece signal. Since it is assumed that the signal waveform has a different shape for each piece signal, this feature quantity that does not depend on extreme points is extracted. Further, it is possible to extract the center of the waveform of the piece signal indicated by the broken line in FIG. 10B and the intermediate point between the waveforms of the piece signals adjacent to each other. Can be obtained as a feature amount. Here, it is necessary that the waveform center of the piece signal has already been calculated. However, as an existing method for obtaining the waveform center of the piece signal, the signal waveform of the piece signal section is extracted and the left and right sides of the signal waveform are inverted. By creating a signal waveform and shifting the position with respect to the original signal and checking the correlation of the waveform (so-called convolution processing), a method with the position having the highest correlation as the center position can be used. Also, the alignment mark center can be obtained by averaging the waveform center coordinates of each piece signal calculated in the same manner.

  The position information of the feature amount including the inflection point position and / or the extreme value position described above is relative coordinates (for example, each piece signal) with the waveform center point of each piece signal and / or the center point of the alignment mark as the origin. It is possible to manage each piece by using the waveform center of each piece as the origin, or coordinates with the alignment mark center point as the origin. The extracted feature points and the corresponding feature values are stored in a storage device (not shown) in association with the analysis conditions and the alignment marks (mark lines constituting the analysis marks).

  Note that the extracted feature points (and feature amounts) are managed in groups for each corresponding alignment mark. In the group, as shown in FIG. 11A, for example, feature points are labeled according to their positions (arrangements). In this example, numbers from 1 to m are attached to m feature points as labels. However, the position where the feature point appears may deviate from the original position or the expected feature point may not be detected due to the difference in shape and deformation of the signal waveform due to the structural difference of the alignment mark, or a detection error. For example, FIG. 11B shows an example in which m groups are configured by k signal waveforms by a plurality of alignment marks. FIG. 11B shows a state in which the p-2th feature point belonging to the second detection signal and the p + 1th feature point belonging to the kth detection signal are not detected among the detection signals. ing.

  A method for performing such grouping will be described with reference to FIG. As described above, main controller 28 stores the mark center positions corresponding to the waveforms of the plurality of alignment marks, and each feature point can be managed in relative position coordinates with the mark center as the origin. Therefore, the position of the feature point included in each waveform can be compared with each other by matching the center position (mark center) of the feature point of each waveform. Here, the positions of corresponding feature points are compared within a comparison range having a predetermined width. That is, the position is within the comparison range, the signs of the slopes (differential values) are the same, and if a plurality of candidates are included, it is confirmed that the position is closest and is determined to be a corresponding feature point. . If the corresponding feature point cannot be confirmed, the labels of other feature points in the group including the feature point are reassigned.

  Returning to FIG. 7, in step 310, a determination amount related to the shape of the signal waveform of the detection signal is obtained using the feature amount obtained in step 308. As shown in FIG. 12, the main controller 28 determines the amplitude (= maximum intensity-minimum intensity) and contrast (= (maximum intensity-minimum) from the maximum and minimum intensity of the detection signal in the piece signal, as shown in FIG. Intensity) / (maximum intensity + minimum intensity)) and the maximum slope of the detection signal at the inflection point, the maximum edge slope (simply called edge slope) is obtained.

  In the next step 312, the measurement reproducibility of the alignment mark detection result is evaluated using the determination amount related to the shape obtained in step 310. As shown in FIG. 13, the main controller 28 evaluates the measurement reproducibility using the amplitude, contrast, and edge slope, which are determination amounts related to the shape. The main controller 28 ranks and evaluates each determination amount in, for example, four stages (A to D). In the example shown in FIG. 13, the amplitude is evaluated as rank A, the contrast is evaluated as rank C, and the edge slope is evaluated as rank B. The main controller 28 evaluates the measurement reproducibility by combining these evaluations. As an example of the evaluation method, there is a method of having a database that can be referred to so that one score corresponds to the relationship between the rank of each determination amount and the given score when three kinds of determination amount ranks are determined. In FIG. 13, an overall score (score) 600 is evaluated with reference to a database prepared in advance for ranks A, B, C, and D.

  The ranking (evaluation) of each determination amount is determined so that an appropriate total point is determined based on, for example, the correlation between each determination amount and measurement reproducibility shown in FIG. Here, the correlation is obtained in advance by an optical teaching simulation or the like. Here, the optical teaching simulation related to measurement reproducibility is to generate an optical image of an alignment mark having various assumed structures under various optical setting conditions (illumination conditions and / or imaging conditions) and focus setting conditions, A signal waveform measured by the image sensor by A / D conversion is obtained in a pseudo manner, a predetermined electrical noise is added, and various feature amounts and / or judgments are made in a signal processing similar to an actual device. It means the process of calculating and evaluating the quantity. By organizing the results of the optical teaching simulation, a database for referring to the score can be created. A range of determination amounts that gives low measurement reproducibility from high measurement reproducibility is sequentially classified into ranks A to D. The determination amount is evaluated using a rank corresponding to the range in which the value is located.

  As is clear from FIG. 14A, the correlation with the measurement reproducibility varies greatly depending on the determination amount. For example, contrast has a weak correlation with measurement reproducibility, and each of amplitude and edge slope has a strong correlation with measurement reproducibility. Therefore, it is preferable to comprehensively evaluate the measurement reproducibility in consideration of the degree of correlation with the measurement reproducibility of each determination amount. In the optical teaching simulation described above, a final database is created in consideration of such points.

It is also possible to define a new determination amount according to the main component of noise included in the detection signal. For example, when the shot noise component (electrical noise that increases in accordance with the amount of received light) by the image sensor in the alignment detection system is dominant, as shown in FIG. 14A, the product of amplitude, contrast, and edge slope is used. The correlation between each of the square roots and the reciprocal of the measurement reproducibility becomes stronger. Therefore, for example, the reciprocal σ _s of the square root of the product of contrast and edge slope can be employed as a new determination amount. Furthermore, when the black level noise component (so-called white noise, which is an electric noise component that is always generated) cannot be ignored, the relationship between the edge slope and the reciprocal of the measurement reproducibility is also ignored as shown in FIG. become unable. Therefore, the square root σ (= √ (α ² · σ _s ² + β ² · σ _b ² )) of the square sum of the reciprocals σ _b and σ _s of the edge slope can be used as a new determination amount. However, α and β (α ² + β ² = 1) are weighting factors, and are obtained in advance by optical teaching simulation or the like. If these determination amounts are used, it is considered that the measurement reproducibility can be predicted fairly accurately from the result of the optical teaching simulation.

  It is desirable that the results of optical teaching simulations be accurate. However, in reality, the alignment mark for evaluation is measured using an actual alignment detection system, and the measurement reproducibility is evaluated. It is desirable to correct the database itself and / or the determination amount to check the measurement reproducibility score value itself and its error.

  Returning to FIG. 7, in step 314, a score is calculated based on the evaluation (result) obtained in step 312. When performing filter processing, a score is calculated for each filter processing.

  In the next step 316, the score calculated in step 314 is stored in the storage device, and then the detection signal feature amount extraction and determination processing subroutine processing ends. As shown in FIG. 15, the main controller 28 displays determination amounts such as amplitude and contrast for each illumination condition and imaging condition. However, the display is not limited to the value of the determination amount, and may be displayed in a form such as a rank based on the determination amount. Further, the main controller 28 displays the measurement reproducibility score stored for each illumination condition and imaging condition, as shown in FIG. The measurement reproducibility is not limited to the above-described score, and may be displayed in a form such as a rank based on the score. Here, it is preferable to highlight and display a recommended condition based on the value or rank of the determination amount. Thereby, an operator (or engineer) or the like can select illumination conditions and imaging conditions that are optimal or suitable for device manufacture. The main controller 28 stores the information on the result of optical teaching displayed on the monitor in the storage device in association with the illumination condition and the imaging condition. When a plurality of filter processes are performed, the display of FIG. 15 and FIG. 16 may be increased and displayed by the number of filter processes so that the effect of the filter process can be confirmed.

  When the subroutine of the feature amount extraction and determination process of the detection signal is completed, the process returns to step 112 of the main routine, and waits for the determination of optimal illumination conditions and imaging conditions. Then, when the operator selects or determines the optimum illumination condition and imaging condition through a pointing device such as a keyboard, a mouse, etc. while viewing the display on the monitor, and inputs the determined illumination condition and determination, the determination in step 112 is performed. Is affirmed and the routine proceeds to step 114.

  In step 114, a recipe file for alignment measurement (EGA measurement) is created using the optimal illumination conditions and imaging conditions selected (or determined). At this time, if filtering is being performed, filtering may be set. Preparation of a recipe file can be easily performed by preparing a recipe file in which the numerical value of the condition to be optimized is blanked in advance and applying the optimum condition determined to this blank. As a result, at the time of actual wafer exposure processing, the main controller 28 can read out the created recipe file and perform fine alignment of the wafer according to the contents.

  In this embodiment, the measurement reproducibility is determined. As described above, at least the items that generally need to be determined are as follows. Measurement reproducibility determination, 2. 2. TIS random error factor determination; It is considered that there are three points: WIS random error factor determination. 2. 3. Of course, it is possible to make a judgment based on a predetermined feature amount to obtain a score, and make a comprehensive judgment by comparing with the score of the measurement reproducibility obtained in the present embodiment.

  Instead of the processing of the operator in step 112, the main controller 28 may select the best illumination condition and imaging condition based on the result of optical teaching. In this case, the main controller 28 selects an illumination condition and an imaging condition that give the best measurement reproducibility so that the overlay accuracy is the best. In step 114, a recipe file (process program) for alignment measurement (EGA measurement) is created using the selected optimization condition. This makes it possible to create an EGA measurement recipe file that includes optimum measurement conditions corresponding to the EGA mark almost completely automatically.

  Until now, in order to avoid making the explanation more complicated than necessary, the creation of the recipe file for EGA measurement has been described. However, the creation of the recipe file for search alignment measurement also involves the creation of the recipe file for EGA measurement. The same can be done. In this case, the contents of processing in setting or changing the illumination wavelength of the alignment detection system AS (step 202) and setting or changing the illumination method and imaging conditions (step 204) are somewhat different, and the detection target is also different. It becomes a search mark instead of the EGA mark. However, except for such differences, a recipe file for search alignment measurement can be created in accordance with substantially the same processing algorithm as in steps 102 to 114, steps 202 to 214, and steps 302 to 316 described above.

  As described above, according to the exposure apparatus 100 according to the present embodiment, the main controller 28 uses the alignment detection system AS to form alignment marks (EGA marks or search marks) formed on the wafer W with a plurality of illuminations. A determination amount related to the shape of the waveform of the detection signal is obtained using a signal processing algorithm (steps 302 to 310), and the reproducibility of detection results of a plurality of marks is based on the determination amount. Evaluation (step 312) makes it possible to select (or determine) optimal illumination conditions and imaging conditions that provide good measurement reproducibility based on the evaluation results (step 314). Therefore, it is possible to perform illumination conditions and imaging conditions optimal for measurement in a short time and / or efficiently. Here, in place of the processing of the operator in step 112, when the main controller 28 selects the best illumination condition and imaging condition based on the result of optical teaching, the optimum illumination condition and It is not necessary to select imaging conditions. In addition, when performing filter processing, it is possible to set optimal filter processing together. If measurement reproducibility is insufficient even under optimal illumination conditions and imaging conditions, measurement processing is performed multiple times for one mark (or signal integration is performed multiple times) to reduce the effect of noise. It is also possible to improve reproducibility.

  Further, according to the exposure apparatus 100 of the present embodiment, the main controller 28 determines the optimized detection condition as a corresponding condition in a process program (recipe file) that defines the alignment measurement processing procedure. Then, an alignment measurement recipe file is created. At the time of wafer processing, the main controller 28 reads the created recipe file, performs wafer search alignment reliably and accurately based on the recipe file, and performs wafer fine alignment (EGA). It becomes possible to carry out with high accuracy.

  Further, in the exposure apparatus 100 of the present embodiment, the fine alignment of the wafer can be performed with high accuracy as described above. Therefore, the pattern of the reticle R is applied on the wafer W during the exposure based on the result of the fine alignment. It is possible to accurately superimpose and transfer to each shot area.

  In the above-described embodiment, when the main controller 28 selects the best illumination condition and imaging condition based on the result of optical teaching, two of “superimposition accuracy priority” and “throughput priority” are selected. The best condition may be selected according to one of the modes. As described above, if the measurement reproducibility is insufficient even when the filter processing is performed under the optimal illumination conditions and imaging conditions, the measurement processing is performed multiple times for one mark (or the signal is divided into multiple times). It is possible to improve the measurement reproducibility by reducing the influence of noise. However, the processing time is increased accordingly, and the throughput is deteriorated. In such a case, in the overlay accuracy priority mode, the number of times of measurement for one mark is increased until sufficient accuracy is obtained, but in the throughput priority mode, measurement is performed with emphasis on throughput at the expense of some measurement reproducibility. And

  It is also possible to prioritize illumination conditions and imaging conditions. For example, the standard condition is a broad frequency of use with respect to the illumination wavelength. Alternatively, the detection conditions are optimized only for illumination conditions and imaging conditions with high priority. For example, the highest priority is the broad frequency of use, followed by green, orange, and red, which are expected to improve measurement reproducibility, and finally cyan, near infrared, which are mainly used in special situations. Attached. For the illumination method and imaging conditions, the best focus position is given the highest priority with the bright field illumination that is frequently used, the optimum focus position with the bright field illumination suitable for the detection of the mark with the next largest step, and finally the alignment detection. Priorities are assigned in the order of the best focus position in dark field illumination in which the occurrence of detection errors (TIS: Tool Induced Shift) derived from the system is suppressed. Alternatively, when the TIS in the bright field illumination is small and the measurement reproducibility in the dark field illumination is high, the best focus position is prioritized in the bright field illumination. Also, when the TIS in bright field illumination is small and the measurement reproducibility in dark field illumination is low, priority is given to the optimum focus position in bright field illumination. When the TIS in bright field illumination is large and the measurement reproducibility in dark field illumination is high, the best focus position is prioritized in dark field illumination. When the TIS in the bright field illumination is large and the measurement reproducibility in the dark field illumination is low, the judgment is left to the operator (or engineer).

  Further, in the above-described embodiment, when the recipe file is created, an example of the detection condition optimization method according to the present invention has been described. However, the present invention is not limited thereto, and in actual wafer processing, Using the pilot wafer or the lot head wafer, the processing excluding the above-described step 114, that is, the alignment mark detection condition optimization processing may be performed. Also in this case, the detection conditions of the EGA mark (and search mark) are optimized, and by following the optimized conditions, it is possible to perform fine alignment (EGA) (and search alignment) of the wafer with high accuracy. Become.

  In the above embodiment, the case where the present invention is applied to an exposure apparatus including a single alignment detection system AS has been described. However, for example, a plurality of alignments disclosed in US Patent Application Publication No. 2008/0088843 and the like. The present invention can also be suitably applied to an exposure apparatus provided with a detection system. In this case, since each of the plurality of alignment detection systems has different optical characteristics such as aberration, it is necessary to optimize the detection conditions for each detection system. However, in order to simplify this procedure, the detection condition is optimized only for one of the plurality of alignment detection systems (referred to as a primary alignment system). Then, the same alignment mark is detected using each of the plurality of alignment detection systems, and the difference in TIS for other alignment systems based on the primary alignment system is obtained from the result. By correcting the detection results of the other alignment systems using this difference, the optimum detection conditions obtained for the primary alignment system can be shared.

  In the above embodiment, the exposure apparatus creates a recipe file for alignment measurement and optimizes the detection conditions at that time. However, the present invention is not limited to this, and the detection condition optimization method and program of the present invention are not limited thereto. The creation method may be performed by an apparatus (mark detection apparatus) including an image processing type alignment sensor (mark detection system) other than the exposure apparatus, for example, an overlay measuring machine. In this case, basically, an effect equivalent to that of the above embodiment can be obtained, and the mark detection conditions can be optimized and a recipe file for alignment measurement can be created with these apparatuses.

  Further, in the above embodiment, for the sake of simplicity of explanation, the main control device 28 performs all processes related to alignment measurement including optimization of detection conditions, creation of a recipe file, and the like. The various processes performed by may be performed by sharing a plurality of hardware. For example, the processing of each step shown in the flowchart of FIG. 2 described above may be performed by appropriately sharing a plurality of microcomputers.

In the above embodiment, an ultraviolet light source such as a KrF excimer laser (output wavelength 248 nm), a pulsed laser light source in the vacuum ultraviolet region such as an ArF excimer laser, or the like is used as the light source. Of course, other vacuum ultraviolet light sources such as F ₂ laser or Ar ₂ laser (output wavelength 126 nm) may be used. Further, for example, not only laser light output from each of the above light sources as vacuum ultraviolet light, but also single wavelength laser light in the infrared region or visible region oscillated from a DFB semiconductor laser or fiber laser, for example, erbium (Er) A harmonic which is amplified by a fiber amplifier doped with (or both erbium and ytterbium (Yb)) and wavelength-converted into ultraviolet light using a nonlinear optical crystal may be used.

  Further, an exposure apparatus that uses EUV light, X-rays, or charged particle beams such as an electron beam or an ion beam as illumination light IL, an exposure apparatus that does not use a projection optical system, such as a proximity exposure apparatus, a mirror projection aligner, and The present invention may also be applied to an immersion type exposure apparatus and the like disclosed in US Patent Application Publication No. 2005/0259234 and the like in which a liquid is filled between the projection optical system and the wafer.

  In each of the above-described embodiments, a light-transmitting mask in which a predetermined light-shielding pattern (or phase pattern / dimming pattern) is formed on a light-transmitting substrate, or a predetermined reflecting pattern on a light-reflecting substrate. However, the present invention is not limited to these. For example, instead of such a mask, an electronic mask (which is a kind of optical system) that forms a transmission pattern, a reflection pattern, or a light emission pattern based on electronic data of a pattern to be exposed may be used. Such an electronic mask is disclosed, for example, in US Pat. No. 6,778,257. Note that the above-described electronic mask is a concept including both a non-light-emitting image display element and a self-light-emitting image display element.

  Further, for example, the present invention can also be applied to an exposure apparatus that exposes a substrate with interference fringes caused by interference of a plurality of light beams, which is called two-beam interference exposure. Such an exposure method and exposure apparatus are disclosed in, for example, WO 01/035168.

  In the above embodiment, the case where the present invention is applied to a scanning exposure apparatus such as the step-and-scan method has been described. However, the present invention is not limited to this, and the step-and-repeat method or the step-and-stitch method is used. The present invention can also be suitably applied to the projection exposure apparatus.

  The present invention is not limited to an exposure apparatus for manufacturing a semiconductor element, but is used for manufacturing a display including a liquid crystal display element and the like, an exposure apparatus for transferring a device pattern onto a glass plate, and a thin film magnetic head. The present invention can also be applied to an exposure apparatus that transfers a device pattern onto a ceramic wafer, and an exposure apparatus that is used for manufacturing an imaging device (CCD or the like), micromachine, organic EL, DNA chip, and the like. Further, in order to manufacture reticles or masks used in not only microdevices such as semiconductor elements but also light exposure apparatuses, EUV exposure apparatuses, X-ray exposure apparatuses, electron beam exposure apparatuses, etc., glass substrates or silicon wafers, etc. The present invention can also be applied to an exposure apparatus that transfers a circuit pattern.

  Heretofore, an exposure apparatus for forming a pattern on a substrate has been described. However, a method for forming a pattern on a substrate by a scanning operation is not limited to the exposure apparatus, and for example, US Pat. No. 6,973,710. It is also possible to use an element manufacturing apparatus provided with an ink jet type functional liquid application device similar to the ink jet head group disclosed in the specification.

  The inkjet head group disclosed in the above-mentioned US patent specification is a substrate (for example, PET, glass, silicon, paper, etc.) by discharging a predetermined functional liquid (metal-containing liquid, photosensitive material, etc.) from a nozzle (discharge port). A plurality of inkjet heads to be applied to What is necessary is just to prepare a functional liquid provision apparatus like this inkjet head group, and to use it for the production | generation of a pattern. In the element manufacturing apparatus provided with this functional liquid application apparatus, the substrate may be fixed and the functional liquid application apparatus may be scanned in the scanning direction, or the substrate and the functional liquid application apparatus may be reversed. You may scan.

  For semiconductor devices, the step of designing the function and performance of the device, the step of manufacturing a reticle based on this design step, the step of manufacturing a wafer from a silicon material, and transferring the reticle pattern to the wafer by the exposure apparatus of the above-described embodiment And a device assembly step (including a dicing process, a bonding process, and a packaging process), an inspection step, and the like. Therefore, the semiconductor device can be manufactured with high productivity.

  As described above, the detection condition optimization method of the present invention is suitable for optimizing the alignment mark detection condition. The program creation method of the present invention is suitable for creating an alignment measurement recipe. Further, the exposure apparatus of the present invention is suitable for overlapping a pattern on a substrate.

DESCRIPTION OF SYMBOLS 16 ... Alignment control apparatus, 28 ... Main control apparatus, 100 ... Exposure apparatus, _Sn ... Shot area | region, 32 ... Fine alignment mark, 33 ... Search alignment mark, AS ... Alignment detection system, W ... Wafer.

Claims (30)

A detection condition optimization method for optimizing a detection condition of a mark for alignment formed on a substrate,
Detecting a plurality of said marks formed on a substrate under a plurality of illumination conditions and imaging conditions using a mark detection system;
A detection signal from the mark detection system is analyzed using a predetermined signal processing algorithm, a determination amount related to a waveform shape of the detection signal is obtained, and reproducibility of detection results of the plurality of marks is determined based on the determination amount. Evaluating and optimizing the plurality of illumination conditions and imaging conditions based on the evaluation results;
A detection condition optimization method including: The predetermined signal processing algorithm includes a plurality of filter processes,
The detection condition optimization method according to claim 1, wherein in the optimization, processing conditions of the plurality of filter processes are further optimized.   The detection condition optimization method according to claim 2, wherein the processing conditions for the plurality of filter processes include an optimal combination of the plurality of filter processes for improving reproducibility of detection results of the plurality of marks.   The detection condition optimization method according to claim 2 or 3, wherein the processing conditions of the plurality of filter processes include respective filter characteristics of the plurality of filters.   The detection condition optimization according to any one of claims 2 to 4, wherein the plurality of filter processes include at least one of a low-pass filter, a high-pass filter, a band-pass filter, and a comb filter. Method.   In the optimization, from the detection signals of the plurality of marks, the position of at least one of the extreme point and the inflection point of the waveform of the detection signal, the intensity and the slope of the detection signal at the position The detection condition optimization method according to claim 1, wherein a feature amount including at least one is extracted, and a determination amount related to the shape is obtained using the feature amount.   The detection condition optimization method according to claim 1, wherein the determination amount related to the shape includes at least one of amplitude, contrast, and edge slope.   The detection condition optimization method according to any one of claims 1 to 7, wherein the determination amount related to the shape includes a function of an arbitrary combination of amplitude, contrast, and edge slope. Prior to the optimization, a correspondence relationship between the determination amount and the reproducibility is obtained,
The detection condition optimization method according to any one of claims 1 to 8, wherein in the optimization, the reproducibility is further evaluated using the correspondence relationship.   The detection condition optimization method according to any one of claims 1 to 9, wherein in the optimization, the determination amount is individually evaluated, and the reproducibility is evaluated by integrating the evaluation results.   The detection condition optimization method according to claim 1, wherein the plurality of illumination conditions include an illumination method or an illumination wavelength of the mark detection system.   The detection condition optimization method according to claim 11, wherein the mark detection system illumination method includes an illumination numerical aperture of the mark detection system.   The detection condition optimization method according to claim 1, wherein the plurality of imaging conditions include a focus position or an imaging numerical aperture of the mark detection system.   The detection condition optimization method according to any one of claims 1 to 13, wherein the plurality of marks include a mark for roughly aligning the substrate and a mark for precisely aligning the substrate.   A program for optimizing detection conditions for detection signals of the plurality of marks by using the detection condition optimization method according to any one of claims 1 to 14, and creating a program for determining an alignment procedure of the substrate How to make. An exposure apparatus that forms a plurality of patterns by superimposing on a substrate,
A mark detection system for detecting a mark for alignment formed on the substrate;
A detection signal obtained by detecting a plurality of the marks under a plurality of illumination conditions and imaging conditions using the mark detection system is analyzed using a predetermined signal processing algorithm, and the waveform of the detection signal is related to An optimization device that obtains a determination amount, evaluates reproducibility of detection results of the plurality of marks based on the determination amount, and optimizes the plurality of illumination conditions and imaging conditions based on the evaluation results;
An exposure apparatus comprising: The predetermined signal processing algorithm includes a plurality of filter processes,
The exposure apparatus according to claim 16, wherein the optimization apparatus further optimizes processing conditions of the plurality of filter processes.   The exposure apparatus according to claim 17, wherein the processing conditions of the plurality of filter processes include an optimal combination of the plurality of filter processes that improves reproducibility of the detection results of the plurality of marks.   The exposure apparatus according to claim 17 or 18, wherein processing conditions of the plurality of filter processes include filter characteristics of the plurality of filters.   The exposure apparatus according to any one of claims 17 to 19, wherein the plurality of filter processes include at least one of a low-pass filter, a high-pass filter, a band-pass filter, and a comb filter.   The optimization device is configured to detect at least one of a position of at least one of an extreme point and an inflection point of a waveform of the detection signal, a strength and a slope of the detection signal at the position, from the detection signals of the plurality of marks. The exposure apparatus according to any one of claims 16 to 20, wherein a feature amount including one is extracted and a determination amount related to the shape is obtained using the feature amount.   The exposure apparatus according to any one of claims 16 to 21, wherein the determination amount related to the shape includes at least one of amplitude, contrast, and edge slope.   The exposure apparatus according to any one of claims 16 to 22, wherein the determination amount related to the shape includes a function of an arbitrary combination of amplitude, contrast, and edge slope.   The exposure apparatus according to any one of claims 16 to 23, wherein the optimization apparatus obtains a correspondence relationship between the determination amount and the reproducibility in advance, and evaluates the reproducibility using the correspondence relationship.   The exposure apparatus according to any one of claims 16 to 24, wherein the optimization apparatus evaluates the determination amount individually and evaluates the reproducibility by integrating the evaluation results.   The exposure apparatus according to any one of claims 16 to 25, wherein the plurality of illumination conditions include an illumination method or an illumination wavelength of the mark detection system.   27. The exposure apparatus according to claim 26, wherein the mark detection system illumination method includes an illumination numerical aperture of the mark detection system.   The exposure apparatus according to any one of claims 16 to 27, wherein the plurality of imaging conditions include a focus position or an imaging numerical aperture of the mark detection system.   The exposure apparatus according to any one of claims 16 to 28, wherein the plurality of marks include a mark for roughly aligning the substrate and a mark for precisely aligning the substrate. A mark detection system for detecting a mark for alignment formed on the substrate;
A detection signal obtained by detecting a plurality of the marks under a plurality of illumination conditions and imaging conditions using the mark detection system is analyzed using a predetermined signal processing algorithm, and the waveform of the detection signal is related to An optimization device that obtains a determination amount, evaluates reproducibility of detection results of the plurality of marks based on the determination amount, and optimizes the plurality of illumination conditions and imaging conditions based on the evaluation results;
A mark detection apparatus comprising: